text
stringlengths 1
6.27k
| id
int64 0
3.07M
| raw_id
stringlengths 2
9
| shard_id
int64 0
0
| num_shards
int64 16
16
|
|---|---|---|---|---|
in 6 months. As of September 2016, no obvious progress was observed. Discussion Collagenomas belong to connective tissue nevi (hamartomas), and are composed predominantly of collagen. They are divided into inherited and acquired. Inherited collagenomas are autosomal dominantly inherited, including familial cutaneous collagenoma and Shagreen patch of tuberous sclerosis. Acquired collagenomas include isolated collagenoma and eruptive collagenoma depending on the number of lesions [15]. Eruptive collagenoma, nevus anelasticus and papular elastorrhexis are closely related entities. Based on the similar clinical and histopathological features, some authors considered they represent a single disease spectrum [3,16,17]. They have common features in terms of peak age of onset, distribution of lesions, and a lack of history The tumor cells showed positive for vim (g), and negative for h-CALD (h, the smooth muscle of muscularis mucosae was positive), CD34 (i, the vascular endothelium was positive). The lesions were blue with Masson Trichrome stain (j, the smooth muscle of muscularis mucosae was stained red), red with PAS stain (k), and pink with Congo red stain (l) of trauma. Histologically, the lesion is composed of plump stellate tumor cells, sometimes with interspersed giant multinucleate cells. Romos et al. studied four cutaneous collagenomas using ultrastructural and immunohistochemical analysis. Their findings confirmed that the tumor cells were fibroblastic cells [18]. Collagen deposition was obvious in almost all of the tumors. Occasionally it was the main component of the tumor. In a research published in 2016, Seung et al. evaluated the status of collagen tissue in eruptive collagenoma, nevus anelasticus and papular elastorrhexis. The selected cases
| 800 |
9111869
| 0 | 16 |
were reclassified into three groups: normal collagen group; fine, dense collagen group; and thick, dense collagen group [19]. This study indicated that different kinds of collagen could be observed in collagenomas probably in accordance with different stages of the tumor. Clinically, all previous cases of eruptive collagenoma were represented with asymptomatic multiple cutaneous nodules. The nodules were skin-colored, dome shaped and usually less than 1 cm in diameter. They occurred mainly on the trunk and upper extremities [2]. The majority age of onset is the first two decades of life. However, there were some cases reported in the later years of life [4,6,20]. In our present case, the patient is an old female. Strangely, the lesions were distributed in the submucosa of whole esophagus and intestine. The features of immunohistochemistry were similar to those of eruptive collagenoma reported previously [18]. Masson Trichrome stain confirmed the masses were composed of fibroblastic cells and collagen. And there were no family history and associated disorders. All of these were in favor of the diagnosis of eruptive collagenoma. The lesions should be differentiated from some other tumors which could be both multinodular and collagenous. Amyloidosis is the first tumor that should be differentiated. It is characterized by abnormal extracellular deposition of specific protein and protein derivatives, which are arranged in a β-pleated sheet structure. There are usually no tumor cells in the lesions. And Congo red stain is red but not pink. These features can distinguish it from eruptive collagenoma. It should also be differentiated from fibromas and plexiform fibromyxoma.
| 801 |
9111869
| 0 | 16 |
Fibromas (fibrous histiocytomas) are ill-defined, characterized by a variable number of spindle and/or rounded cells. A variable admixture of inflammatory cells, coarse collagen bundles in haphazard array are present. Plexiform fibromyxoma (plexiform angiomyxoid myofibroblastic tumor) is a rare benign mesenchymal tumor. The plexiform growth of bland spindle cells in a richly vascularized fibromyxoid stroma is distinctive. GIST is another tumor that should be distinguished. It is the most common primary mesenchymal tumor of the gastrointestinal tract and has a broad morphological spectrum. The tumor cells are usually positive for CD117 and CD34. So far, the pathogenesis of eruptive collagenoma was not clear. Uitto et al. commented that collagen accumulation is probably associated with a reduced collagenase [21]. Some other investigators considered that hormone may play a role in this disease [5,22]. Our patient is with a slightly abnormal level of IgG, IgA, complement C3 and complement C4. Whether it is associated with immune system disorder is worthy of further studying. Conclusion We describe a case of eruptive collagenoma occurring in the submucosa of esophagus and intestine, which have never been reported with eruptive collagenoma. We estimate that there are some changes of collagen in the submucosa of gastrointestinal tract that are similar to those on the skin.
| 802 |
9111869
| 0 | 16 |
Hysteresis and metastability of Bose-Einstein condensed clouds of atoms confined in ring potentials We consider a Bose-Einstein condensed cloud of atoms which rotate in a toroidal/annular potential. Assuming one-dimensional motion, we evaluate the critical frequencies associated with the effect of hysteresis and the critical coupling for stability of the persistent currents. We perform these calculations using both the mean-field approximation and the method of numerical diagonalization of the many-body Hamiltonian which includes corrections due to the finiteness of the atom number. Recently, the phenomenon of hysteresis has also been investigated in an annular potential [9]. In this experiment a Bose-Einstein condensate of sodium atoms that was initially at rest was stirred, and as the rotational frequency of the stirring potential increased, the cloud was observed to make a transition to a state with one unit of circulation at a critical frequency, Ω 1 . On the other hand, in the reverse process (i.e., starting with the gas having one unit of circulation and decreasing the frequency of the stirrer) the system was observed to return to the state with zero circulation at a different critical frequency, Ω 2 , which is a clear indication of hysteresis. Motivated by the above experiments, we consider here the phenomenon of hysteresis in a Bose-Einstein condensed gas of atoms confined in a ring potential [10][11][12][13][14][15] as well as the stability of persistent currents [16,17]. One of the main results of our study is the effect of the finiteness of the atom number on the phenomenon of hysteresis and on the
| 803 |
118447200
| 0 | 16 |
stability of the persistent currents. To attack this problem we use the method of diagonalization of the many-body Hamiltonian. Contrary to the mean-field approximation -which makes the implicit assumption of a large particle number -the diagonalization approach includes corrections due to a finite number of atoms. In addition, it avoids the assumption of a simple product state for the many-body wavefunction that is central to the mean-field approach. As a result, this approach captures correlations that are built when the atom number is very low or when the diluteness condition is violated. We stress that in various recent experiments it has be-come possible to trap and detect very small numbers of atoms, which can even be of order unity see, e.g., Ref. [18]. Indeed, there appears to be a more general tendency in the field of cold atoms to move towards the study of small systems. Interestingly, the vast majority of the theoretical studies which have been performed on the superfluid properties of cold atomic gases and on the phenomenon of hysteresis assume the opposite limit of large particle numbers, since they are based on the mean-field Gross-Pitaevskii approximation. As a result, very little is known about the effect of the finiteness of systems with a small number of atoms. In the following we first present our model in Sec. II and comment on the phenomena of hysteresis and of metastability. Then, we evaluate in Sec. III the critical frequencies associated with the phenomenon of hysteresis within the mean-field approximation. In Sec. IV we go beyond
| 804 |
118447200
| 0 | 16 |
the mean-field approximation to consider corrections of order 1/N (and lower) due to the finiteness of the atom number N . In Sec. V we investigate the same question regarding the critical coupling for metastability and the matrix element for the decay rate of persistent currents in a small system. In Sec. VI we make contact with recent experiments on the phenomenon of hysteresis and of metastability, and finally in Sec. VII we present our conclusions. II. MODEL AND GENERAL CONSIDERATIONS In the present study we assume one-dimensional motion of bosonic atoms under periodic boundary conditions, as in a ring potential. This model is expected to be valid in an annular/toroidal trap as long as the interaction energy is much smaller than the energy of the trapping potential in the transverse direction. If c m and c † m are annihilation and creation operators of an atom with angular momentum mh, the Hamiltonian has the form Here M is the atom mass, R is the mean radius of the torus/annulus, S is its cross section (in the transverse direction), with R ≫ √ S, and U = 2h 2 a/(M RS) is the matrix element for elastic s-wave atom-atom collisions, with a scattering length a. In analysing the phenomenon of hysteresis and of the metastability of superflow [16,17], the main feature to be considered is the dispersion relation [10][11][12][13][14][15], i.e., the energy of the system as a function of the angular momentum. Let E(ℓ) denote the total energy where ℓh ≡ Lh/N is the angular momentum
| 805 |
118447200
| 0 | 16 |
per atom and Lh is the total angular momentum. According to Bloch's theorem [10] E(ℓ) consists of a periodic part plus a quadratic part which comes from the motion of the center of mass. Thus, one needs consider only 0 ≤ L ≤ N (0 ≤ ℓ ≤ 1); the remainder of the spectrum follows trivially as a consequence of Bloch's theorem. In the absence of interactions E(ℓ) consists of straight lines. In the intervals q ≤ ℓ ≤ q + 1 where q is an integer, E(ℓ)/N = (2q + 1)|ℓ|h 2 /(2M R 2 ) (in what follows below we assume for simplicity that ℓ ≥ 0). Obviously, at the end points of each interval the first derivative of E(ℓ) is discontinuous. In the presence of repulsive/attractive interactions these discontinuities remain, while the curvature is negative/positive, respectively. Figure 1 shows a schematic picture of the dispersion relation E(ℓ) for the repulsive interactions which we consider here. Such a spectrum will give rise to hysteresis. If one goes to the rotating frame and considers E rot (ℓ)/N = E(ℓ)/N −ℓhΩ, there are competing local minima as the rotational frequency of the trap Ω is varied. These competing minima give rise to discontinuous transitions and thus to hysteresis. The two critical frequencies Ω 1 and Ω 2 of the hysteresis loop correspond to the value of the slope of the dispersion relation E(ℓ) for ℓ → 0 + and ℓ → 1 − , respectively. The effect of hysteresis is thus a generic feature of this
| 806 |
118447200
| 0 | 16 |
problem. On the other hand, for an effective attraction between the atoms, hysteresis is absent, since the curvature of E(ℓ) is positive, and thus there are no discontinuous transitions as the rotational frequency of the trap is varied. It is convenient to write (in the interval 0 ≤ ℓ ≤ 1) the total energy per particle E(ℓ)/N as [19] In the case of the non-interacting problem the first term on the right gives the kinetic energy, and e(ℓ) vanishes. Due to Bloch's theorem, e(ℓ) is symmetric around ℓ = 1/2 and a periodic function with a period equal to unity. In the dispersion relation we show the energy in the lab frame, E(ℓ), (middle curve) as well as in the rotating frame, Erot(ℓ)/N = E(ℓ)/N − ℓhΩ, for the two critical frequencies Ω1 and Ω2 for which the slope of Erot vanishes for ℓ → 0 + and ℓ → 1 − , respectively. The arrows in the upper plot indicate the instability that results from the disappearance of the energy barrier; in the lower plot they indicate the hysteresis loop as the rotational frequency varies. Therefore, it is crucial to determine the value of ε. Interestingly, the differenceh(Ω 1 − Ω 2 ) is equal to 2ε. Furthermore, the sign of Ω 2 determines the stability of persistent currents. Specifically, the condition Ω 2 = 0 represents the critical value of the coupling for metastability of the currents, and metastability will be present if Ω 2 < 0. III. HYSTERESIS IN THE MEAN-FIELD APPROXIMATION We
| 807 |
118447200
| 0 | 16 |
begin with the mean-field approximation and consider the limit ℓ → 1 − . One can construct a Taylor-series expansion of the energy as a function of the small parameter 1 − ℓ. Since we are interested in the slope of the dispersion relation, we only need the linear term in the expansion for the energy. To get that, it suffices to consider only the dominant state in the order parameter Ψ, which is φ 1 , with φ m (θ) = e imθ / √ 2π as well as the neighbouring modes φ 0 and φ 2 . This is due to the fact that there is a cross term in the energy that comes from the scattering of two atoms with m = 1 resulting an atom with m = 0 and another on with m = 2. This term can be negative and thus lowers the energy [20]. Therefore, we write the order parameter as where the coefficients are real variational parameters and also |c 1 | is of order unity, while |c 0 | and |c 2 | are both of order 1 − ℓ. We stress that a completely analogous calculation holds for ℓ → 0, in which case one should We should also mention that one may work more generally with the three states φ 1−κ , φ 1 and φ 1+κ , with κ = 2, 3, . . ., however the fact that the kinetic energy of the states φ m scales as m 2 necessarily implies that
| 808 |
118447200
| 0 | 16 |
κ = 1. The coefficients appearing in Eq. (3) must satisfy the normalization condition, c 2 0 + c 2 1 + c 2 2 = 1, and the constraint of fixed angular momentum, c 2 1 + 2c 2 2 = ℓ or The expectation value of the energy per particle in the above state is where c 0 and c 2 have been assumed to have opposite signs in order to minimize the energy. Here, γ/2 = N U/(2ǫ) = 2N aR/S is the ratio between the interaction energy of the gas with a homogeneous density distribution and the kinetic energy ǫ ≡h 2 /(2M R 2 ). After linearisation, the above expression may also be written as the value of θ that minimizes the energy is θ 0 = (1/4) ln(2γ + 1). Therefore, the minimized energy is The derived value of ε is thus ε/ǫ = √ 2γ + 1 − 1 and therefore while where ω = ǫ/h. We note here that Ω 2 will vanish if γ = 3/2. This is the well-known result for the stability of persistent currents in a single-component gas, see, e.g., Ref. [21]. One can generalize the above results (using Bloch's theorem) in the interval q ≤ ℓ ≤ q + 1, where and From the last equation it follows trivially that the critical value of the coupling for stability of persistent currents (for ℓ = q + 1) is γ = (2q + 1)(2q + 3)/2, as Bloch's theorem implies. IV. HYSTERESIS BEYOND THE
| 809 |
118447200
| 0 | 16 |
MEAN-FIELD APPROXIMATION We now examine the same problem beyond the meanfield approximation. To do this, we use the method of diagonalization of the many-body Hamiltonian. To get some insight, we start with the truncated space containing the single-particle states φ 0 , φ 1 , and φ 2 [i.e., the states used in Eq. (3)]. The eigenstates may be written in the form where n = 0, 1, 2, . . . denotes the excited state with index n. Here the states |p are defined as |0 p , 1 N −2p , 2 p , where the notation |0 N0 , 1 N1 , 2 N2 indicates that N 0 atoms occupy the state φ 0 , etc. Clearly, the states |p are eigenstates of the number operator and of the angular momentum for a system of N atoms with angular momentum L = N . Again, one can work more generally with the three states φ 1−κ , φ 1 and φ 1+κ , with κ = 2, 3, . . ., however the corresponding problem becomes block diagonal, with the triplet of the states with κ = 1 giving the slope we are looking for [21]. One can diagonalize the Hamiltonian in this truncated space using the Bogoliubov transformation to obtain the eigenvalues E n (L), which are E n (L = N )/ǫ − γ(N − 1)/2 = N − (γ + 1) + 2γ + 1(1 + 2n). (12) Considering the states |p ′ = |0 p+1 , 1 p−2m ,
| 810 |
118447200
| 0 | 16 |
2 p−1 with N atoms and L = N − 2 units of angular momentum, one can follow the same procedure as before to find that From the lowest eigenvalues of each of the last two equations it follows that Ω 2 /ω = 2 − √ 2γ + 1, in agreement with the result of the mean-field approximation, Eq. (8). The approach considered above has assumed that N is ≫ 1, while the expectation value of m [in Eq. (11)] is of order unity. To find the finite-N corrections for the critical values of Ω 1 and Ω 2 , we have diagonalized the many-body Hamiltonian numerically without making any approximations beyond the truncation to some set of single-particle states φ m with −m max ≤ m ≤ m max . Figure 2 shows the result of such a calculation for N = 5 atoms, 0 ≤ L ≤ 10, γ = N U/ǫ = 0.5, and m max = 4, where we plot a few eigenvalues for each value of L. The dispersion relation satisfies Bloch's theorem. The fact that the form of this figure is the same as that of the schematic plot of Fig. 1 indicates the presence of hysteresis. We stress that for the small values of N that we consider here one can easily reach the Tonks-Girardeau limit. In this limit γ is at least of order N 2 . Thus, in order for the mean-field approximation to be valid, γ has to be much less than N 2 .
