Deep-Manager: a versatile tool for optimal feature selection in live-cell imaging analysis.

One of the major problems in bioimaging, often highly underestimated, is whether features extracted for a discrimination or regression task will remain valid for a broader set of similar experiments or in the presence of unpredictable perturbations during the image acquisition process. Such an issue is even more important when it is addressed in the context of deep learning features due to the lack of a priori known relationship between the black-box descriptors (deep features) and the phenotypic properties of the biological entities under study. In this regard, the widespread use of descriptors, such as those coming from pre-trained Convolutional Neural Networks (CNNs), is hindered by the fact that they are devoid of apparent physical meaning and strongly subjected to unspecific biases, i.e., features that do not depend on the cell phenotypes, but rather on acquisition artifacts, such as brightness or texture changes, focus shifts, autofluorescence or photobleaching. The proposed Deep-Manager software platform offers the possibility to efficiently select those features having lower sensitivity to unspecific disturbances and, at the same time, a high discriminating power. Deep-Manager can be used in the context of both handcrafted and deep features. The unprecedented performances of the method are proven using five different case studies, ranging from selecting handcrafted green fluorescence protein intensity features in chemotherapy-related breast cancer cell death investigation to addressing problems related to the context of Deep Transfer Learning. Deep-Manager, freely available at https://github.com/BEEuniroma2/Deep-Manager , is suitable for use in many fields of bioimaging and is conceived to be constantly upgraded with novel image acquisition perturbations and modalities.

Accelerating the experimental responses on cell behaviors: a long-term prediction of cell trajectories using Social Generative Adversarial Network.

The incremented uptake provided by time-lapse microscopy in Organ-on-a-Chip (OoC) devices allowed increased attention to the dynamics of the co-cultured systems. However, the amount of information stored in long-time experiments may constitute a serious bottleneck of the experimental pipeline. Forward long-term prediction of cell trajectories may reduce the spatial-temporal burden of video sequences storage. Cell trajectory prediction becomes crucial especially to increase the trustworthiness in software tools designed to conduct a massive analysis of cell behavior under chemical stimuli. To address this task, we transpose here the exploitation of the presence of "social forces" from the human to the cellular level for motion prediction at microscale by adapting the potential of Social Generative Adversarial Network predictors to cell motility. To demonstrate the effectiveness of the approach, we consider here two case studies: one related to PC-3 prostate cancer cells cultured in 2D Petri dishes under control and treated conditions and one related to an OoC experiment of tumor-immune interaction in fibrosarcoma cells. The goodness of the proposed strategy has been verified by successfully comparing the distributions of common descriptors (kinematic descriptors and mean interaction time for the two scenarios respectively) from the trajectories obtained by video analysis and the predicted counterparts.

Discovering the hidden messages within cell trajectories using a deep learning approach for in vitro evaluation of cancer drug treatments.

We describe a novel method to achieve a universal, massive, and fully automated analysis of cell motility behaviours, starting from time-lapse microscopy images. The approach was inspired by the recent successes in application of machine learning for style recognition in paintings and artistic style transfer. The originality of the method relies i) on the generation of atlas from the collection of single-cell trajectories in order to visually encode the multiple descriptors of cell motility, and ii) on the application of pre-trained Deep Learning Convolutional Neural Network architecture in order to extract relevant features to be used for classification tasks from this visual atlas. Validation tests were conducted on two different cell motility scenarios: 1) a 3D biomimetic gels of immune cells, co-cultured with breast cancer cells in organ-on-chip devices, upon treatment with an immunotherapy drug; 2) Petri dishes of clustered prostate cancer cells, upon treatment with a chemotherapy drug. For each scenario, single-cell trajectories are very accurately classified according to the presence or not of the drugs. This original approach demonstrates the existence of universal features in cell motility (a so called "motility style") which are identified by the DL approach in the rationale of discovering the unknown message in cell trajectories.

Communicative reprogramming non-curative hepatocellular carcinoma with low-dose metronomic chemotherapy, COX-2 inhibitor and PPAR-gamma agonist: a phase II trial.

Systemic therapy for advanced hepatocellular carcinoma (HCC) is still challenging. A biomodulatory therapy approach targeting the communicative infrastructure of HCC, including metronomic low-dose chemotherapy with capecitabine, pioglitazone and rofecoxib, has been evaluated in patients with non-curative HCC. Altogether 38 patients were evaluable in this one-arm, multicenter phase II trial. The primary endpoint, median progression-free survival was 2.7 months (95% CI: 1.6-3.79) for all evaluable patients and 8.4 months (95% CI: 0-18.13) for patients ≥ 6 weeks on protocol. Median overall survival (OS) was 6.7 months (95% CI: 4.08-9.31) and 9.4 months (95% CI: 4.82-13.97), respectively. Most common adverse events were edemas grade 3, which were commonly related to the advanced stage, with 66% of the patients suffering from liver cirrhosis. Exploratory data analyses showed significant impact of ECOG performance status grade 0 versus 1 and CLIP score 0/1 versus > 1 on OS, 9.8 months (95% CI: 4.24-15.35) versus 2.7 months (95% CI: 1.03-4.36; P = 0.002), and 9.8 months (95% CI: 3.23-16.37) versus 4.4 months (95% CI: 3.14-5.66; P = 0.009), respectively. Preceding tumor surgery had significant beneficial impact on survival, as well as maximal tumor diameter of < 5 cm. The correlation of C-reactive protein decrease with significantly improved OS underlines the close link between inflammation and tumor control. Biomodulatory therapy in advanced HCC may be a low toxic, efficacious treatment and principally demonstrates that such approaches should be followed further for treatment of advanced HCC.