| 811 |
118447200
| 0 | 16 |
Having diagonalized the Hamiltonian, we extract the slope of the dispersion relation from the difference E 0 (L = 1) − E 0 (L = 0) to determine Ω 1 . Finally, by varying the atom number, 2 ≤ N ≤ 5 we find that Ω 1 can be approximated as for γ = 0.1. A subtle point in this calculation is the fact that the interactions strength increases with increasing N . This results in a greater depletion of the condensate. Thus, in order to extract the critical frequencies associated with the hysteresis, we keep γ fixed or equivalently allow U to scale like 1/N . In obtaining Eq. (14) m max was set equal to 5. Clearly, m max must be sufficiently large so that the fitting parameters have saturated. The differences in these parameters due to changing m max = 4 to m max = 5 are in the seventh, third, and second significant figures respectively. The value of the leading term is remarkably close to the value of √ 1 + 2γ ≈ 1.09544 found in Eq. (7), which is the asymptotic value of Ω 1 for N → ∞. Similar calculations for γ = 1 yield Although the leading term is still reasonably close to √ 1 + 2γ ≈ 1.73205, the agreement is materially worse. This is presumably because of the larger depletion of the condensate due to the stronger interaction. One general observation that emerges from the above analysis is that the effect of the finiteness of the system
| 812 |
118447200
| 0 | 16 |
and of the correlations, captured by the method of diagonalization, is to decrease the value of Ω 1 from its asymptotic value (and thus to increase the value of Ω 2 ). We comment on this observation in the following section. Last but not least, we mention that the value of the angular momentum for which the winding number of the order parameter changes is exactly ℓ = 1/2. In the equivalent language of solitary waves [23] the lowest-energy state with this value of the angular momentum corresponds to a "dark" solitary wave (i.e., a solitary wave with a node) which, although dark, still has a finite propagation velocity due to the finiteness of the ring [22,23]. Assuming without loss of generality that the center of the solitary wave is located at θ = π, the real part of the order parameter has a fixed sign. Its minimum value (at θ = π) vanishes as ℓ → (1/2) − . The imaginary part of the order parameter has sinusoidal behaviour and vanishes at θ = 0, π, and 2π. This necessarily implies that the net phase change is zero. On the other hand, for ℓ → (1/2) + , the minimum value of the real part of the order parameter, which remains θ = π, is negative and approaches zero from below. This tiny change in the minimum value of the real part of the order parameter from slightly positive to slightly negative is sufficient to change the winding number of the phase. We stress that
| 813 |
118447200
| 0 | 16 |
this tiny change can be described perturbatively and, although there is a violent rearrangement of the phase of the order parameter, this rearrangement can in no way prevent hysteresis. V. METASTABILITY OF PERSISTENT CURRENTS IN A SMALL SYSTEM The dispersion relation can develop an energy barrier for sufficiently strong and repulsive interatomic interactions which separates the state with L = N from the state with L = 0 [16]. While E 0 (L = N ) will always have a higher energy than E 0 (L = 0) [in fact, E 0 (L = N ) − E 0 (L = 0) = N ǫ], the state with L = N is then metastable. As a result, if the system is prepared in the state L = N , it will require an exponentially long time for the system to decay since this process must occur via quantum tunnelling. Furthermore, the energy and the angular momentum of the gas must be dissipated by small non-uniformities in the trapping potential. In this section we investigate two different questions. The first is the critical value of the coupling required for the system to develop an energy barrier with particular concern for finite-N effects. The second question is how the matrix element of a symmetry-breaking singleparticle operator ∆V , that can connect the two eigenstates of lowest energy, |L = N and |L = 0 , depends on the atom number N (for reasons that we explain below). Starting with the first question, according to Eq. (8) the critical
| 814 |
118447200
| 0 | 16 |
value of γ for the existence of a local minimum for ℓ → 1 − is γ cr = 3/2. This is an asymptotic result, which does not include finite-N corrections. To find these corrections, we choose a fixed value of the atom number N and identify the critical value of U , U cr , which gives a zero slope in the dispersion relation for ℓ → 1 − , i.e., E 0 (L = N ) = E 0 (L = N − 1). The result of this calculation is given in Fig. 3 where we plot the number of atoms on the x axis and the product N U cr ≡ γ cr on the y axis, for m max = 4. These results can be fit as γ cr ≈ 1.5106 + 0.6020/N + 8.2820/N 2 − 34.8262/N 3 +73.3879/N 4 . (16) The small deviation of the asymptotic value of γ cr in the above expression from the expected value of 3/2 is presumably due to the truncation, m max = 4, the limited number of atoms we have considered, N ≤ 10, and correlations which are absent in the calculation within the mean-field approximation. Interestingly, as seen from Fig. 3, the value of γ cr for a finite number of atoms is higher than 3/2. Since this is determined by the slope E 0 (L = N ) − E 0 (L = N − 1), we conclude that the correlations which are captured within the present approach (but are
| 815 |
118447200
| 0 | 16 |
absent within the mean-field approximation) lower the energy of the state with L = N − 1 more than that of the state with L = N . Thus, a higher value of γ is necessary to stabilize the currents in the state with L = N . The same mechanism which increases γ cr is also responsible for the decrease (increase) of Ω 1 (Ω 2 ) found in the previous section. We turn now to the second question regarding the decay rate of the persistent current. In order for the energy barrier (which develops for sufficiently strong interatomic interactions) to prevent the decay of the currents and render them metastable with an exponentially long decay time, the matrix element of any symmetry-breaking single-particle operator ∆V connecting the states |L = N and |L = 0 must be vanishingly small [24]. Otherwise the presence of the energy barrier becomes irrelevant and the currents will decay. To investigate this problem, we consider a singleparticle operator ∆V = V 0 N i=1 δ(θ i ), which is a sum of delta function potentials intended to mimic irregularities in the trap [12]. This potential breaks the axial symmetry of the Hamiltonian and induces transitions between the two states |L = N and |L = 0 . We thus evaluate the matrix element L = N |∆V |L = 0 , making use of the lowest-energy states |L = 0 and |L = N that we get from the diagonalization of the axially-symmetric Hamiltonian. Clearly the only terms which
| 816 |
118447200
| 0 | 16 |
give a nonzero contribution to this matrix element are those that raise the angular momentum by L = N units when acting on |L = 0 , In the absence of interactions, when all the atoms are in the single-particle state φ 1 and φ 0 , respectively, this matrix element vanishes for all N > 1. This is also the case in the mean-field approximation. To get a non-vanishing matrix element it is necessary to consider non-zero interactions that deplete the condensate and a finite number of atoms. Figure 4 shows the value of | L = N |∆V |L = 0 /V 0 | as function of N . Again, we keep γ = gN fixed for the reasons stated above. Here, we have chosen γ = gN = 0.1, while the states |L = 0 and |L = N have been evaluated for m max = 5. As seen from this plot, this matrix element shows an exponential decay as function of N . To get an understanding of this decay we recall that the operator ∆V excites atoms, increasing their angular momentum by N units. Furthermore, the amplitudes c m in the expression of Eq. (11) decay very rapidly with m, as seen in Fig. 5 for N = 50 atoms with a rate that does not depend on N . This is a more general result that also holds in more extended spaces. The fact that the amplitudes of the states contributing to |L = 0 and |L = N decrease
| 817 |
118447200
| 0 | 16 |
rapidly as one moves away from |0 N and |1 N along with the nature of ∆V , which induces singleparticle excitations by N units of angular momentum, combine to make this decay matrix element extremely sensitive to N . Thus, the main result of this section is, quite generally, that a combination of sufficiently strong interatomic interactions and a finite number of atoms enhances the size of the matrix element and thus reduces the timescale that is associated with the decay rate of the persistent currents. This result may be interesting to explore experimentally in small systems with an interaction whose strength can be tuned. VI. CONNECTION WITH THE EXPERIMENTS ON HYSTERESIS AND METASTABILITY In order for our assumption of one-dimensional motion to be valid, the interaction energy must be much smaller than the quantum of energy associated with the motion of the atoms in the transverse direction (or, equivalently, the coherence length must be much larger than the transverse dimensions of the annulus/torus). However, this assumption is violated under current typical conditions, and thus the motion is not quasi-one-dimensional. For example, in the experiment of Ref. [9], where 23 Na atoms were used, the chemical potential is µ/h ≈ 2π×1.7 kHz, while the frequencies of the annular-like trapping potential (in the transverse direction) are ω 1 ≈ 472 Hz and ω 2 ≈ 188 Hz. (As a result, it has been argued that vortex-antivortex pairs form in this experiment.) Thus, it is not possible to make neither a quantitative nor a qualitative comparison of
| 818 |
118447200
| 0 | 16 |
the present theory and the experiment of Ref. [9]. An investigation of this problem using a more realistic model is underway and will be described in a future publication. If one wants nonetheless to get an estimate for the critical frequencies of hysteresis for the parameters of Ref. [9] using the present theory, it follows for a radius of R ≈ 19.5 µm, that ω =h/(2M R 2 ) ≈ 3.6 Hz. Given that a ≈ 28Å, N ≈ 4 × 10 5 , and S = πa 1 a 2 with a i = h/(M ω i ), i.e., a 1 ≈ 2.42 µm and a 2 ≈ 3.83 µm, the dimensionless parameter γ = 2N aR/S has the value γ ≈ 1500.0. It then follows from Eqs. (7) and (8) that Ω 1 ≈ 197.2 Hz and Ω 2 ≈ −190.0 Hz. Clearly, these large frequencies (as compared to the observed frequencies, which are on the order of 10 Hz) are due to the very large value of γ, which is the ratio between the interaction energy of a homogeneous cloud with a density n 0 = N/(2πRS) and the kinetic energy associated with the motion in the ring,h 2 /(2M R 2 ). It is also interesting to make estimates for the case where the motion is quasi-one-dimensional. Consider, for example, the case where the experimental conditions are identical to those of Ref. [9] but where the number of atoms is reduced by, e.g., a factor of 4 × 10 4 to the
| 819 |
118447200
| 0 | 16 |
value N = 10. This would reduce the interaction energy to the extent that the conditions for one-dimensional motion would be fulfilled. This reduction in N would also reduce the value of γ to ≈ 0.04. The corresponding critical frequencies would become Ω 1 ≈ 3.7 Hz and Ω 2 ≈ 3.5 Hz. While the difference between Ω 1 and Ω 2 is small, ≈ 2γω, it would still be of interest to investigate their dependence on N , which, according to the results of Sec. IV, is 1/N to leading order. It would also be interesting to investigate the effect of finite system size on the critical value for stability of the persistent currents in such small systems. According to the results of Sec. V, the value of γ cr also scales as 1/N to leading order. Last but not least, the decay time of the currents would show a much more rapid -and thus more pronounced -decrease as N decreases. VII. CONCLUSIONS In the present study we have investigated the phenomenon of hysteresis and of metastability in a Bose-Einstein condensed cloud of atoms which are confined in a ring potential. Interestingly, this problem has recently been examined experimentally [9], while many other experiments have focused on the question of persistent currents in such topologically nontrivial potentials [1][2][3][4][5][6][7][8]. In the phenomenon of hysteresis the main question is the evaluation of the critical frequencies. As we have shown, in a purely one-dimensional system these two frequencies are related as a consequence of Bloch's theorem. Further, we
| 820 |
118447200
| 0 | 16 |
have evaluated those both within the meanfield approximation and beyond mean field (i.e., by numerical diagonalization of the many-body Hamiltonian) in order to determine finite-N corrections. We have also performed calculations of the critical coupling for the metastabiliity of superflow and of the matrix element associated with the decay rate in a finite system of atoms. As we have argued, the depletion of the condensate due to the interaction combined with the finiteness of the atom number can cause the decay rate to increase exponentially with decreasing N . Thus, the general tendency is that the finiteness of a system makes the supercurrents more fragile, in the sense that it increases the decay rate of the currents, and it also increases the critical coupling for metastability. Given the recent experimental activities on the problems of hysteresis and of metastability, and also given the more general tendency in the community of cold atoms to move to small systems (i.e., systems with a small atom number N ) the present results, which we believe are of theoretical interest, may will become experimentally relevant in the near future.
| 821 |
118447200
| 0 | 16 |
Flavor Oscillations in Field Theory Neutrino flavor oscillations are discussed in terms of two coupled Dirac fields. The interacting Lagrangian is diagonalized to obtain the exact eigenvalues and eigenfunctions. Flavor wave functions are then derived directly from the quantized neutrino fields. Probability density obtained by squaring these wave functions upon taking into account the neutrino chirality are in agreement with the standard neutrino oscillation probabilities. I. INTRODUCTION The theory and phenomenology of neutrino flavor mixing has been extensively studied mainly in the framework of quantum mechanics [1][2][3]. Only very recently, a quantum field theoretical analysis of flavor mixing has been considered using the LSZ formalism [4]. Given the conflicting experimental results which have been obtained from a variety of neutrino observations, it is important to investigate the problem of neutrino propagation in the general context of field theory to reassure ourselves that there are indeed no differences between these results and standard ones obtained from the quantum mechanical treatment. Otherwise said, we need to firmly establish the approximations necessary to derive the expressions used phenomenologically. It is perhaps not superfluous to point out that in the Dirac theory, the contribution of negative energy states becomes substantial, an aspect of the problem totally absent in the quantum mechanical treatment. Our discussion in the present paper will be through the following Lagrangian L =ψ e (iγ · ∂ − m e )ψ e +ψ µ (iγ · ∂ − m µ )ψ µ − δ(ψ e ψ µ +ψ µ ψ e ), (1.1) consisting of two coupled
| 822 |
14225282
| 0 | 16 |
Dirac neutrino fields ψ e , ψ µ with masses m e and m µ . The interaction is provided through a lepton number violating term with a coupling constant δ. The model allows for exact diagonalization. Neutrino and anti-neutrino flavor wave functions can be directly obtained as matrix elements of the quantized neutrino fields. The fully interacting Lagrangian should also include the weak interaction term. It is ignored here (as is done usually), since we are interested in free propagation and possible oscillation in flavor alone. For its production, we simply assume that the neutrino is created through some weak interaction process. Such a breakup is of course an approximation. The effect of parity non-conservation due to the weak interaction term on the other hand, is taken into account by considering the left-handed (right-handed) components of the neutrino (anti-neutrino) flavor wave functions. The model discussed here represents an example (albeit simple) of an interacting field theory which is exactly solvable. This may be of some help in elucidating the properties of interacting theories, which can be normally studied only in perturbation theory. For example, the interaction between the electron and muon neutrino fields produces a change in their masses: the "experimental" masses m 1 , m 2 depend on the "bare" muon and electron neutrino masses m e , m µ and the coupling constant δ. The paper is organized as follows. In Sec. 2, the standard neutrino oscillation phenomenology is reviewed. In Sec. 3, the two coupled Dirac equations, obtained from the Lagrangian (1.1)
| 823 |
14225282
| 0 | 16 |
are solved. The electron and muon neutrino fields are then quantized in terms of the two free uncoupled fields, which diagonalize Eq. (1.1), as described in Sec. 4. The total conserved charge is the sum of the electron and muon flavor charges, which are not conserved separately. The last section closes with some concluding remarks. II. STANDARD TREATMENT We will critically review the standard quantum mechanical treatment of neutrino oscillations to bring out the essential approximations made implicitly and the boundary conditions which are imposed. A state vector |ψ > is introduced as a linear combination of the flavor eigenstates |e > and |µ > (assuming just two flavors) with C e =< e|ψ >, C µ =< µ|ψ > and C e and C µ then become the amplitudes for detecting an electron neutrino and a muon neutrino respectively. To derive the time evolution of the coefficients C e (t) and C µ (t), the state vector |ψ > is written as a superposition of the energy (mass) eigenstates |ν 1 > and |ν 2 > |ψ >= C 1 |ν 1 > +C 2 |ν 2 >, where C 1 and C 2 are the amplitudes for finding the neutrino in the energy states E 1 and E 2 respectively. These coefficients evolve in time as Introducing the rotation matrix between flavor and mass eigenstates it is easy to see that the following relation between the energy and flavor amplitudes holds Hence, the time evolution of the coefficients C e and C µ is
| 824 |
14225282
| 0 | 16 |
given by In Eq. (2.6), the boundary condition that at production we have only a given flavor must be imposed. Suppose for example that at t = 0 a muon neutrino is produced, i.e. ¿From Eq. (2.5) at t = 0 we obtain The time evolution of the flavor amplitudes is obtained by substituting Eq. (2.8) in Eq. (2.9b) Space and therefore momentum is introduced by assuming in Eqs. (2.9) E 2 1 = m 2 1 + p 2 , E 2 2 = m 2 2 + p 2 and x = t. The probability of finding a given flavor is obtained by squaring The assumption that the muon neutrino is created with a definite momentum p is only an approximation as has been pointed out previously [5][6][7][8]. It is in contradiction with four momentum conservation, for example for the reaction π → µν. Each of the possible energy eigenstates has a somewhat different momentum p i . In the rest frame of the pion, energy conservation dictates that (i= 1, 2) Therefore, if we introduce momentum, i.e. space in Eqs. (2.9) we should write In the relativistic approximation x ≃ t the squared moduli of the amplitudes C e (x, t), C µ (x, t) reduce to the standard neutrino oscillation probabilities given by Eqs. (2.10), as described in Ref. [5]. The semiclassical approximation that a neutrino moves at a velocity close to c (assuming very small neutrino masses) "on a classical path" (x = ct) can be a good approximation if the
| 825 |
14225282
| 0 | 16 |
neutrino travels over a macroscopic distance. Hence, if a muon neutrino is produced at the space-time (x ≃ ct = 0), the probability of observing an electron neutrino is maximum for those space-time points x ≃ ct at which |C e | 2 = 1. If some matter is present at these points, processes such as ν e + n → p + e − , will occur, but not processes of the type ν µ + n → p + µ − indicating therefore flavor oscillations. III. FIELD THEORETICAL DISCUSSION Since neutrinos are relativistic particles of spin 1/2, it is important to derive a relativistic equation of motion which can describe such flavor mixing. Energy eigenfunctions can be derived from this equation and one proper way to deal with states of negative energies is to quantize the field. As stated in the Introduction, we will consider the following interacting The parameter δ is an extra mass (energy) related to the small amplitude that a neutrino can flip flavor. The following two coupled Dirac equations However, since there is some amplitude that a neutrino, which is produced as an electron neutrino becomes later a muon neutrino, the possible rest energies of the system are not simply m e and m µ , but are functions of the flipping energy. It is easy to see that the conserved current is Thus, the separate electron and muon flavor currents are not conserved, only their sum is. In order to determine the energy eigenvalues and eigenfunctions of the system
| 826 |
14225282
| 0 | 16 |
of equations (3.2) and (3.3) we consider the ansatz where P is the four-momentum P = (E, p), which is unknown and is to be determined so that the system of differential equations (3.5) and (3.6) is satisfied. The coefficients a and b are Dirac spinors, which can be written as where χ 1,2 and ϕ 1,2 are two component vectors. Substituting Eqs. (3.5), (3.6) into Eqs. (3.2), (3.3), we obtain the system of linear homogeneous equations where σ are the Pauli matrices. The system of Eqs. Solving Eq. (3.9), we obtain (p = |p|) with m 1,2 given by and R = (m µ − m e ) 2 + 4δ 2 . The solutions of Eqs. (3.8) can be written as (3.14) As stated above, for a given value of the momentum p, there are four different energies Corresponding to the positive energy solution E 1 = m 2 1 + p 2 , we have the following two where s = 1, 2 is the spin index and φ 1 (s, p) is given by and u 1 (s, p) is the Dirac spinor For the other positive energy solution E 2 = p 2 + m 2 2 , we have with φ 2 (s, p) given by and M 2 defined as We notice here that because M 1 M 2 = −1 we can write φ 2 (s, p) in terms of M 1 as and u 2 (s, p) is the Dirac spinor Similarly for the solutions of negative
| 827 |
14225282
| 0 | 16 |
energies −E 1,2 we have the following eigenfunctions: with φ 3 (s, p) given by and to make the transformation on the field which uncouples the interacting Lagrangian given by Eq. (3.1) Following relations are useful to see that U uncouples Eq. (3.1) Therefore the fields φ 1 and φ 2 describe the "normal modes". To quantize ψ ν , we expand ψ e and ψ µ in terms of the normal modes (energy eigenfunctions) found in Sec. 3 where the number operators b i and d i (i=1,2) satisfy the anti-commutation relations and spin s. The vacuum state is defined by The total charge operator is (4.8) The following relations hold Hence, for a given value of the momentum p and spin s, there are four possible normal mode These states differ for the charge ( ±1, ±2) Q|1 ps >= |1 ps >, Q|2 ps >= |2 ps >, (4.11a) The above states allow us to construct wave functions in space-time. For example, the wave function associated with the state |1 ps > is where A and B specify the amount of each normal mode state of positive energy present in the state |φ + >. Similarly, a general state of negative charge, momentum p and spin s is given by where C and D specify the amount of each normal mode of negative energy present in the state |φ − >. The matrix element gives the probability amplitude of finding a neutrino of momentum p and spin s at the space-time point (x,
| 828 |
14225282
| 0 | 16 |
t) with the electron flavor. In the same way, the matrix element is the probability amplitude for the muon flavor. To take into account the fact that neutrinos (anti-neutrinos) are created with negative (positive) chiralities, we define the "observable wave functions" as where ψ e (x, t) and ψ µ (x, t) are given by Eqs. (4.15). Hence, the observable flavor neutrino wave functions are Hence the probability densities of finding the electron and muon neutrino flavor are given respectively by (4.25b) The coefficient [2M 1 /(1 + M 2 1 )] 2 is equivalent to sin 2 (2θ) in Eqs. (2.10). Therefore the field theory treatment reduces to the standard quantum mechanical treatment described in Sec. 2. As another example, we consider the case in which the neutrino eigenfunctions of different masses have different momenta p 1 and p 2 with E 1 = p 2 1 + m 2 1 , E 2 = p 2 2 + m 2 2 . where we have assumed that at a given space-time point (x = 0, t = 0), we have only the muon flavor. The probability densities of finding at the space-time point (x, t) the electron and muon neutrino flavors reduce to Eqs. (2.10) in the relativistic approximation |x| ≃ t. V. CONCLUSIONS We have discussed an explicit model of neutrino flavor mixing in the framework of quantum field theory. In this model, the equations of motion for the interacting fields are solved directly and the system is diagonalized in terms of the
| 829 |
14225282
| 0 | 16 |
two uncoupled free fields φ 1 and φ 2 of mass m 1 and m 2 respectively. We notice here that because we can directly diagonalize the Lagrangian we do not need to write the interacting fields in terms of the free asymptotic fields ψ 0e , ψ 0µ of mass m e and m µ respectively. We have also derived neutrino flavor wave functions in such a way that the total flavor charge is constant. The probability densities, derived from these wave functions, are in agreement with the standard neutrino oscillation probabilities, if we take into account the neutrino chirality. Also, since explicit plane wave solutions for all normal modes have been obtained, wave packets corresponding to these can be constructed via standard techniques [9].
| 830 |
14225282
| 0 | 16 |
MICAL1 facilitates breast cancer cell proliferation via ROS‐sensitive ERK/cyclin D pathway Abstract Molecule interacting with CasL 1 (MICAL1) is a multidomain flavoprotein mono‐oxygenase that strongly involves in cytoskeleton dynamics and cell oxidoreduction metabolism. Recently, results from our laboratory have shown that MICAL1 modulates reactive oxygen species (ROS) production, and the latter then activates phosphatidyl inositol 3‐kinase (PI3K)/protein kinase B (Akt) signalling pathway which regulates breast cancer cell invasion. Herein, we performed this study to assess the involvement of MICAL1 in breast cancer cell proliferation and to explore the potential molecular mechanism. We noticed that depletion of MICAL1 markedly reduced cell proliferation in breast cancer cell line MCF‐7 and T47D. This effect of MICAL1 on proliferation was independent of wnt/β‐catenin and NF‐κB pathways. Interestingly, depletion of MICAL1 significantly inhibited ROS production, decreased p‐ERK expression and unfavourable for proliferative phenotype of breast cancer cells. Likewise, MICAL1 overexpression increased p‐ERK level as well as p‐ERK nucleus translocation. Moreover, we investigated the effect of MICAL1 on cell cycle‐related proteins. MICAL1 positively regulated CDK4 and cyclin D expression, but not CDK2, CDK6, cyclin A and cyclin E. In addition, more expression of CDK4 and cyclin D by MICAL1 overexpression was blocked by PI3K/Akt inhibitor LY294002. LY294002 treatment also attenuated the increase in the p‐ERK level in MICAL1‐overexpressed breast cancer cells. Together, our results suggest that MICAL1 exhibits its effect on proliferation via maintaining cyclin D expression through ROS‐sensitive PI3K/Akt/ERK signalling in breast cancer cells. contains flavin mono-oxygenase activity and is responsible for majority of MICAL1's function. 9 Recently, overexpression of MICAL2
| 831 |
3789521
| 0 | 16 |
and MICAL-L2, the other members of MICAL family, has been confirmed to be related to multiple invasive phenotype of cancer cells such as increased motility, proliferation, as well as inducing epithelial-tomesenchymal transition (EMT). 10,11 Domain architecture of MICAL1 is closely related to Drosophila MICAL 4 ; however, to date, only a few reports characterizing the functions of MICAL1 in human cancer progression have been published. Sustaining proliferative signalling and resistant cell death are important hallmarks of cancer. 12 More and more cellular molecules are identified as essentials for regulating those progresses. [13][14][15] Previous studies have reported the anti-apoptosis effect of MICAL1 in human melanoma cells. The mechanism was demonstrated to be associated with MICAL1's negative control of mammalian Ste-20-like kinase 1 (MST1)-nuclear-Dbf2-related kinase (NDR) apoptotic signalling by competing with MST1 for NDR binding. 5,16 Despite its characteristic on anti-apoptosis, whether MICAL1 could influence cancer cell proliferation and the underlying molecular mechanism remains unclear. Recent immunohistochemical studies revealed that MICAL1 is highly expressed in hBRAFV600E human melanomas which display constitutive activation of the AKT, ERK pathway and abnormal melanoma growth. 5 MICAL1 has been identified exert its effect on promoting breast cancer cell invasion with RAB protein. 17 In this study, we will address the role of MICAL1 in breast cancer cell proliferation and provide evidence for a mechanism describing its regulation. Our previous work provided evidence that MICAL1 plays an essential role in the activation of ROS/Akt signalling and cell invasive phenotype and identified a novel link between RAB35 and MICAL1 in promoting breast cancer cell
| 832 |
3789521
| 0 | 16 |
invasion. 17 In the current study, our results suggest that MICAL1 exhibits its positively regulatory function on breast cancer cell proliferation via maintaining cyclin D expression through ROS-sensitive PI3K/Akt/ERK signalling, which implicates an essential role for MICAL1 in breast cancer pathogenesis. | Cell and plasmids Human breast cancer cell lines MCF-7 and T47D were originally obtained from the Cell Biology Institute of Chinese Academy of Sciences (Shanghai, China). Cells were cultured in Dulbecco's modified Eagle's medium (DMEM, high glucose) (Hyclone) supplemented with 10% (v/v) foetal bovine serum (FBS) (Hyclone) and antibiotics (100 U/ mL streptomycin and 100 lg/mL penicillin) (Invitrogen) in a humidified incubator at 37°C with 5% CO 2 . Cells were grown on coverslips for fluorescence staining and on plastic dishes for protein extraction. | MTT assays Cell viability was determined by 3-(4,5-dimethylthiazol-2-yl)-2,5diphenyltetrazolium bromide (MTT) assay, as described previously. 18 In brief, cells were seeded at a density of 5 9 10 3 cells per well into 96-well plate and transfected with siRNA or plasmids as indicated. Ten replicas were made for each group. After cultured for the indicated time, cells were washed and MTT (Sigma) was added. The plate was then incubated in the dark for 4 hours, followed by measurement at 490 nm using a microplate absorbance reader (Bio-Tek, Elx800, USA). | Colony formation assays In brief, 1 9 10 3 stably transfected cells were seeded into 6-well dish and cultured for up to 2 weeks. Colonies were visualized by crystal violet staining, and then, colonies were photographed and counted. | Flow
| 833 |
3789521
| 0 | 16 |
cytometry analysis Cell cycle analysis was performed by flow cytometry. Briefly, cells were harvested and fixed in 80% ice-cold ethanol overnight. Then the cells were incubated with RNase A and propidium iodide staining solution at 37°C for 30 minutes in darkness. Subsequently, the stained cells were analysed using flow cytometry. | Cytoplasmic and nuclear protein extraction Cytoplasmic and nuclear proteins were obtained using the Nuclear and Cytoplasmic Protein Extraction Kit (Beyotime) according to the manufacturer's instructions. 19 In brief, cells were harvested by centri- The pellet was vortexed again and centrifuged for 5 minutes at 5000 rpm. Supernatant was collected as cytoplasmic extract. The insoluble (pellet) fraction was suspended by nuclear extraction agent. After vortexed several times, the mixture was centrifuged for 10 minutes at 12000 rpm. The supernatant was collected as nuclear extract. | Western blotting assays Subconfluent cells were washed with PBS and lysed for 20 minutes on ice in a RIPA buffer containing 1% protease inhibitor cocktail (Beyotime). The protein was collected, and its concentration was determined by the BCA protein assay reagent kit (Thermo Fisher Scientific) then separated by SDS-PAGE and transferred to nitrocellulose membrane. The membrane was then blocked with 5% non-fat milk for 1 hour at room temperature and incubated with primary antibody overnight at 4°C. The following antibodies were used: | Immunofluorescence assays Cells for immunostaining were fixed in ice-cold methanol for 20 minutes, then permeabilized in 0.1% Triton X-100 and blocked in PBS containing 1% BSA (Sigma) for 1 hour at room temperature. The cells were incubated with
| 834 |
3789521
| 0 | 16 |
primary antibody overnight at 4°C followed by incubation with rhodamine-conjugated secondary antibody (Life Technologies) for 1 hour at room temperature within a moist chamber. After washing with PBS, the samples were mounted with DAPI Fluoromount G (Southern Biotech). Images were acquired using an Olympus BX51 microscope coupled with an Olympus DP70 digital camera, and fluorescence intensity was quantified by ImageJ. Five images in each group were analysed. | 5-Ethynyl-2-deoxyuridine (EdU) incorporation assays Cell proliferation was measured using EdU staining kits (Keygen) according to the manufacturer's instruction. Five replicas were made for each group. In brief, cells were cultured on coverslips until reaching 70% confluence, then EdU was added to the culture media for 2 hours. After labelling, cells were washed three times with PBS followed by formaldehyde fixation. Then, the cells were incubated with glycine and washed with PBS containing 0.5% Triton X-100. After the cells were counterstained with DAPI, cells were mounted and imaged by fluorescence microscopy (Olympus BX 51, Tokyo, Japan) coupled with an Olympus DP70 digital camera. | Hoechst staining Cell apoptosis was measured using Hoechst staining according to the manufacturer's instruction (Beyotime). Three replicas were made for each group. In brief, the Hoechst 33258 staining solution was prepared by diluting the Hoechst stock solution 1:2000 in PBS. After removing the medium, sufficient staining solution was added to the cells, and then, the cells were protected from the light for 30 minutes at room temperature. The staining solution was removed and cells were washed with PBS for three times. Observation and photography were
| 835 |
3789521
| 0 | 16 |
performed by a fluorescence microscope. Images were collected using a fluorescence microscope. | Statistical analysis Statistical analysis was performed using the SPSS. Error bars represent the standard error of mean SEM, and the significance of difference between the two groups was analysed by Student's t test. P < .05 represents statistical significance, and P < .01 represents sufficiently statistical significance (two-tailed). All experiments were repeated at least three times independently. and T47D cells and found that by silencing MICAL1, cell growth rate in both cell lines was inhibited ( Figure 1B). In contrast, induced MICAL1 expression resulted in an accelerated cell proliferation rate compared with the controls (Figure 1C,D). The colony formation assay showed that MCF-7 cells with stable MICAL1 expression increased colony number by 1.5-fold compared to control cells (P < .05) ( Figure 1E). Edu staining analysis showed that MCF-7 and T47D cells with low MICAL1 expression exhibited significant reduction in cell proliferation rate ( Figure 1F). Hoechst staining showed that MICAL1 did not significantly alter cell apoptosis ( Figure 1G). All these results indicated a stimulatory function of MICAL1 in proliferation of breast cancer cells. | MICAL1 depletion decreases expression of cyclin D To further verify the impact of MICAL1 expression on breast cancer proliferation, we assessed cell cycle distribution. As shown in The progression of cell proliferation needs to pass through cell cycle which is strictly regulated by key regulatory proteins. A number of key markers, such as cyclin-dependent kinases (CDK2, CDK4, CDK6) and cyclin (A, D, E), reportedly dynamically modulate cell
| 836 |
3789521
| 0 | 16 |
cycle. We noticed that, among these molecules, the levels of cyclin D and CDK4 were markedly blocked after silencing MICAL1. Consistently, cyclin D and CDK4 were higher when MICAL1 overexpressed in T47D cells ( Figure 2E,F) and MCF-7 cells (Figure 2G,H). These results indicate that cyclin D and CDK4 play key roles in MICAL1induced cell proliferation. | MICAL1-mediated cell proliferation is independent of wnt/b-catenin, NF-jB and mTOR pathways To determine whether the effects of MICAL1 on cell proliferation were associated with wnt/b-catenin or NF-jB pathways, we | Involvement of p-ERK signalling in MICAL1induced proliferation Next, we examined the effect of p-ERK inhibitor U0126 on cyclin D expression and cell proliferation. As shown in Figure 4A, overexpression of MICAL1 resulted in increased cyclin D in breast cancer cells. In addition, the down-regulated expression of cyclin D was induced by silencing of MICAL1. Here, increased cell proliferation rate ( Figure 4B) as well as higher expression level of cyclin D and c-myc was found in MICAL1 overexpression cells, and this change was markedly blocked after treatment with U0126 ( Figure 4C). As shown in Figure 4D,E, lower levels of p-ERK were found both in cytoplasm and in nucleus parts in the MICAL1-depletion T47D and MCF-7 cells. In contrast, the expression of p-ERK in the MICAL1-overexpression T47D and MCF-7 cells was increased greatly when compared to the control group. Combined with the above results, it suggests that MICAL1 may support ERK activation and then increases cyclin D expression which is involved in cell cycle and proliferation regulation in breast
| 837 |
3789521
| 0 | 16 |
cancer cells. | ROS mediates ERK/cyclin D signals via phosphorylated Akt MICAL1 is well characterized in ROS generation, which has been associated with cancer cell invasion. 17 We transfected MCF-7 and T47D cells with siMICAL1 and then examined its effects on ROS production. As shown in Figure 5A, ROS level in those cells was inhibited by siMICAL1 transfection. ROS scavenger NAC was used here to eliminate ROS production ( Figure 5B). To probe the involvement of PI3K/Akt activation in MICAL1-induced cell proliferation, we transfected T47D cells with MICAL1 and p-Akt expression was detected by Western blotting assays. Our observations yielded evidence that the level of p-Akt was markedly increased after the overexpression of MICAL1. Moreover, pre-treatment with NAC inhibited p-Akt level when MICAL1 was overexpressed ( Figure 5C). NAC and LY294002 pre-treatment also delayed the increased p-ERK, c-myc and cyclin D levels induced by MICAL1-overexpression ( Figure 5C, D). MICAL1-induced cell proliferation was also inhibited by U0126 and LY294002 in both MCF-7 and T47D cells ( Figure 5E,F). Collectively, these data indicated that MICAL1 may facilitate ROS generation, which leads to PI3K/Akt/ERK/cyclin D signalling activation and breast cancer cell proliferation ( Figure 6). | DISCUSSION MICALs are multidomain proteins previously known for their roles in organization of cytoskeleton and synaptic structures. For example, Drosophila MICAL is identified as an essential for neuronal growth by interactions with the cytoplasmic region of plexin and activating plexin-and semaphorin-mediated axonal signalling. 20 Recent studies also reported that MICAL1 promoted the development of hippocampal mossy fibre connections through its ability to
| 838 |
3789521
| 0 | 16 |
control actin cytoskeleton rearrangement. 21 The notion that cell cytoskeleton rearrangement contributes to cell proliferation has been confirmed by many research groups 15 ; however, up to now, no study yet addressed the function of MICAL1 in the human cancer cell proliferation process. In contrast to previous findings which indicated that the expression of MICAL-L2, another member of MICAL family, is related to the clinical stage and histologic grade of ovarian cancer, 11 this study showed a novel link between MICAL1 and cell proliferation for the first time. Therefore, it is interesting to investigate the signalling mechanisms underlying the effect of MICAL1 on promoting breast cancer cell proliferation. Cyclin-dependent kinase is protein kinase that could regulate cell cycle by binding to cyclin. CDK4-Cyclin D complex is a major integrator of cell proliferation which is required for progression through G1 phase and entry into S phase of the cell cycle. Here, we noticed that knockdown of MICAL1 in breast cancer cells significantly downregulated both CDK4 and cyclin D protein levels, whereas MICAL1 overexpression could reverse those effects on CDK4 and cyclin D. In addition, the numbers of breast cancer cell in S phase of the cell cycle were significantly lower after silencing of MICAL1 compared with control group, while MICAL1 overexpression improved them. The above results indicate that MICAL1 expression may lead to an increase in CDK4-Cyclin D complex that stimulates cellular G1/S transition, which is probably involved in MICAL1-induced breast cancer cell proliferation. Cyclin D is regulated by the upstream pathways including NF-jB, PI3K/Akt/mTOR and Wnt/b-catenin.
| 839 |
3789521
| 0 | 16 |
[22][23][24][25][26] MICAL1 has an important role in triggering Akt phosphorylation, which is a key signal that can endow breast cancer cells with an invasive phenotype. 17 We have noticed that MICAL1 selectively activates PI3K/Akt signalling pathway. To our surprise, p-S6K protein level, the main downstream effector of PI3K/Akt/mTOR cascade, was increased after MICAL1 depletion and decreased after MICAL1 overexpression, suggesting that mTOR may not the target of MICAL1 for proliferation regulation. To find out whether the effects of MICAL1 were dependent on Wnt/b-catenin and NF-jB signalling pathways or not, we explored the expression of p-b-catenin, p-GSK-3b and the distribution of NF-jB in cytoplasm and nucleus. Results showed that MICAL1 depletion did not significantly alter expression or location of those markers, indicating that the inhibitory effects of MICAL1 were also independent on those signalling pathways. MAPKs are highly conserved Phosphorylation of ERK and its translocation to nucleus by ROS have been proven to be an important mechanism to mediate breast cancer cell migration by LPA. 28 We detected an up-regulation of p-ERK in the nucleus upon MICAL1 expression. In addition, inhibition of ERK activation by U0126 results in an decrease in cyclin D expression in MICAL1-overexpressed cells, indicating that MICAL1induced breast cancer cell proliferation might be due to cyclin D expression by a p-ERK dependent way. More and more evidence showed that MICAL uses its FAD domain to either oxidize proteins or produce ROS such as H 2 O 2 . MICAL proteins may oxidize actin at methionine residues directly, causing depolymerization of actin filaments. 8 Microtubule
| 840 |
3789521
| 0 | 16 |
assembly can also be altered indirectly by the ROS produced by MICAL. 29 Oxidative stress originates from an imbalance between the generation and scavenging of ROS, activates aberrant signalling cascades and leads to tumorigenesis. A recent study reported that increased ROS production was triggered by overexpression of constitutively active MICAL1 mutants. 16 On the contrary, ROS levels were significantly attenuated upon transfection of the enzymatically impaired FAD domain mutant. 4 Consistent with these findings, we found that knockdown of MICAL1 decreased the production of ROS, suggesting that ROS may be a major downstream effect of MICAL1. PI3K/Akt and ERK are particularly sensitive to redox reaction. 28,30 Additionally, ROS scavenger NAC significantly reversed MICAL1 overexpression induced up-regulation of p-Akt and p-Akt inhibitor LY294002 significantly reversed MICAL1 overexpression induced up-regulation of p-ERK, CDK4 as well as cyclin D in breast cancer cells. In summary, these data suggest that MICAL1 may promote breast cancer cell proliferation by ROS-PI3K/Akt signalling pathway activation. The present study was a continuation of our previous study where we have proven that ROS-PI3K/Akt signalling pathway is selectively responsible for MICAL1-induced breast cancer cell invasion, 17 providing a basis for further exploring the role of PI3K/Akt signalling in MICAL1-induced malignant phenotype. In this study, we examined the effects of MICAL1 on breast cancer cell proliferation and found that MICAL1 exhibits its effects by regulating cyclin D expressions via ROS-sensitive PI3K/Akt/ERK signalling. Our observations described MICAL1's effect on breast cancer cell proliferation and may help to better understand how deregulation of MICAL1 contributes to breast cancer progression.
| 841 |
3789521
| 0 | 16 |
AEWAE: An Efficient Ensemble Framework for Concept Drift Adaptation in IoT Data Stream —With the evolution of the fifth-generation (5G) wireless network, smart technology based on the Internet of Things (IoT) has become increasingly popular. As a crucial component of smart technology, IoT systems for service delivery often face concept drift issues in network data stream analytics due to dynamic IoT environments, resulting in performance degradation. In this article, we propose a drift-adaptive framework called Adaptive Exponentially Weighted Average Ensemble (AEWAE) consisting of three stages: IoT data preprocessing, base model learning, and online ensembling. It is a data stream analytics framework that integrates dynamic adjustments of ensemble methods to tackle various scenarios. Experimental results on two public IoT datasets demonstrate that our proposed framework outperforms state-of-the-art methods, achieving high accuracy and efficiency in IoT data stream analytics. I. INTRODUCTION With the evolution of the fifth-generation (5G) wireless network, the Internet of Things (IoT) has become a revolutionary technique that enables a diverse number of features and applications [1]. The IoT has been applied to many areas of daily life, including smart cities, Smart Logistics [2], intelligent transportation systems (ITS), smart healthcare, smart agriculture and so on [3]. Data from a variety of sources and domains, such as IoT sensors and intelligent devices, can be gathered in the realm of IoT. This information is then sent to centralized servers for analysis via multiple communication protocols, such as WiFi, Bluetooth, and ZigBee, among others [3]. Utilizing data analytics methods on IoT data streams allows for the extraction of
| 842 |
258615490
| 0 | 16 |
significant insights, which can be applied to a wide array of IoT applications, such as identifying anomalies [4]. The general structure of IoT data analytics can be visualized in Figure 1. Currently, IoT data analytics faces two main challenges. Firstly, IoT systems are vulnerable to most existing cyberthreats since IoT devices are connected mostly over wireless networks and are typically utilized in an unattended fashion. In this type of environment, an attacker may easily gain both physical or logical access to these devices illegally [5]. On the other hand, real-world IoT traffic data typically consists of dy-* Scientific Research Projects of Guangdong Provincial Education Department (Specialized in Key Areas) under Grant 2022ZDZX1015. IoT Devices Fig. 1. The architecture of IoT data stream analytics namic data streams, which are continuously generated in everchanging, non-stationary IoT environments. Consequently, in practical applications, IoT data analytics frequently encounter the challenge of concept drift [6], as data distributions evolve over time. The presence of concept drift introduces significant challenges when developing machine learning models, since their learning performance may progressively degrade owing to data distribution changes [4]. Thus, it is crucial to develop online adaptive analytics models capable of accommodating both predictable and unpredictable shifts in the IoT data, ensuring accurate and robust performance in real-world applications [7]. Therefore, this article proposes a drift adaptive framework named Adaptive Exponentially Weighted Average Ensemble (AEWAE). It consists of three stages: IoT data preprocessing, base model learning, and online ensembling. The proposed framework can be implemented on IoT cloud servers for processing large-scale data
| 843 |
258615490
| 0 | 16 |
streams transmitted from IoT end devices using wireless communication strategies, as illustrated in Fig. 1. The effectiveness and efficiency of the proposed adaptive framework are evaluated using two public IoT cybersecurity datasets: IoTID20 [8] and CICIDS2017 [9], to evaluate the proposed framework in intrusion detection use cases as an example for IoT data stream analytics. The main contributions of this article can be summarized as follows: 1) It proposes an optimized adaptive framework for IoT anomaly detection based on ARF-ADWIN, ARF-EDDM, OPA, KNN-ADWIN, PSO, and AEWAE and the implementation code is open access on GitHub. 2) It proposes adaptive exponentially weighted average ensemble (AEWAE) to address the concept drift issues, a novel ensemble drift adaptation strategy for online learning on dynamic data streams . 3) The proposed framework is assessed through a case study using two public IoT security datasets, and its performance is compared with a variety of state-of-the-art online learning techniques. The remainder of this paper is organized as follows: Section II provides a literature review of state-of-the-art IoT anomaly detection and concept drift adaptation methods. Section III describes the proposed AEWAE drift adaptation framework. Section IV presents and discusses the experimental results. Section V concludes the paper. II. RELATED WORK Owing to the dynamic nature of IoT environments, IoT streaming data is susceptible to various shifts in data distribution. For instance, the physical events captured by IoT sensors may evolve over time, while the sensing components themselves may age or require periodic updates. The associated alterations in IoT data distribution are named concept
| 844 |
258615490
| 0 | 16 |
drift [6] [7]. The presence of concept drift may undermine the decisionmaking capabilities of IoT data analytics models, leading to significant repercussions in IoT systems. To tackle concept drift, it is essential to employ effective methods that can detect such drifts and adapt to the changes accordingly. This section introduces and discusses existing methods for concept drift detection and adaptation. A. Concept Drift Detection One key difficulty in drift detection is dealing with the various types of concept drift, including sudden and gradual drifts [7]. Another challenge stems from the wide range of factors contributing to concept drift in IoT systems, which can be attributed to IoT-related events (e.g., system updates, device substitutions, and various network incidents) as well as time-series aspects (e.g., seasonal patterns and trends). Window-based methods and performance-based methods are two potential solutions for drift detection [7]. Adaptive Windowing (ADWIN) is a prevalent window-based approach that employs adaptive sliding windows to identify concept drift by examining the statistical discrepancy between two neighboring sub-windows [7]. While it performs effectively for gradual drifts and long-term changes, ADWIN may inadvertently lose valuable historical data, leading to false alarms and unnecessary model updates. Additionally, the Early Drift Detection Method (EDDM) [10] is a widely-used performance-based drift detection technique that identifies concept drift by tracking the extent of model performance deterioration. While EDDM is proficient at detecting drifts that result in model degradation, particularly sudden drifts, its performance in detecting gradual drifts falls short compared to distribution-based methods. Thus, employing different methods can efficiently detect various types of
| 845 |
258615490
| 0 | 16 |
concept drifts. B. Concept Drift Adaptation Upon detecting a concept drift, learning models need to adapt to new concepts and improve their performance. Current drift-adaptive learning techniques can be broadly classified into three main categories: adaptive algorithms, incremental learning, and ensemble learning methods [7]. Adaptive algorithms for drift adaptation typically monitor the model's performance on new data and make adjustments to its parameters based on feedback, enabling the system to learn from incoming data, detect changes in the data distribution, and update the model accordingly. They typically consist of a machine learning model paired with a method for detecting concept drift. For example, Losing et al. [11] proposed Knearest neighbors with ADWIN drift detector (KNN-ADWIN). KNN-ADWIN is based on the K-nearest neighbor algorithm and uses the ADWIN algorithm to detect and update changes in the data distribution. Incremental learning refers to the process of sequentially learning from incoming data samples, where the learning model is partially updated upon the arrival of new data or detection of drift. Hoeffding trees (HTs) [7] is a fundamental incremental learning technique that utilizes the Hoeffding inequality to establish the minimum number of data instances needed for each decision node, enabling nodes to adapt to new instances. Another incremental learning algorithm is online passive-aggressive (OPA) [12], which adjusts to concept drift by passively accepting accurate predictions and aggressively addressing errors. Hoeffding Anytime Tree (HATT) [13] is an incremental learning approach that selects and divides nodes as soon as the confidence threshold is met, rather than determining the optimal split as in
| 846 |
258615490
| 0 | 16 |
Hoeffding trees. This approach makes HATT more efficient and precise in adapting to concept drift. Ensemble learning is a learning approach that combines the output of multiple base learners to create an ensemble model, which boasts improved generalization ability and performance in adapting to concept drift. Adaptive random forest (ARF) [14] and streaming random patches (SRP) [15] are two excellent ensemble online learning techniques that utilize multiple Hoeffding Trees (HTs) as base models and incorporate a drift detector (such as EDDM) for each HT to tackle concept drift. ARF employs a local subspace randomization approach for tree construction, whereas SRP leverages global subspace randomization to create random feature subsets for model training. PWPAE [4] and MSANA [16] are two ensemble frameworks proposed by L. Yang et al. In PWPAE, weights are calculated based on the real-time accuracy of individual samples, whereas MSANA's ensemble method incorporates the idea of windows, determining weights by examining the trial error rate within a specific window. Ensemble models, which maintain both historical and new data patterns through various base learners, are effective at adapting to concept drift in dynamic data stream analytics. Although they typically yield better performance than incremental learning methods, their execution time tends to be higher. This study seeks to introduce an ensemble framework that effectively balances the trade-off between model performance and learning speed. Additionally, existing approaches primarily concentrate on dynamically selecting models or preprocessing techniques. However, there is limited research on dynamically adapting ensemble methods when confronted with varying datasets. Thus, this article presents a comprehensive
| 847 |
258615490
| 0 | 16 |
data analytics framework that incorporates dynamic adjustments of ensemble methods to address different scenarios. III. PROPOSED FRAMEWORK A. System Overview The overview of the proposed framework for detecting anomalies in IoT systems by leveraging data stream analytics is shown in Fig. 2. It consists of three main phases: IoT data preprocessing, base model learning and online ensembling. At the first stage, to obtain a more representative subset of the incoming IoT data streams, the k-means cluster sampling method is employed to extract a highly informative sample, which is discussed in Section III-B.1. And then it will be subjected to Z-score normalization to standardize its data distribution and bring all features to the same scale which is discussed in Section III-B.2. Second, four base learners known as ARF-ADWIN , ARF-EDDM, OPA, and KNN-ADWIN are developed to handle concept drift adaptation and perform preliminary anomaly detection which is discussed in Section III-C. These base learners are designed to adapt to the changing data distribution and identify potential anomalies in the data stream in real-time fashion. In the last phase, the proposed AEWAE method integrates the outputs of selected base learners by utilizing their prediction probabilities and real-time error rates for online model ensemble. Additionally, the Particle Swarm Optimization (PSO) algorithm is employed to optimize the hyperparameter of AEWAE, as discussed in Section III-D. B. Data Preprocessing Data preprocessing is essential for enhancing the quality of data streams to improve model learning performance. In IoT data streams, the challenges of handling large amounts of continuously generated data make it
| 848 |
258615490
| 0 | 16 |
infeasible and unnecessary to utilize all data samples for model development. Consequently, effective data sampling and normalization methods, such as K-means clustering and Z-score normalization, should be employed to select highly representative data samples. 1) Data Sampling: In the context of IoT data streams, Kmeans clustering serves as an efficient data sampling technique that partitions the data into a predefined number of clusters. By selecting representative samples from each cluster, the algorithm ensures the chosen data points capture the underlying distribution of the entire dataset effectively, thus maintaining the data's essential characteristics while reducing its volume. Owing to the substantial volume of IoT traffic datasets, the K-means cluster sampling method is employed to select a mere 1% of the original data samples for evaluating the proposed framework. The size of the resulting subset can be flexible, depending on factors such as the rate at which IoT data is produced and the processing capabilities of the server machines. In contrast to alternative sampling techniques, Kmeans cluster sampling can produce a high-quality and highly representative subset by primarily discarding data points that are redundant. 2) Data Normalization: Machine learning and data analytics techniques frequently place greater importance on features with larger values. Utilizing data scaling or normalization methods, which adjust features within a dataset to a consistent scale, can help circumvent biased models and enhance learning performance. Z-score is a commonly used scaling technique for data analytics problems. Compared to alternative normalization techniques, Z-score normalization effectively standardizes features, enabling models to better focus on relevant patterns and can
| 849 |
258615490
| 0 | 16 |
handle outliers of IoT data streams [17]. The Z-score normalization method scales the feature value of each data sample x to a normalized value z as [18] where z represents the Z-score, x is the original data point, µ is the mean of the dataset, and σ is the standard deviation of the dataset. C. Drift Adaptation Base Learner Selection In the proposed framework, lightweight online learning methods, as described in Section II-B, are employed to learn from data streams. Initially, these learning methods process a small training set to generate the primary base learners. As new samples from the online test set are encountered, the learners predict and adapt, updating themselves if concept drift is detected. Two detection methods, ADWIN [7] and EDDM [10], are incorporated into the framework to identify concept drift. While ADWIN compares the mean values of two consecutive data windows, EDDM focuses on monitoring changes in model performance using drift and warning thresholds, denoted by β and α, respectively, as expressed in the following equation: Where P max represents the maximum performance measure achieved, α is the warning threshold, and β is the drift threshold. One of the basic online learners in the framework is the Hoeffding Tree (HT), which adapts to concept drift by employing the Hoeffding bound. The Hoeffding bound is defined as: In this equation, ε denotes the Hoeffding bound, R symbolizes the range of the output variable, δ represents the desired level of confidence, and n stands for the number of samples. The Adaptive Random Forest (ARF)
| 850 |
258615490
| 0 | 16 |
[14] is chosen as an advanced ensemble model, which uses local subspace randomization to build HTs for drift adaptation, and incorporates a drift detector for identifying concept drift. As ARF demonstrates efficiency and effectiveness across various data stream analytics problems, two ARF models with different drift detectors, ARF-ADWIN and ARF-EDDM, are selected as two base online learning models. Next, upon carefully evaluating the online learning methods, including HT [7], HATT [13], OPA [12], SRP [15], and KNN-ADWIN [11], the selection process took into account both performance and execution time. While HT and HATT computational costs are low, they were ultimately not chosen due to some performance constraints. Similarly, SRP, though it can perform well among many online learning tasks, is not selected for the ensemble framework, as its high computational complexity. As a result, the final online learning methods selected for the proposed framework are OPA, KNN-ADWIN, ARF-ADWIN and ARF-EDDM. The reasons for choosing them as base learners are as follows: 1) As described in Section II-A, ADWIN works well with gradual drift and EDDM is effective in detecting sudden drift, using both of them enables the detection of different types of concept drift. 2) ARF-ADWIN and ARF-EDDM, both extensions of the Adaptive Random Forest (ARF) algorithm, have demonstrated improved performance in handling concept drift compared to other existing drift adaptation methods by experimental studies in [14]. The use of these two variants in the ensemble model allows the proposed framework to benefit from their enhanced adaptability to concept drift, further improving the overall performance of
| 851 |
258615490
| 0 | 16 |
the system. 3) One of the key advantages of the selected models, ARF-ADWIN, ARF-EDDM, OPA and KNN-ADWIN, is their ability to deliver strong performance while maintaining shorter execution times [16]. This balance between efficiency and effectiveness makes them well-suited for real-time IoT data stream analysis and handling concept drift. 4) An effective ensemble model should maintain high diversity among base learners, increasing the likelihood of performance improvement. Despite using different learning algorithms, the chosen models, ARF, OPA, and KNN-ADWIN, exhibit sufficient diversity in their construction processes to result in a more robust and diverse ensemble model. D. Drift Adaptation Ensemble Framework: AEWAE Upon acquiring the base learning models, their predictive outputs are combined to create an enhanced ensemble model using a novel ensemble strategy called Adaptive Ensemble with Exponentially Weighted Average Errors (AEWAE), as shown in Algorithm 1, to boost performance. AEWAE combines base models by allocating dynamic weights to the predictive likelihoods of these models, subsequently averaging the weighted probabilities. The category with the highest average probability value, indicating the most assured outcome, is chosen as the ultimate prediction result. Given a data stream D = {(x 1 , y 1 ),..., (x n , y n )} , with c different target classes, y ∈ 1, ..., c, the final predicted target classŷ for each input data x can be denoted by the following formula: Where b j=1 represents the summation symbol that iterates through all the base models; b is the number of base models and in the proposed ensemble model, b =
| 852 |
258615490
| 0 | 16 |
4; w j indicates the weight of the j th base model; p j (y j = i|L j , x) is the prediction probability of class i for the data sample x using the j th base model L j ; Finally, L j represents the j th base model. The weight w j for each base model L j is calculated using the inverse of the model's real-time error rate, as described below: where is a small constant used to avoid a denominator of 0. Consequently, greater weights are allocated to the base models exhibiting lower error rates and superior performance. The EWA(Exponentially Weighted Average) error rate for each base model L j can be calculated by where i represents the i th data in (X test, y test), and α indicates the weighting decay factor for historical errors when calculating the exponentially weighted moving average. Train base learners using X train, y train; 5: for each instance (x i , y i ) in X test, y test do 6: for j = 1 to 4 do 7: y pred j ← hat j (x i ); // Predict label 8: y prob j ← hat j .proba(x i ); // Predict probabilities 9: hat j ← hat j .learn(x i , y i ); // Update base learner 10: error j ← I(y pred j = y i ); // Compute error for each base learner 11: cumulative error j ← α · cumulative error j + (1 − α) · error
| 853 |
258615490
| 0 | 16 |
j ; // Compute cumulative error 12: ewa error rate j ← cumulative errorj 1−α (i+1) ; // Compute exponentially weighted average error rate 13: end for 14: ea ← 4 k=1 1 ewa error rate k + ; // Compute ensemble accuracy 15: for j = 1 to 4 do y pred ← I(y prob0 < y prob1 ); // Make final prediction based on the probability comparison between the two classes 23: Compute and update Accuracy, Precision, Recall, and F1-score using y pred and y i ; α (i+1) ) is to normalize the cumulative error, so that ewa error rate j is scaled to the range between 0 and 1. This scaling is necessary to make ewa error rate j comparable across different iterations and samples, and to enable it to serve as a reliable indicator of the algorithm's prediction accuracy. The cumulative error rate for each base model L j can be denoted by cumulative error j = α · cumulative error j + (1 − α) · error j where α is a weighting factor or a smoothing parameter that return M axAcc, HP opt ; // The best accuracy and hyperparameters 12: end function determines the contribution of the previous cumulative error to the current value of the cumulative error. The cumulative error rate is a variation that draws inspiration from the exponentially weighted average calculation method. Exponentially Weighted Moving Average (EWMA) charts were initially proposed in [19] for detecting an increase in the mean of a sequence of random
| 854 |
258615490
| 0 | 16 |
variables. The study began to consider real-time dynamic control of processes using discrete data [20]. By employing this method to calculate the weight, it can further enhance our adaptability to dynamic IoT data streams. The error rate associated with each base model L j , can be expressed as follows: where y pred j denotes the predicted label from the base model j; and y i represents the actual label for the data point i. The AEWAE model exhibits dynamic adaptability to diverse scenarios by leveraging the optimized α to make ensemble method adjustments. Consequently, Particle Swarm Optimization (PSO) is employed to fine-tune this hyperparameter in order to achieve an optimized ensemble model. The complete HPO method is given by Algorithm 2. The reasons for choosing PSO to tune the hyperparameter are as follows [21]: 1) PSO is simple to implement and exhibits rapid convergence. 2) PSO is an efficient HPO method as it has a low computational complexity of O (N log N ) and support for parallel execution. 3) PSO demonstrates effectiveness in handling various hyperparameter types including the continuous category that the AEWAE hyperparameter falls into. In order to execute AEWAE, the configuration space for the hyperparameter is provided to the "HPO" function. Subsequently, PSO identifies the most suitable hyperparameter that yields the maximum overall accuracy. The discovered optimal hyperparameter is then supplied to the "DriftAdaptation" function, which constructs the optimized model for precise IoT data analysis. The primary contributor to the time complexity of the ensemble model is the complexity of the
| 855 |
258615490
| 0 | 16 |
base learners. The time complexity of the AEWAE algorithm itself is limited to O(nck), where n represents the number of samples, c is the number of class values in the target variable, and k indicates the number of base learners. Since both c and k are typically small values, the algorithm's overall time complexity remains relatively low. Compared to other existing drift-adaptive online learning methods, the proposed AEWAE approach has the following advantages: 1) In contrast to numerous existing ensemble learning methods that employ the hard majority voting strategy, the proposed framework utilizes the confidence probability of each base classifier for every class, offering a more resilient and adaptable approach. By taking into account the uncertainty of each base classifier for every data sample, the framework avoids making arbitrary decisions, leading to improved overall performance and more accurate predictions. 2) By utilizing an exponentially weighted average for weight calculation and Particle Swarm Optimization (PSO) for hyperparameter optimization, the proposed framework enables dynamic adjustments of ensemble methods when confronted with diverse scenarios. This approach leads to a highly effective adaptation to concept drift in dynamic IoT data streams. 3) Opting for lightweight base learning models allows for the creation of an efficient ensemble model, addressing a IV. PERFORMANCE EVALUATION significant limitation of many existing ensemble models: their high complexity. By using simpler base learners, the ensemble model can maintain high predictive performance while reducing computational demands, making it more suitable for real-time applications and large-scale IoT data processing. A. Experimental Setup The proposed framework was implemented in
| 856 |
258615490
| 0 | 16 |
Python 3.9 on the Google Colaboratory platform by extending the River [22] library. The system was run on a computer equipped with an Intel(R) Xeon(R) CPU @ 2.20GHz and 12.7 GB of memory, representing an IoT central server machine for big data analytics purposes. We assessed the proposed approach using two public IoT security datasets: IoTID20 [8] and CICIDS2017 [9]. IoTID20, a more recent IoT dataset, consists of IoT network traffic data generated from both benign and malicious IoT devices, featuring 83 distinct network attributes. On the other hand, CICIDS2017 is a public network security dataset provided by the Canadian Institute of Cybersecurity, encompassing stateof-the-art cyberattack scenarios. Given that the CICIDS2017 dataset was produced by executing various attack types across different time frames, the attack patterns within the dataset have changed, leading to six concept drifts, as depicted in Fig. 3. For this study, we selected a representative IoTID20 subset containing 6,252 data points and a sampled CICIDS2017 subset with 14,155 records. These subsets were obtained using the K-means method for model evaluation purposes. Anomaly detection systems primarily aim to differentiate between cyberattacks and regular system states. As a result, the datasets employed are binary, labeled either as "normal" or "attack". Two evaluation methods, hold-out validation and prequential validation, are utilized to assess the proposed framework. In the hold-out validation approach, the initial 10% of the data is designated for model training, while the remaining 90% is allocated for online testing. On the other hand, prequential validation, also known as test-and-train validation, entails using each sample
| 857 |
258615490
| 0 | 16 |
from the online test set to first evaluate the learning model, followed by its integration into model training or updating. This process ensures that the model is continuously refined and updated as new data is introduced. The experimental results of the proposed framework are thoroughly analyzed from various angles to offer a comprehensive understanding. The analysis considers two main perspectives: effectiveness and efficiency. From the effectiveness perspective, four performance metrics are employed to assess the proposed framework: accuracy, precision, recall, and F1-score. Due to the imbalanced nature of IoT anomaly detection datasets, precision, recall, and F1-score are used alongside accuracy to deliver a well-rounded evaluation of the model's performance. From the efficiency perspective, two machine learning (ML) and data analytics-related quality of service (QoS) parameters, latency, and throughput, are utilized to gauge the learning efficiency of the proposed framework. Latency refers to the time taken by learning models to infer, calculated as the average test time per data sample or packet. Throughput, on the other hand, represents the number of processed data samples or packets per unit of time (e.g., per second). Low latency and high throughput are crucial performance requirements for ML and data analytics models [23]. In addition to these parameters, the probability density function is employed to assess the efficiency of the proposed model. For real-time analytics to be feasible, efficient learning models need to strike a balance between prediction accuracy and response time. Table I illustrate the performance comparison of the proposed AEWAE method against other state-ofthe-art online adaptive learning methods presented
| 858 |
258615490
| 0 | 16 |
in Section II-B, including ARF-ADWIN [14], ARF-EDDM [14], OPA [12], SRP [15], KNN-ADWIN [11], HT [7], HATT [13], PWPAE [4], and MSANA [16]. As illustrated in Figures 3 and 4, as well as Table I, the four base models, ARF-ADWIN, ARF-EDDM, KNN-ADWIN, and OPA, achieve high accuracy while maintaining quite low latency on the CICIDS2017 and IoTID20 datasets. This demonstrates why they were chosen as base learners. C. Experimental Setup As depicted in Figure 3 and Table I, on the CICIDS2017 dataset, there are six concept drifts that occurred in the experiments. The proposed AEWAE outperforms all other evaluated online learning models when the α is set to 0.0856, achieving the highest accuracy of 99.69%, precision of 99.52%, recall of 98.57% and F1-score of 99.03%. Moreover, the inference time for AEWAE is a mere 3.19 ms per packet, which is faster than other ensemble techniques such as PWPAE and MSANA. The throughput of the proposed AEWAE model also remains impressively high at 313.16 samples per second. The evaluation of online learning models on the IoTID20 dataset, as depicted in Figure 4 and Table I, reveals that three minor drifts occurred during the early stages of the experiment. The results demonstrate that the proposed AEWAE model excels in comparison to other evaluated models when the α is set to 0.096. It achieves the highest accuracy at 99.72%, precision at 99.7%, recall at 100%, and an F1-score of 99.85%. Additionally, AEWAE's inference time is a mere 2.88 ms per packet, outpacing other ensemble techniques like PWPAE and MSANA.
| 859 |
258615490
| 0 | 16 |
The throughput of the proposed AEWAE model is also remarkably high, reaching 347.33 samples per second. Furthermore, to assess the efficiency and feasibility of the proposed framework, the probability density functions (PDFs) of the inference time/latency for the CICIDS2017 and IoTID20 datasets are presented in Figures 5 and 6, respectively. Figures 5 and 6 demonstrate that nearly all data samples were processed within 8 ms and 7 ms for the respective datasets, showcasing the effectiveness of the proposed framework. In conclusion, the proposed AEWAE framework has proven to be highly efficient and effective in processing data samples for both the CICIDS2017 and IoTID20 datasets. The experimental results highlight the effectiveness and robustness of the AEWAE model for IoT streaming data analytics. These findings demonstrate the feasibility of implementing the proposed framework in real-time environments. V. CONCLUSION The development of fifth-generation (5G) wireless networks has led to a notable growth in smart technology adoption based on the Internet of Things (IoT), providing great convenience to humans. However, as an essential aspect of smart technology, IoT systems for service delivery frequently encounter concept drift challenges in network data stream analytics, owing to the dynamic nature of IoT environments. This issue can lead to a decline in system performance. In this article, we propose a drift-adaptive framework designed to efficiently manage the processing of ever-changing IoT data streams while dynamically adapting to concept drifts. The proposed framework consists of IoT data preprocessing, base model learning, and online ensembling using a novel AEWAE approach with the PSO hyperparameter optimization method.
| 860 |
258615490
| 0 | 16 |
According to performance evaluations conducted on two benchmark IoT streaming datasets, IoTID20 and CICIDS2017, our proposed framework has demonstrated its effectiveness in processing dynamic IoT streams, achieving higher accuracy rates of 99.72% and 99.69%, respectively, compared to other state-ofthe-art methods. In the future work, the proposed framework will be applied to other domains, such as the Internet of Vehicles (IoV). Yafeng Wu was born in Nanjing, China. He received a B.S. degree of Software Engineering from Nantong University Xinglin College in 2021 and is currently pursuing an M.S. degree in network and information security engineering at Guangdong Polytechnic Normal University. His main research interests include network information security, IoT security, and machine learning.
| 861 |
258615490
| 0 | 16 |
PHYTOREMEDIATION OF SOILS CONTAMINATED WITH USED LUBRICATING OILS USING SUNFLOWER Lubricating oil is one of the most important derivatives from petrol industry, However its life cycle finished with the generation of the used oil, ham,ful to the environment and must be suitable disposed, A Brazilian law authorized the use of 30% of re-refined lubricating oil with the new oil, This market leads to the establishment of several new industries which use small and medium trucks to collect used oil at gas stations where consumers change the lubricant oil, On the last years an intense flux of these trucks can be observed at the main road of Brazil, which connects Rio de Janeiro and Sao Paulo, the two biggest cities of our country, This scenario can lead to possible accidents and oil spills, This paper relates the use of sunflower to remediate soils collect at the margin of this important road, It was tested soils contaminated with 2, 4, and 6%, The soils were extracted with Soxhlet extraction using cyclohexane and analyzed by gas chromatography with flame ionization detection, The first three months of the experiments indicated a phytoremediation of 86, 82, and 65% for the soils contaminated with 2, 4, and 6%, respectively, I INTRODUCTION Lubricating oil is one of the most important products from petrol industry, by its value, several uses, technical requirements, and developments in its fabrication to follow the advances in engines, However, its life cycle ends with the generation of a residue, very harmful to the environment otherwise not suitable dispose, with
| 862 |
221701809
| 0 | 16 |
metals and degraded organic and inorganic compounds in its composition, Data from the distributors companies of these products indicated that lubricating oil market is greater than one million of cubic meters per year in BraziLlt is estimated that 50% of the used lubricating oils can be collected to be reprocessed, once some types of this product can not be regenerated, as the soluble oils and others that is burned together with the fuels, As reported by the distributors these oil collection is near 50% of the available used oil at the great cities as the metropolitan area of Rio de Janeiro and Sao Paulo, This quantity lead to a difficult management of the collection process, since that this collection is made by small companies, and is very difficult to follow, In 1993, with the purpose to minimize the inadequate discharge of the used lubricating oil, the National Environment Council (CONAMA) enact a law (CONAMA 09/93) that establishes basic rules to the dispose of the used or contaminated oils, to avoid the environment impacts, This resolution was upgraded by the resolution CONAMA 464/2007, that still in validity. This resolution establish that the disposition of these oils must done by recycling using the re refining process, to remove contaminants, degradation products, unusable additives, conferring to the new products basic properties to be used as raw material to be used as a new lubrication oil in a limit of 30% in volume. This resolution also instruct the collection and destination activities of the used of contaminated oils, establishing the
| 863 |
221701809
| 0 | 16 |
responsibilities of the manufacturers, importers, resellers, and consumers of these products. It is unquestionable that used lubrication oils, when unsuitable discharged may lead to harmful consequences to the environment. Despite this Brazilian law is considered actual and more restrictive than similar of developed countries as France and Italy, some factors difficult the complete execution. The used lubricating oils contain deterioration products as oxygenated hydrocarbons ( organic acids and carbonyls compounds), polynuclear aromatic hydrocarbons (PAH), and resinous substances. Other inorganic compounds are also present as additives to adjust the oils properties as viscosity, density, superficial tension, detergent proprieties, among several others. Others compounds may be present as metals removed from the structural parts of engines, sealants, particulates and water. This scenario leads to a multi-contamination event that must be studied in detail. A recent search in the literature found no research in this area using the keywords "lubricating oil" and "phytoremediation". EXPERIMENTAL PHASE The vegetable used to remediate soils contaminated with lubricating oils was the Helianthus annus (sunflower). The soil tested was collected at five different places (20 km from each one) at the margin of the Presidente Dutra Highway, the most important highway in Brazil, which connect Sao Paulo do Rio de Janeiro. The collected soils were mixed and used in all experiments, and were classified as lato soils. The used lubricating oil was collect from the engines from 3 different automobiles and mixed. The experiments were conducted in an open place, with a roof to protect from rain. Eight experiments were done, in triplicate, totalizing
| 864 |
221701809
| 0 | 16 |
24 experiments, as follow: Experiment I, 2, and 3: only soil. Experiment 4, 5, and 6: soil with sunflower. Experiment 7,8,and 9: soil with 2% of used lubricating oil. Experiment 10, I 1, and I 2: soil with 4% of used lubricating oil, Experiment I 3,I 4,and 15: soil with 6% of used lubricating oil, Experiment 16, 17, and 18: soil with 2% of used lubricating oil and sunflower. The experiments were conducted during three months at Resende, a city located at the south of the Rio de Janeiro State (22 ° 29' S and 44 ° 28' W Gr) at 440 m of altitude. The region climate is dry tropical with 1200 mm of precipitation concentrated during October to March, The mean temperature of the experiment was 24,7 ° C and the humidity of7 I%, For each month, portions of I 0g of contaminated soil were removed to evaluate the process. The soil samples were put inside filter paper folders and the extraction were done using 200 ml of cyclohexane in a Soxhlet extractor, during 6 hours, After the extraction the samples were concentrated in an evaporative device to a volume of IO mL The chemical analyses were carried using a gas chromatography with a flame ionization detector (FID) (Agilent GC 6890), No identification or quantification of the contaminants was done, The evaluation of the phytoremediation process was done comparing the total area obtained in the respective chromatograms of each experiment A capillary column (DB-I, 30 m x 0.25 mm x 0.25 µm) was used
| 865 |
221701809
| 0 | 16 |
to separate the species and 1 Helium was used as carrier gas at a flow rate of 1.5 ml min-, The injection temperature was 300 ° C in splitless condition with 1.0 µl of injection volume. The oven temperature starts at 60 ° C and programmed to 320 ° C at I 0 ° C min-1 , and the FID was set to 340 ° C The soil samples were collected 30, 60, and 90 days after the beginning of the experiments, Sunflower plants were inserted into contaminated soils when they were with 8-10 cm of height RES UL TS AND DISCUSSION The experiments done only with soil were to investigate the possible presence of previous contaminants, The experiments with soil and used lubricating oil without sunflower were done to evaluate the loss of contaminants by volatilization. The results indicated a very low remediation in the first month, basically by the adaptation of the vegetable in the contaminated soiL After 60 days some remediation can be visualized in the range of 30%, but in 90 days the remediation is better than 60%, as can be seen at Figure At the first 60 days of the experiment it can be observed a greater remediation of the soil contaminated with higher oil content, This can be an indicative of a strong volatilization process promoted by the sunflower, After 90 days of the experiment the best results were obtained by the remediation of the soil contaminated with 2% of lubricating oil, The sunflower of the experiment with 6 %
| 866 |
221701809
| 0 | 16 |
of oil after 90 days, and in a minor extent the 4 %, were affected by these oil contents, and some damages were observed in the leafs_ A secondary result obtained by this work indicated that the remediation process is more intense on the heavier compounds, by the observation of a more intense reduction of the chromatograms area of the peaks obtained at the end of the chromatogram, It is interesting to observe that the results presented at Figure I are only a mean reduction of near 120 chromatogram peaks, Some compounds peak area did not experiment a reduction and some others increase, possible by the inter-conversions of one substance to other, Also some compounds experimented I 00% of reduction, These individual results are now being investigated and will be published soon, CONCLUSIONS This work is an initial research to investigate the possible contamination of used lubricating oil of soils from highways margins, A more careful set of experiments need to be done, as oil and soil characterization, and the sunflower or other specie must be follow in all its parts, as roots and leafs_ But these primary results indicated a good potential to use the Phytoremediation technique, A more detailed study must be done to evaluate the contamination limit Studies are also necessary to investigate the remediation of inorganic compounds, mainly the metals associated with the used oil, These metals probably are being remediated by phytoaccumulating process,
| 867 |
221701809
| 0 | 16 |
Ophiopogonin B-induced autophagy in non-small cell lung cancer cells via inhibition of the PI3K/Akt signaling pathway Ophiopogonin B (OP-B) is a bioactive component of Radix Ophiopogon Japonicus, which is often used in Chinese traditional medicine to treat pulmonary disease. However, whether or not OP-B has any potential antitumor activity has not been reported. Here, we show that the non-small cell lung cancer (NSCLC) cell lines NCI-H157 and NCI-H460 treated with OP-B grow more slowly and accumulate vacuoles in their cytoplasm compared to untreated control cells. Flow cytometric analysis showed that the cells were arrested in G0/G1 phase. Nuclear morphology, Annexin-V/PI staining, and expression of cleaved caspase-3 all confirm that OP-B does not induce apoptosis. Instead, based on results from both transmission electron microscopy (TEM) and the expression of microtubule-associated protein 1 light chain 3-II (LC3-II), we determined that OP-B treatment induced autophagy in both cell lines. Next, we examined the PI3K/Akt/mTOR signaling pathway and found that OP-B inhibited phosphorylation of Akt (Ser473, Thr308) in NCI-H157 cells and also inhibited several key components of the pathway in NCI-H460 cells, such as p-Akt(Ser473, Thr308), p-p70S6K (Thr389). Additionally, insulin-mediated activation of the PI3K/Akt/mTOR pathway provides evidence that activation of this pathway may correlate with induction of autophagy in H460 cells. Therefore, OP-B is a prospective inhibitor of PI3K/Akt and may be used as an alternative compound to treat NSCLC. Introduction Gefitinib and erlotinib, epidermal growth factor receptor tyrosine kinase inhibitors (EGFR-TKIs), have been widely used to treat NSCLC in the clinic. However, their efficacy has been limited by both
| 868 |
1144379
| 0 | 16 |
natural and acquired resistance. Autophagy is known as a type II programmed cell death. It has been found that cell death can occur concomitantly with features of autophagy, and excessive stimulation of autophagy through over-expression of beclin1 suppresses tumorigenesis (1,2). Autophagy is a multi-step process consisting of initiation, autophagosome formation (nucleation, elongation, and completion), maturation, and degradation (3). Autophagy initiation is complete with the accumulation of the ULK1/2-ATG13-FIP200 complex, which results in development of the isolation membrane, also known as a phagophore. The generation of the complex is regulated by mammalian target of rapamyacin (mTOR), which lies downstream of the class I phosphatidylinositol 3-kinase (PI3K)/Akt pathway. mTOR senses mitogenic stimuli, nutrient conditions, and ATP. The development of the autophagosome is dependent on the class III PI3K complex, which consists of the proteins Vps-34, beclin1, and p150, which all localize to the phagophore and recruit further autophagyrelated genes (ATGs) to allow for elongation and completion of the autophagosome. Once the auto phagosome is developed, its maturation is complete upon fusion with a lysosome to form an autophagolysosome (4,5). Constitutive activation of the PI3K/Akt pathway occurs in 90% of NSCLC cell lines, thus, promoting cell survival and resistance to chemotherapy or γ-irradation (6). As a result, inhibition of PI3K/Akt signaling is not only important for induction of autophagic cell death but also essential for finding new treatment for NSCLC. In our preliminary screening, OP-B was found to be effective in reducing the viability of a panel of human NSCLC cells. Further investigation of its anticancer mechanisms in NCI-H157
| 869 |
1144379
| 0 | 16 |
and H460 cells showed that OP-B primarily induces autophagy but not apoptosis. Examination of the PI3K/Akt/ mTOR signaling pathway showed that OP-B selectively inhibits phosphorylation of Akt both at Ser473 and Thr308 in both of the two cell lines, suggesting that OP-B may be a potential inhibitor of the PI3K/Akt pathway for the treatment of NSCLC. Materials and methods Materials and reagents. Ophiopogonin B was purchased from Nanjing Ze Lang medical technology company. The compound was initially dissolved in dimethyl sulfoxide (DMSO) (Sigma, USA) as a stock solution before use. For treatment of cells, it was diluted in culture medium to the appropriate concentrations, and the final concentration of DMSO was less than 0.01%. The chemicals used were rapamycin, LY294002 (Cell Signaling Technology), staurosporine, insulin, PI, Alamar blue, and Hoechst 33258 (Sigma). We also used the Alexa Fluor 488 Annexin-V/ Dead cell apoptosis kit (Invitrogen, USA). In vitro viability assay. Cells were seeded into 384-well plates using a Liquid dispenser in a bio-safety cabinet. Using the liquid handling system, cells were treated with drug the next day for 72 h. The final concentrations used in the assay were 50, 25, 12.5, 6.25, 3.125, 1.56, 0.78 and 0.39 µmol/l in triplicate. A volume of 5 µl/well Alamar blue was transferred into the assay plates for a final concentration of 10%. The plates were exposed to an excitation wavelength of 530 nm, and the emission at 590 nm was recorded to determine whether any of the test compounds fluoresce at the emission wavelength and thus interfere with the
| 870 |
1144379
| 0 | 16 |
assay. Plates were returned to the incubator and the fluorescence was read at 4 h. Data were calculated as the percentage of cell viability by the following formula: the percentage of cell viability = (At/As) x100%. At and As indicated the absorbance of the test substances and solvent control, respectively. The mean value and standard error for each treatment were determined and the % cell viability relative to control (0.01% DMSO) was calculated. The IC 50 is defined as the concentration of drug that kills 50% of the total cell population as compared to control cells at the end of the incubation period. Cell cycle analysis and apoptosis detection. Cells were treated with or without OP-B for 24 h, and then harvested by centrifugation, washed with ice-cold phosphate-buffered saline (PBS), and fixed in ice-cold 70% ethanol overnight. The cells were then treated with 40 µg/ml RNase at 37˚C and then stained with 40 µg/ml PI for 30 min. The percentage of cells in each phase (SubG1, G0/G1, S, and G2/M) was calculated (Becton Dickinson). For Hoechst 33258 nuclear staining, exponentially growing cells were seeded at a density of 10 5 cells per well onto heat-sterilized coverslips in 6-well plates. After attachment, cells were treated with or without 10 µmol/l OP-B or 1 µmol/l Staurosporine for 24 h. Subsequently, the cells were fixed (methanol: glacial acetic acid = 3:1) for 10 min, and then dyed (Hoechst 33258, 10 µg/ml) for another 10 min at 37˚C. After washing three times with PBS, the cells were observed under a
| 871 |
1144379
| 0 | 16 |
fluorescence microscope (Olympus, Japan). Apoptosis and dead cells induced by OP-B were assayed using the high content screening (HCS) Kinetic Scan Reader (ThermoFisher Scientific, USA). The principle of the assay is that cells are labeled with a cocktail of fluorescent dyes (including Hoechst 33258 and Alexa Fluor 488 Annexin-V) that indicate the cellular properties of interest, including nuclear structure, cell membrane permeability, and early and late apoptosis. All procedures were performed according to the manufacturer's instructions. The cells were plated at a density of 8x10 3 cells/well in each well of a 96-well plate. After culturing for 24 h, cells were incubated with 10 µmol/l OP-B for another 24 h. Thirty minutes before the completion of incubation, a cocktail of fluorescent dyes was added to each well. The cells were then fixed with prewarmed Fixation Solution and washed twice with PBS. Plates were then sealed and ran on an HCS Reader to acquire images. Images were analyzed with HCS software, and the fluorescence intensity of Hoechst 33258 and Annexin-V/PI were calculated. Transmission electron microscopy (TEM). After being exposed to 10 µmol/l OP-B for 48 h, the cells were trypsinized, washed with PBS and fixed in 2.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.2) overnight at 4˚C. The next day, cells were washed three times with 0.1 M phosphate buffer. Thereafter, the cells were fixed in 1% aqueous osmium, dehydrated with increasing concentrations of ethanol (30, 50, 70, 80, 90 and 100%), and embedded in araldite. The ultrathin sections were prepared with a microtome (Leica, Germany)
| 872 |
1144379
| 0 | 16 |
and mounted on copper grids. The samples were stained with 2% aqueous uranyl acetate and lead citrate and observed in a transmission electron microscope (Jeol, Japan). Statistical analysis. Unless otherwise stated, data were expressed as the mean ± SD, and analyzed by Student's t test. P-value <0.05 was considered statistically significant. Results Effect of OP-B on proliferation of NSCLC cells. NSCLC, including squamous carcinoma, adenocarcinoma, and large cell carcinoma, represents ~80-87% of all lung cancer cases (7). To determine whether OP-B (structure shown in Fig. 1A) has any therapeutic effect on NSCLC cells, we performed a cell viability assay using eleven human NSCLC cell lines. After 72 h of treatment, OP-B significantly decreased cell viability of all cell lines tested in a dose-dependent manner, and the IC 50 was <4 µmol/l (~3.87 µmol/l) (Fig. 1B). Effect of OP-B on cell morphology, cell cycle, and apoptosis in NCI-H157 and H460 cells. NCI-H157 and H460 cells represent the main subtypes of NSCLC and are derived from squamous cell carcinoma and large cell carcinoma, respectively. In our experiment, we found that these two cell lines were more sensitive to OP-B (IC 50 of H157 and H460 was 2.86 and 4.61 µmol/l, respectively) than all other NSCLC cell lines tested. Therefore, we chose these lines to further investigate the pharmacological effect of OP-B. Following 10 µmol/l OP-B treatment for 24 h, many vacuoles appeared in both cell lines but especially in H157 cells ( Fig. 2A). When the concentration of OP-B was increased to 20 µmol/l, the vacuoles became even
| 873 |
1144379
| 0 | 16 |
larger and occupied almost all of the space outside the nucleus in NCI-H157 cells. Flow cytometric analysis of cells stained with propidium iodide showed a mild increase in the cell population in G0-G1 phase after the cells were treated with different concentrations of OP-B for 24 h in medium containing 10% FBS (Fig. 2B). To further investigate whether the cell cycle arrest was associated with apoptosis, we measured levels of caspase-3 and Bcl-2. Treatment with staurosporine served as a positive control for cleaved caspase-3 staining. The results showed that there was no change compared to the negative control (Fig. 2C), and cleaved caspase-3 was only slightly increased in NCI-H460 cells following 48 h of treatment with OP-B (Fig. 2D); however, no change was detected in NCI-H157 cells (data not shown). Nuclear staining with Hoechst 33258 also showed no characteristics of apoptosis, such as cell shrinkage, nuclear condensation, and fragmentation (Fig. 2E). Furthermore, the cells labeled with a cocktail of fluorescent dyes (including Hoechst 33258 and Alexa Fluor 488 Annexin-V/Dead cell apoptosis kit) and scanned with high content screening (HCS) Kinetic Scan Reader (ThermoFisher Scientific) (Fig. 2F) also suggested that OP-B did not induce apoptosis or necrosis in either of the two cell lines. OP-B induces autophagy in NCI-H157 and H460 cells. Results from transmission electron microscopy (TEM) showed that the cytoplasmic vaculoes had double-layered membranes and that many of them contained cytoplasmic organelles or myelin figures. Furthermore, the vacuoles increased in size and number and fused into larger vacuoles, while the nucleus remained intact (Fig. 3A).
| 874 |
1144379
| 0 | 16 |
Detection of LC3 by immunoblotting showed that OP-B treatment increased the conversion of LC3-I to LC3-II in a dose-and time-dependent manner (Fig. 3B-E). Thus, we speculated that treatment with OP-B induced autophagy of NCI-H157 and H460 cells. Effects of OP-B on PI3K/Akt/mTOR/p70S6K signaling pathway and induction of autophagy in NCI-H157 and H460 cells. The PI3K/Akt/mTOR/p70S6K signaling pathway, which is often associated with tumorigenesis and activated in numerous tumors, is well-known to regulate autophagy (8-10). Thus, the pathway was examined in relation to OP-B-induced autophagy in NCI-H157 and H460 cells. As shown in Fig. 4A and B, when cells were treated with at least 10 µmol/l OP-B for at least 1h, p-Akt (Ser473) was significantly inhibited in both NCI-H157 and H460 cells, but p-p70S6K (Thr389) was inhibited only in H460 cells under the same conditions. Under all treatments, p-4EBP1 (Thr37/46) was not affected. LY294002 is a well-characterized inhibitor of PI3K, and rapamycin is an inhibitor of mTORC1. Next, we tested the effect of OP-B on the PI3K/Akt/mTOR/p70S6K pathway in both H157 and H460 cells. In H157 cells, similarly to LY294002, OP-B inhibited phosphorylation of Akt both at Ser473 and Thr308, and it weakened the feedback activation of rapamycin on p-Akt at Ser473. In contrast, p-PDK1 (Ser241), p-p70S6K (Thr389), and LC3, were not affected by any of the above treatments (Fig. 4C). Similarly, in H460 cells, OP-B showed enhanced inhibition of p-Akt (Ser473 and Thr308) after co-treatment with LY294002. Additionally, it weakened the feedback activation of rapamycin on the two sites. Unlike in H157 cells, the phosphorylation of
| 875 |
1144379
| 0 | 16 |
p70S6K (Thr389) was inhibited by treatment with OP-B, LY294002, or rapamycin. Conversion of LC3 І to LC3-II was induced by OP-B, inhibited by LY294002, and enhanced by co-treatment with OP-B and LY294002. In fact, co-treatment with OP-B and rapamycin had an even more significant effect than single treatment with OP-B or rapamycin alone (Fig. 4D and E). The above results show that in NCI-H157 and H460 cell lines, OP-B has similar pharmacological effects on the inhibition of p-Akt (Ser473and Thr308). However, there are varying degrees of inhibition on the PI3K/Akt/mTOR/p70S6K signaling pathway. Taken together, NCI-H460 was more sensitive to OP-B not only based on inhibition of the pathway but also based on induction of autophagy. Correlation between inhibition of the PI3K/Akt/mTOR/p70S6K pathway and induction of autophagy from OP-B in H460 cells. Insulin upregulates PI3K and its downstream targets, including Akt and mTOR, and also suppresses autophagy (11)(12)(13). As shown in Fig. 5B, 30 min of insulin treatment significantly phosphorylates Akt at Ser473 and p70S6K atThr389. In contrast, when cells were pretreated with OP-B and then stimulated with insulin, the phosphorylation was significantly inhibited. Otherwise, no significant differences in the LC3-II/ actin ratio between OP-B treatment and OP-B treatment with insulin were observed (Fig. 5). Discussion OP-B is a natural active compound extracted from the Chinese herbal medicine ophiopogon. In this study, we found that OP-B successfully inhibited cell proliferation in a panel of NSCLC cell lines. The IC 50 in all lines tested was <4 µmol/l. In order to further investigate the pharmacological effect of OP-B
| 876 |
1144379
| 0 | 16 |
on NSCLC, we chose NCI-H157 and NCI-H460 cells as our cell line models since they represent the most commonly used NSCLC cells and because they were all sensitive to OP-B in our preliminary studies. Of note, we found that after 24 h of treatment with at least 10 µmol/l OP-B in medium containing 10% FBS, a large number of vacuoles accumulated in the cytoplasm of cells. From the volume of the vacuoles, we judged that NCI-H157 seemed to be more sensitive to OP-B than NCI-H460 ( Fig. 2A). Cell cycle analysis by flow cytometry showed that OP-B induced a modest increase in G0/G1 phase in both cell lines (Fig. 2B). However, expression of caspase-3 and Bcl-2, detected by western blot (Fig. 2C), nuclear morphology (stained by Hoechst 33258), and fluorescence intensity (labeled by AnnexinV/PI) detected by high content screening (HCS) Kinetic Scan Reader (Fig. 2F) all showed that OP-B did not induce apoptosis or necrosis in either cell line. In order to determine if the vacuoles were associated with autophagy, we measured several markers. Importantly, TEM is able to distinguish autophagic cytoplasmic vacuoles from cellular vesicles, such as endosomes, lysosomes, and apoptotic blebs (14). We first observed cell morphology. The vacuoles had double membranes, the internal contents were degraded by lysosomal hydrolases, and only some myelin figures remained (Fig. 3A). These results are consistent with earlier publications (15). The presence of LC3 in autophagosomes, and the conversion of LC3 I to LC3-II are known indicators of autophagy (16). Detection of LC3 using western blot showed that
| 877 |
1144379
| 0 | 16 |
in both cell lines, the OP-B increased the conversion of LC3 I to II in a dose-and time-dependent manner ( Fig. 3B and D). Unexpectedly, the conversion rate of LC3 I to II was more significant in NCI-H460 than in H157 (Fig. 3C and E). To further investigate the reason behind this, we assessed the PI3K/Akt/mTOR signaling pathway, which is the main pathway involved in the regulation of autophagy. Within 4 h of OP-B treatment in NCI-H157 cells, p-Akt was inhibited and autophagy was not induced. However, in H460 cells, p-PDK1, p-Akt, and p-p70S6K were all inhibited by OP-B and autophagy was induced ( Fig. 4C and D). In addition, activation of the pathway using Insulin showed that the autophagy induced by OP-B correlated with an active signaling pathway (Fig. 5). Since activation of PI3K/Akt occurs in 90% of NSCLC cell lines, it has become an important target for the development of anticancer drugs. It is well known that LY294002 A and B) The NCI-H157 and H460 cells treated with 0, 2.5, 5, 10 µmol/l OP-B for 1.5 h or 10 µmol/l of OP-B for 0, 0.5, 1 or 2 h were analyzed by immunoblotting with antibodies against p-Akt (Ser473), p-p70S6K (Thr389), p-4EBP1 (Thr37/46), Akt, p70S6K, 4EBP1, and actin. (C and D) The NCI-H157 and H460 treated with 10 µmol/l OP-B, 10 µmol/l LY294002 or 10 µmol/l Rapmycin for 4 h were analyzed by immunoblotting with antibodies against p-PDK1 (Ser241), p-Akt (Thr308), p-Akt (Ser473), p-p70S6K (Thr389), LC3, and actin. (E) Densitometry analysis of LC3-II levels relative
| 878 |
1144379
| 0 | 16 |
to actin in H460 cells was performed using three independent experiments. Error bars, SD; ** p<0.01; *** p<0.001. and rapamycin are inhibitors of PI3K and mTORC1, respectively. However, these compounds have some drawbacks. For example, LY294002 does not distinguish between class I and class III PI3K, and its inhibition of class III PI3K also inhibits autophagy (17)(18)(19). Rapamycin inhibits mTORC1, but has a negative feedback on Akt (20). Herein, we found that at least in NCI-H460 cells, OP-B was an ideal inhibitor of the PI3K/ Akt/mTOR/p70S6K pathway. It inhibited all components of the pathway and even had a synergistic effect with LY294002 on Akt. It also decreased the activation of rapamycin on Akt and had a synergistic effect on induction of autophagy. Taken together, OP-B displayed significant cytotoxicity on a panel of NSCLC cell lines at a relatively low concentration. In NCI H157 and H460 cells, it inhibited p-Akt both at Ser308 and Thr473 and significantly induced autophagy. In NCI-H460 cells, it inhibited the PI3K/Akt/mTOR/p70S6K pathway more thoroughly than in H157 cells. Thus, we speculate that OP-B may be an alternative agent in the classification and treatment of NSCLC.
| 879 |
1144379
| 0 | 16 |
Phenylmethimazole abrogates diet-induced inflammation, glucose intolerance and NAFLD Nonalcoholic fatty liver disease (NAFLD) is the hepatic manifestation of both metabolic and inflammatory diseases and has become the leading chronic liver disease worldwide. High-fat (HF) diets promote an increased uptake and storage of free fatty acids (FFAs) and triglycerides (TGs) in hepatocytes, which initiates steatosis and induces lipotoxicity, inflammation and insulin resistance. Activation and signaling of Toll-like receptor 4 (TLR4) by FFAs induces inflammation evident in NAFLD and insulin resistance. Currently, there are no effective treatments to specifically target inflammation associated with this disease. We have established the efficacy of phenylmethimazole (C10) to prevent lipopolysaccharide and palmitate-induced TLR4 signaling. Because TLR4 is a key mediator in pro-inflammatory responses, it is a potential therapeutic target for NAFLD. Here, we show that treatment with C10 inhibits HF diet-induced inflammation in both liver and mesenteric adipose tissue measured by a decrease in mRNA levels of pro-inflammatory cytokines. Additionally, C10 treatment improves glucose tolerance and hepatic steatosis despite the development of obesity due to HF diet feeding. Administration of C10 after 16 weeks of HF diet feeding reversed glucose intolerance, hepatic inflammation, and improved hepatic steatosis. Thus, our findings establish C10 as a potential therapeutic for the treatment of NAFLD. Introduction Obesity is the single most important risk factor for the development of nonalcoholic fatty liver disease (NAFLD), which is the most prevalent liver disease in the western hemisphere (Lazo & Clark 2008, Bellentani et al. Kanwal 2014. NASH is associated with increased mortality not only from vascular disease but also
| 880 |
4954767
| 0 | 16 |
from complications of cirrhosis and hepatocellular cancer. Thus, targeting hepatocellular inflammation is expected to significantly prevent the progression of the disease and reduce mortality in patients with NAFLD (Younossi et al. 2011). Activation of TLR4-mediated inflammation also exacerbates hepatic lipid accumulation, although the exact mechanism is still unknown. Mice deficient in TLR4 demonstrate HF diet-induced weight gain but are protected against inflammation, hepatic steatosis and insulin resistance (Shi et al. 2006, Suganami et al. 2007, Tsukumo et al. 2007, Davis et al. 2008, Spruss et al. 2009, Pierre et al. 2013, Jia et al. 2014, Ferreira et al. 2015. Liver-specific TLR4-knockout (TLR4 LKO ) mice become obese when placed on a HF diet but remain insulin sensitive and are protected from the development of steatosis (Jia et al. 2014). The attenuation of steatosis and insulin resistance is most likely due to reduced proinflammatory gene expression in liver and adipose tissue of both global and liver-specific TLR4-deficient mice (Jia et al. 2014). Even with the acknowledged epidemic of obesity and associated NAFLD, there is an overwhelming failure (1) to clinically recognize the disease in the early stages due to the lack of specific diagnostic indicators or (2) to initiate treatment as there are no effective medications which specifically attenuate the early systemic inflammatory processes of NAFLD. This leaves patients and physicians only with long-term weight loss through diet to treat NAFLD, which is effective but very difficult to sustain (Gelli et al. 2017), or bariatric surgery, which can be very effective at reducing hepatic fat content (Hannah
| 881 |
4954767
| 0 | 16 |
& Harrison 2016, Schwenger et al. 2018) but can have significant associated complications (Chang et al. 2017). In light of this and studies suggesting a direct involvement of TLR4-mediated inflammation in the development of HF diet-induced hepatic steatosis and insulin resistance, there is a concerted effort directed at developing therapeutics targeting TLR4 signaling. We have developed a library of small-molecule inhibitors of inflammation that potently block TLR signaling, including FFA-and gut-derived LPSinduced TLR4 signaling (Harii et al. 2005, McCall et al. 2007, Schwartz et al. 2009, Deosarkar et al. 2014. Our lead compound, phenylmethimazole (C10), is a derivative of methimazole that inhibits inflammation resulting from TLR3 and TLR4 signaling in both immune and non-immune cells by blocking homodimerization of IRF3 and thus blocking its nuclear translocation and transcriptional activation activity (Courreges et al. 2012). Thus, we hypothesized that C10 will prevent and/or reverse HF diet-induced hepatic and adipose tissue Phenylmethimazole (C10) (Concord Biosciences, Cleveland, OH, USA) was prepared as a 200 mM stock solution in 100% (v/v) DMSO (Sigma-Aldrich) and further diluted to achieve the working concentration indicated in individual experiments. Mice and experimental design This work was conducted with approval from the Ohio University Institutional Animal Care and Use Committee in accord with accepted standards of humane animal care. Experimental procedures Six-week-old C57BL/6J male mice were purchased from Jackson Labs and housed 4 per cage in an environment controlled for temperature (18-22°C) and humidity on a 14:10-h light/darkness cycle. Mice were allowed to acclimate for 1 week prior to diet placement and C10 and control
| 882 |
4954767
| 0 | 16 |
treatments. Reversal study Prior to the start of the experiment, mice were randomly assigned to a diet group: LF diet group or HF diet group. After 16 weeks on their respective diet, an IPGTT was performed to evaluate glucose tolerance in each mouse. Any mouse in the LF diet group that was glucose intolerant and any mouse in the HF diet group that were glucose tolerant were removed from the study. Inclusion/exclusion criteria were as follows: If the IPGTT curve for a HF diet-fed mouse was identical or very similar to that of the LF diet-fed group, it was excluded. Similarly, if the IPGTT curve for a LF diet-fed mouse was identical or very similar to that of the HF diet-fed group, it was excluded. One week following the IPGTTs (i.e. after 17 weeks on respective diets), mice fed the HF diet were randomly assigned to a treatment group; HF diet + sham injection, HF diet + DMSO, HF diet + 1 mg/kg C10 in 10% DMSO and PBS. Mice received once daily IP injections for 14 weeks. Weights were recorded weekly. Body composition was obtained as described earlier. Another IPGTT was performed on mice after 12 weeks of C10 or control treatments. Intraperitoneal glucose tolerance tests (IPGTTs) Intraperitoneal glucose tolerance tests (IPGTTs) were performed on 12-h fasted mice. Body weight and blood glucose (Freestyle Freedom Blood Glucose monitoring System, Abbott Laboratories) was measured prior to IP injection of glucose (Sigma-Aldrich) (1-2 g/kg body weight). Subsequent blood glucose measurements were performed at time 0 and at
| 883 |
4954767
| 0 | 16 |
20/30, 60, 90, 120 and 180 min post IP injection of the glucose bolus. Histological analysis For microscopic examination of liver morphology and steatosis, liver tissue was fixed in 10% buffered formalin for 12-24 h. Formalin-fixed tissues were dehydrated in ethanol and embedded in paraffin for hematoxylin and eosin staining. Liver sections for histological staining were cut to 5 µm. Tissue preparation for histological analysis was performed by Ohio University Heritage College of Osteopathic Medicine Histological Core Services. Hepatic TG quantification Hepatic TG content was evaluated using a protocol based on the Salmon and Flatt method of lipid saponification (Salmon & Flatt 1985, List et al. 2009). Glycerol concentration was plotted against absorbance. The concentration of glycerol (mg glycerol/g tissue) was calculated by multiplying the determined concentration from the equation of the graph by the dilution factor and the number 5.31. The number (5.31) was used to correct the conversion of glycerol to TG by units of glycerol (mg/dL) to units of TG (mmol/L) and to mg/g tissue. TG content in AML-12 cells was presented as a ratio of total protein determined by BCA. Serum TG and total cholesterol Serum TG and total cholesterol were measured using blood collected at the experimental endpoint from non-fasted mice. The commercially available colorimetric Triglyceride Quantification Assay Kit (Abcam, cat #ab65336) was performed according to the manufacturer protocol to quantify serum TG. The commercially available colorimetric Cholesterol/Cholesteryl Ester Quantitation Assay Kit (Abcam, cat #ab65359) was used according to the manufacturer protocol to measure serum total cholesterol. Statistical analysis Statistical analysis
| 884 |
4954767
| 0 | 16 |
was performed using GraphPad Prism 7 for Mac. Statistical differences were determined using a one-way or two-way ANOVA followed by a Tukey-Kramer or Bonferroni test for post hoc comparison. Figure 1 C10 inhibits hepatic inflammation in addition to triglyceride accumulation in cell culture. AML-12 and HepG2 cells were treated with 100 µM C10 or DMSO (control) to determine if C10 could prevent hepatic inflammation in the presence of 0.75 mM palmitate or 10 ng/mL LPS. Treatment with C10 prevented palmitate-(A) and LPS (B)-induced pro-inflammatory cytokine (Ifnb1 and Tnfa) expression. Inhibition of palmitate-induced pro-inflammatory cytokine expression was also observed in HepG2 cells (C). Treatment with C10 prevented palmitateinduced triglyceride accumulation in AML-12 cells (D). Bars indicate mean + s.e.m. Significance was determined using a one-way ANOVA followed by Tukey's post hoc analysis for multiple comparison; *P < 0.05 between Untreated and Palmitate + C10 treated groups compared to both Palmitate and Palmitate + DMSO groups (A, C and D) or P < 0.05 between Untreated and LPS + C10 treated groups compared to LPS and LPS + DMSO groups (B). # Different from LPS, P < 0.05. + Different from Palmitate and Palmitate + DMSO, P < 0.05 (D). C10 halts HFD-induced inflammation and NAFLD C10 prevents FFA-and LPS-induced inflammation in both murine and human hepatocytes in culture and TG accumulation in murine hepatocytes In previous studies, we have shown that C10 prevents palmitate-and LPS-induced pro-inflammatory cytokine expression in murine macrophages (RAW264.7 cells) and differentiated 3T3-L1 adipocytes by inhibiting TLR4 signaling, specifically by blocking transcriptional activity of
| 885 |
4954767
| 0 | 16 |
IRF3 (McCall et al. 2010). In HFD-induced NAFLD, TLR4 expressed in hepatocytes is activated by both FFAs and LPS (Matsumura et al. 2000, Reyna et al. 2008. Stimulation of the MyD88-dependent pathway leads to pro-inflammatory cytokine expression, specifically Tnfa and Il6, while activation of the MyD88-independent TLR4 pathway leads to direct upregulation of type 1 interferons (Ifnb1) and indirect upregulation of Tnfa (Paik et al. 2003, Takeda et al. 2003, Shi et al. 2006, Miura et al. 2013, O'Neill et al. 2013. As indicated in Fig. 1A and B, treatment with palmitate and LPS leads to the upregulation of Ifnb1 and Tnfa in murine hepatocytes (AML-12 cells) compared to the untreated control groups and C10 prevents LPS-and palmitate-induced upregulation C10 does not prevent weight gain or an increase in fat mass due to HF-feeding. Seven-week old C57BL/6J male mice were fed either LF or HF diet and treated once daily with sham, DMSO, or C10 intraperitoneal injection for 18 weeks. Total body weights were measured weekly and body composition was measured every 2 weeks for the duration of the study. Adipose (mesenteric, subcutaneous, epididymal, and retroperitoneal) tissue weight was measured after tissue harvest at 18 weeks. (A) HF diet feeding promoted a marked increase in body weight when compared to LF-fed mice. (B) Additionally, % Fat mass was increased in HF-fed mice when compared to LF-fed controls. Percent Lean Mass was increased in LF-fed mice when compared to HF-fed groups. Percent Fluid Mass was no different between LF-and HF-fed mice. (C) HF-fed mice displayed increased adipose
| 886 |
4954767
| 0 | 16 |
tissue weights after 18 weeks on HF diet when compared to LF-fed mice. Data points on line graphs (A and B) indicate mean and error bars indicate +/− s.e.m. and bars on bar graphs (C) indicate mean + s.e.m. Significance was determined using ANOVA followed by Tukey's post hoc analysis for multiple comparison; *Different from HF-fed groups; P < 0.05, n = 8. C10 halts HFD-induced inflammation and NAFLD A Patton et al. 237:3 Journal of Endocrinology of Ifnb1 and Tnfa. Treatment with C10 also prevents palmitate-induced pro-inflammatory cytokine expression in HepG2 cells, a human hepatocellular carcinoma cell line (Fig. 1C). The solvent control, DMSO, is known to exhibit anti-inflammatory effects by repressing pro-inflammatory cytokine production (Elisia et al. 2016). Although DMSO had some anti-inflammatory activity, we show that C10 has a greater anti-inflammatory effect by inhibiting proinflammatory cytokine production when compared to the palmitate-and LPS-stimulated DMSO groups. Inflammation is associated with enhanced hepatic de novo lipogenesis and TG accumulation (Feingold & Grunfeld 1987, Grunfeld et al. 1988, 1991, Feingold et al. 1990, 1992. Exogenous Tnfa in mice and rats has caused increased TG production and storage in the liver (Feingold & Grunfeld 1987, Feingold et al. 1990). In addition to preventing FFAand LPS-induced pro-inflammatory cytokine expression in vitro, C10 also reduced palmitate-mediated accumulation of TG in mouse hepatocytes (AML-12 cells) (Fig. 1D). C10 does not affect weight gain, body composition or adipose weight in HF diet-fed C57BL/6J male mice It is already known that TLR4-deficient mice develop obesity when fed a HF diet. Thus, we
| 887 |
4954767
| 0 | 16 |
were interested in the effect of C10 treatment on weight and body composition in a HF diet-induced model of obesity (DIO model). Seven-week-old C57BL6/J male mice were fed either a LF diet (10% fat) or HF diet (60% fat). Mice were dosed once daily with intraperitoneal (IP) sham injections, IP injections of DMSO (vehicle) or IP injections of 1 mg/kg C10 for 18 weeks. During the 18-week study (Prevention Study), mice were evaluated for the development of obesity by measurement of body weight and body composition. A HF diet challenge resulted in significantly more weight gain compared to the LF-fed mice ( Fig. 2A). Body composition revealed increased fat mass as a percentage of total body weight in the HF-fed mice when compared to the LF-fed mice (Fig. 2B). Obesity was also assessed by adipose tissue weight. At the end of the 18-week study, HF-fed mice had increased adipose tissue weight in mesenteric, subcutaneous, epididymal and retroperitoneal depots when compared to the LF sham group (Fig. 2C). C10 and DMSO treatment had no significant effect on weight, body composition or adipose tissue weight of HF-fed mice (Fig. 2). C10 blocks hepatic TG deposition in HF diet-fed C57BL/6J male mice Our in vitro experiments demonstrated that C10 prevented palmitate-induced TG accumulation in mouse hepatocytes (Fig. 1D). Thus, we sought to determine if C10 could prevent C10 prevents HF diet-induced hepatic steatosis. Hematoxylin and eosin staining was performed on liver tissue sections prepared after 18 weeks of HF diet feeding. Liver triglyceride content was determined by biochemical analysis
| 888 |
4954767
| 0 | 16 |
(A) Histological examination revealed that C10 prevents hepatic lipid accumulation. All images in (A) were taken at 400× magnification. Scale bar, 40 µm. (B) Treatment with C10 decreased hepatic triglyceride content when compared to HF-fed sham and DMSO groups but had no effect on serum triglyceride levels. Dotted lines represent the mean and error bars indicate + s.e.m. Significance was determined using ANOVA followed by Tukey's post hoc analysis for multiple comparison; *Different from HF-fed groups; P < 0.05. # Different from HF sham and HF DMSO groups; P < 0.05. hepatic TG accumulation in vivo. After 18 weeks of HF diet feeding and C10 treatment (Prevention Study), histological examination of the liver revealed the absence of steatosis in C10-treated mice when compared to livers of the HF sham and HF DMSO control mice. (Fig. 3A). To further quantify C10 inhibition of hepatic lipid accumulation, TG content was quantified biochemically which revealed increased TG content from liver samples of HF-fed mice when compared to LF-fed mice. Treatment with C10 reduced hepatic TG content compared to HF sham and HF DMSO groups (Fig. 3B). There was no difference noted among adipose depot weights of HF diet groups; however, the difference in hepatic TG content in the liver of C10-treated mice when compared to the HF controls indicate that C10 may have a localized effect on hepatic lipid metabolism. Serum TG and cholesterol levels were measured in non-fasted mice after 16 weeks of HF diet feeding ( Fig. 4A and B, respectively). Total serum TG remained unchanged (Fig.
| 889 |
4954767
| 0 | 16 |
4A) while total serum cholesterol was elevated in HF diet-fed mice when compared to LF sham controls; C10 did not affect serum cholesterol levels in HF-fed mice (Fig. 4B). C10 protects C57BL/6J male mice from HF diet-induced glucose intolerance Ectopic fat deposition in insulin target tissues impairs the function of insulin signaling and thus impairs glucose homeostasis (i.e. induces glucose intolerance/insulin resistance). Previous in vitro studies have shown that treatment with C10 prevent palmitate-induced IRS1 serine 307 phosphorylation, a process known to mediate insulin resistance in insulin-stimulated 3T3L1 adipocytes (McCall et al. 2010). To determine the effect of C10 on glucose tolerance in HF diet-fed mice, we performed a 2-h intraperitoneal glucose tolerance test (IPGTT) 13 weeks post diet and treatment initiation. The HF-fed control animals exhibited glucose intolerance, with significantly higher glucose levels at 20, 60, 90, 120 and 180 min following glucose administration during the IPGTT compared to LF-fed mice (Fig. 5), and the area under the curve (AUC) was significantly higher in the HF-fed mice compared to LF-fed mice (Fig. 5). HF-fed mice receiving C10 treatment had improved glucose tolerance despite their obesity as compared to the HF sham and HF DMSO-treated mice (Fig. 5). C10 prevents HF diet-induced inflammation in liver and mesenteric adipose tissue from C57BL/6J male mice Inflammation, specifically due to cytokines and chemokines produced by FFA and gut-derived LPS activation of TLR4 signaling, leads to systemic glucose intolerance by impairing insulin signaling in target tissues including adipose and liver. Activation of the TLR4 signaling pathways leads to expression of
| 890 |
4954767
| 0 | 16 |
proinflammatory cytokines, in particular, Tnfa and type 1 interferons (Ifnb1). Our model of HF diet feeding promotes an increase in circulating FFAs, gut-derived LPS, as well as an increase in fat deposition and accumulation in adipose tissue and the liver (Fraulob et al. 2010). Pathologic exposure of adipose and liver tissue to FFAs and gut-derived LPS activates TLR4 signaling and induces a local inflammatory tissue response marked by an increase in pro-inflammatory Total serum cholesterol is elevated in HF diet-fed mice, but serum triglyceride is unchanged. Total serum triglyceride and cholesterol were measured from non-fasted mice after 16 days of diet challenge and/or C10 treatment using commercially available colorimetric assays. (A) Serum triglyceride levels were unaffected by HF diet feeding and C10 treatment. (B) Total serum cholesterol was elevated in HF-fed mice when compared to LF sham controls. Dotted lines represent the mean and error bars indicate + s.e.m. Significance was determined using ANOVA followed by Tukey's post hoc analysis for multiple comparison; *Different from HF-fed groups; P < 0.05, n = 6-8. C10 halts HFD-induced inflammation and NAFLD A Patton et al. 237:3 Journal of Endocrinology cytokine gene expression. We observed that palmitate treatment induces expression of Tnfa and Ifnb1 in AML-12 cells, which was inhibited by treatment with C10 ( Fig. 1). Additional studies in our laboratory have shown that C10 prevents transcriptional activity of NFKB and IRF3 thereby inhibiting the upregulation of cytokine and chemokine production (McCall et al. 2010, Deosarkar et al. 2014. To determine the efficacy of C10 to prevent HF
| 891 |
4954767
| 0 | 16 |
diet-induced inflammation, adipose and liver tissue were collected from the mice in this study for analysis of pro-inflammatory gene expression. In both adipose and liver tissue, there was a significant increase in Tnfa and Ifnb1 expression in our HF sham and HF DMSO groups (Fig. 6). However, as in our in vitro studies, this HF diet-mediated increase in Tnfa and Ifnb1 mRNA levels was inhibited in both liver and adipose tissues from the HF diet-fed C10-treated animals ( Fig. 6A and B, respectively). Additionally, mRNA levels of F4/80, which encodes a cell surface macrophage marker remained low in adipose tissue of C10-treated mice despite HF diet feeding (Fig. 6C). C10 reverses HF diet-induced glucose intolerance, hepatic and adipose inflammation and hepatic steatosis The observation that C10 prevents HF diet-induced hepatic steatosis and inflammation as well as adipose inflammation and glucose intolerance in our DIO mouse model led us to question if C10 could reverse established glucose intolerance, hepatic steatosis and hepatic inflammation in these mice. To address this, male C57BL/6J mice were put on either a LF diet or a C10-treated mice maintain glucose tolerance despite obesity. A 3 h intraperitoneal glucose tolerance test was performed after 13 weeks of HF diet feeding. C10 prevented HF diet-induced glucose intolerance. Area under the curve was significantly lower in the LF sham group as well as the HF C10-treated group throughout the duration of the glucose challenge. Data points on the line graph indicate mean and error bars indicate +/− s.e.m. and bars on the bar graph indicate
| 892 |
4954767
| 0 | 16 |
mean + s.e.m. Significance was determined using ANOVA followed by Tukey's post hoc analysis for multiple comparison; *Different from HF sham and HF DMSO groups; P < 0.05, n = 8. Figure 6 C10 prevents HF diet-induced inflammation in vivo. Inflammatory gene expression was measured in liver and mesenteric adipose tissue after 18 weeks of HF diet feeding. Hepatic Ifnb1 and Tnfa expression were reduced in C10-treated mice when compared to HF sham and DMSO groups (A). Additionally, C10 prevented an upregulation of Ifnb1 and Tnfa in mesenteric adipose tissue (B) as well as Emr1(F4/80), a macrophage marker (C). Bars indicate mean + s.e.m. Significance was determined using ANOVA followed by Tukey's post hoc analysis for multiple comparison; *Different from HF sham and HF DMSO groups; P < 0.05, n = 7-8. HF diet for 16 weeks after which glucose tolerance was evaluated in each mouse via IPGTT. Mice on the LF diet that were glucose tolerant remained in the study and were maintained on the LF diet for the duration of the experiment. Mice on the HF diet that were glucose intolerant remained in the study and were maintained on the HF diet for the remainder of the study. As can be seen in Figs 7 and 8, the LF-fed mice weighed significantly less than the HF-fed mice (Fig. 7A, Week 0) and the IPGTT revealed that the LF-fed mice were glucose tolerant, whereas the HF-fed mice were glucose intolerant (Fig. 8A). One week later (after 17 weeks on the diets), the glucose-intolerant HF-fed mice
| 893 |
4954767
| 0 | 16 |
were randomly divided into HF sham, HF DMSO and HF C10 (1 mg/kg) treatment groups as described earlier for the 'prevention study' and were treated as indicated for 14 weeks. There was no difference in weights between the HF-fed treatment groups (Fig. 7A), although all HF-fed groups continued to gain weight and become more obese over the course of the 14-week treatment period (Fig. 7); however, the HF-fed C10-treated mice were now glucose tolerant (Fig. 8B), indicating that despite a continued rise in obesity, C10 reversed the glucose intolerance that was present in the mice prior to C10 treatment ( Fig. 8A and B). Hepatic steatosis (Fig. 9A) was also reduced as well as total serum cholesterol (Fig. 9B). Serum TG remained unchanged (Fig. 9B). Moreover, hepatic inflammation (Tnfa, Ifnb1 and Il6) (Fig. 9C) was dramatically reduced in C10-treated mice compared to the HF controls. Figure 7 C10 did not reverse body weight increase due to HF diet feeding. Weeks 0-14 are represented as the start and end points of the C10 'reversal study' post 16 weeks of HF or LF diet feeding. Total body weight was measured weekly and body composition was measured every 2 weeks during the duration of the study. At the end of the study, adipose tissue weight was measured. Discussion Inflammation is widely recognized as a key factor in the pathogenesis of metabolic diseases, specifically obesityrelated diseases such as T2DM and NAFLD. Obesity alone is considered the most important risk factor for development of NAFLD and is the driver of inflammation
| 894 |
4954767
| 0 | 16 |
in this disease that is responsible for its progression (Wild et al. 2004, Dabelea et al. 2014, Imes & Burke 2014. The major inflammatory signaling pathway in chronic inflammation in a state of obesity is TLR4 (Davis et al. 2008, Pierre et al. 2013, Jia et al. 2014, Sawada et al. 2014. TLR4 is abundantly expressed in insulin target tissues such as adipose tissue, liver and skeletal muscle and is now accepted as a key player in obesity-induced insulin resistance and T2DM (Jialal et al. 2014). Earlier studies suggested that the stimulation of TLR4 seen in obesity/ insulin resistance/T2DM results from gut-derived LPS (Lassenius et al. 2011, Jayashree et al. 2014, Velloso et al. 2015; however, it is now evident that FFAs derived from HF diets can also trigger TLR4 signaling in these target tissues (Reyna et al. 2008, Kim et al. 2012 leading to NAFLD and insulin resistance. In our model, we used a HF diet to promote the development of insulin resistance and hepatic steatosis. HF diets increase circulating levels of FFAs, which deposit in adipose tissue and other tissues such as the liver. Acceleration of hepatic FFA deposition occurs in obesityinduced NAFLD due to an increase in dietary fatty acids or lipolysis of adipose tissue. We hypothesized that C57BL/6J mice fed a HF diet would develop hepatic inflammation, steatosis and insulin resistance, which would be prevented and/or reversed with C10 treatment. Human and mouse hepatocyte cell lines demonstrated an inflammatory response when exposed to LPS and FFAs. In our in vitro system, C10
| 895 |
4954767
| 0 | 16 |
exhibited potent anti-inflammatory properties by preventing FFA-and LPS-induced pro-inflammatory cytokine expression measured by a reduction in Tnfa and Ifnb1 mRNA levels in hepatocytes in culture. Similar findings were observed in vivo as pro-inflammatory cytokine expression was also significantly reduced by C10 in liver and mesenteric adipose tissue of mice fed a HF diet. In addition to anti-inflammatory effects, C10 treatment prevented glucose intolerance (an indirect measure of insulin resistance) and hepatic steatosis in mice fed a HF diet. Inhibition of FFA-induced hepatic lipid accumulation by C10 treatment was also observed in vitro. Furthermore, and most clinically relevant, HF diet-induced insulin resistance was reversed by C10 intervention. C10-treated mice also had significantly reduced hepatic inflammation and decreased hepatic TG content, albeit the latter effect was not as dramatic as was observed in the 'prevention study'. The modest effect of C10 on hepatic TG content in the 'reversal study' may be due to the fact that the HF diet used in this study induced an overwhelming amount of hepatic steatosis due to the HF diet containing 60% fat. If the C10 treatment had continued for a longer duration or the diet changed to regular chow, we anticipate this effect would be more pronounced, especially given the C10 treatment reverses HF diet-induced glucose intolerance. A 3 h intraperitoneal glucose tolerance test was performed just prior to the initiation of treatment (A) and after 12 weeks from the start date of the reversal study (B). (A) At the beginning of the study all HF-fed mice were glucose intolerant compared to
| 896 |
4954767
| 0 | 16 |
LF-fed mice. (B) Following treatment, blood glucose levels remained elevated in the HF sham and HF DMSO groups when compared to the LF sham and HF C10-treated mice. Area under the curve was significantly lower in the C10-treated group when compared to the HF sham and DMSO groups and was nearly indistinguishable from the LF sham group. This indicates that C10 reverses glucose intolerance due to the HF diet feeding. Data points on line graphs indicate mean and error bars indicate +/− s.e.m. and bars on bar graphs indicate mean + s.e.m. Significance was determined using ANOVA followed by Tukey's post hoc analysis for multiple comparison; (A) * Different from LF sham, (B) * Different from HF sham and DMSO groups; P < 0.05, n = 5-6. fact that hepatic inflammation and insulin resistance was significantly reduced. In this regard, we have previously shown that continued HF feeding after intensive insulin therapy, in this same mouse model, prevents the 'Legacy Effect' of early insulin treatment in new-onset T2DM (Guo et al. 2015). While not evaluated in this study, it would be of interest in future studies to see if other models of NAFLD (e.g. ob/ob or db/db mice) also respond similarly to C10 treatment. Currently, there are no therapeutic interventions to prevent the inflammation associated with NAFLD. Because NAFLD is associated with metabolic disease and is often considered the hepatic manifestation of metabolic syndrome, pharmacological agents that target the lipid accumulation or insulin resistance component of NAFLD are used as front-line therapies. Certain anti-diabetic therapies including
| 897 |
4954767
| 0 | 16 |
pioglitazone (Ratziu et al. 2008), acarbose (Chiasson et al. 2002), metformin (Haukeland et al. 2009) and possibly statins (Eslami et al. 2013) exhibit anti-inflammatory properties and are effective at treating NAFLD. However, there is a real need for novel, new classes of anti-inflammatory drugs for the prevention and treatment of the localized inflammation associated with NAFLD as the inflammation in the presence of steatosis is what leads to NASH and the more severe stages of the disease that result in death. The pathogenesis of NAFLD is now considered to be 'multiple-hit' due to hepatic insults that occur in parallel, which ultimately leads to increased lipid accumulation and immune infiltration (Day & James 1998). The early stage of C10 treatment reverses HF diet-induced hepatic steatosis and hepatic and adipose inflammation and reduces serum cholesterol. (A) Treatment with C10 decreased hepatic triglyceride content when compared to HF-fed sham and DMSO groups. (B) Total serum cholesterol levels were reduced in the HF-fed C10-treated mice when compared to HF-fed control mice, however, serum triglyceride levels were unchanged. Dotted lines represent the mean and error bars indicate + s.e.m. (C) Hepatic Ifnb1, Tnfa and Il6 expression was reduced in C10-treated mice when compared to HF sham and DMSO groups. Bars indicate mean + s.e.m. Significance was determined using ANOVA followed by Tukey's post hoc analysis for multiple comparison; # Different from HF sham and DMSO groups; P < 0.05. *Different from HF-fed groups; P < 0.05, n = 4-6. C10 halts HFD-induced inflammation and NAFLD A Patton et al. 237:3 Journal
| 898 |
4954767
| 0 | 16 |
of Endocrinology NAFLD (steatosis) is considered benign; however, it is now believed to be an active state of inflammation and metabolic dysfunction (Tilg & Moschen 2010, Buzzetti et al. 2016. Moreover, we now know that inflammation occurs in hepatic steatosis and thus it can be targeted by pharmacological agents before the onset of NASH. Activation of TLR4 signaling is a key mediator of HF diet-induced hepatic inflammation. In addition to being a critical mediator of TLR4 signaling, MyD88 has recently been shown to be critical for maintaining mammalian target of rapamycin (mTOR) activation (Chang et al. 2013). Given the metabolic consequences of NAFLD (i.e. obesity, insulin resistance and T2DM), mTOR involvement is essential in light of its role in cardiovascular diseases such as atherosclerosis, coronary heart disease and stroke (Tarantino & Capone 2013, Patil & Sood 2017. Moreover, a key mechanism linking inflammation to altered glucose and lipid metabolism is that visceral adipocytes and associated macrophages produce and release copious amounts of inflammatory cytokines into both the portal and systemic vasculature, which cause insulin resistance in insulin target tissues (i.e. liver, muscle and fat). Thus, the novel findings presented herein that C10 can reverse HF dietinduced hepatic steatosis, glucose intolerance, as well as hepatic and visceral adipose inflammation, coupled with the finding that C10 inhibits Tnfa-induced Vcam1 expression and reduces monocytic cell adhesion to endothelial cells (Dagia et al. 2004), an important process in the pathogenesis of atherosclerosis and other chronic inflammatory diseases, suggests that C10 may have a more profound clinical impact than the treatment
| 899 |
4954767
| 0 | 16 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.