Melatonin promotes Bax sequestration to mitochondria reducing cell susceptibility to apoptosis via the lipoxygenase metabolite 5-hydroxyeicosatetraenoic acid.

Extra-neurological functions of melatonin include control of the immune system and modulation of apoptosis. We previously showed that melatonin inhibits the intrinsic apoptotic pathway in leukocytes via stimulation of high affinity MT1/MT2 receptors, thereby promoting re-localization of the anti-apoptotic Bcl-2 protein to mitochondria. Here we show that Bcl-2 sequesters pro-apoptotic Bax into mitochondria in an inactive form after melatonin treatment, thus reducing cell propensity to apoptosis. Bax translocation and the anti-apoptotic effect of melatonin are strictly dependent on the presence of Bcl-2, and on the 5-lipoxygenase (5-LOX) metabolite 5-hydroxyeicosatetraenoic acid (5-HETE), which we have previously shown to be produced as a consequence of melatonin binding to its low affinity target calmodulin. Therefore, the anti-apoptotic effect of melatonin requires the simultaneous, independent interaction with high (MT1/MT2) and low (calmodulin) affinity targets, eliciting two independent signal transduction pathways converging into Bax sequestration and inactivation. MT1/MT2 vs. lipoxygenase pathways are activated by 10(-9) vs. 10(-5)M melatonin, respectively; the anti-apoptotic effect of melatonin is achieved at 10(-5)M, but drops to 10(-9)M upon addition of exogenous 5-HETE, revealing that lipoxygenase activation is the rate-limiting pathway. Therefore, in areas of inflammation with increased 5-HETE levels, physiological nanomolar concentrations of melatonin may suffice to maintain leukocyte viability.

Lipoxygenase-mediated pro-radical effect of melatonin via stimulation of arachidonic acid metabolism.

We have shown that melatonin immediately and transiently stimulates intracellular free radical production on a set of leukocytes, possibly as a consequence of calmodulin binding. We show here that melatonin-induced ROS are produced by lipoxygenase (LOX), since they are prevented by a set of LOX inhibitors, and are accompanied by increase of the 5-LOX product 5-HETE. LOX activation is accompanied by strong liberation of AA; inhibition of Ca(2+)-independent, but not Ca(2+)-dependent, phospholipase A2 (PLA2), prevents both melatonin-induced arachidonic acid and ROS production, whereas LOX inhibition only prevents ROS, indicating that PLA2 is upstream with respect to LOX, as occurs in many signaling pathways. Chlorpromazine, an inhibitor of melatonin-calmodulin interaction, inhibits both ROS and arachidonic acid production, thus possibly placing calmodulin at the origin of a melatonin-induced pro-radical pathway. Interestingly, it is known that Ca(2+)-independent PLA2 binds to calmodulin: our results are compatible with PLA2 being liberated by melatonin from a steady-state calmodulin sequestration, thus initiating an arachidonate signal transduction. These results delineate a novel molecular pathway through which melatonin may participate to the inflammatory response.

Oxidation-dependent maturation and survival of explanted blood monocytes via Bcl-2 up-regulation.

Monocytes isolated and cultured according to standard procedures from the blood of 22 healthy donors display an activation process, monitored as adhesion and increased exposure of CD11. Starting from very early time points, monocytes undergo a deep redox modulation, i.e., they increase reactive oxygen species (ROS) formation and decrease glutathione content; at the same time, the anti-apoptotic protein Bcl-2 is substantially up-regulated. The cause-effect relationship between these parameters was investigated. On the one side, pharmacological glutathione depletion with BSO further increases ROS formation and Bcl-2 levels. On the other side, scavenging of ROS by Trolox prevents Bcl-2 up-regulation. Two lipoxygenase (LOX) inhibitors (CAPE or AA861) prevent ROS increase and, accordingly, also prevent Bcl-2 up-regulation. All this evidence supports the redox-sensitivity of Bcl-2 regulation. Trolox, CAPE and AA861, i.e., all treatments that abolish ROS increase and prevent Bcl-2 up-regulation, increase the rate of cell loss, whereas BSO, increasing Bcl-2, reduces cell loss and induces chemo-resistance. Thus, explanted healthy monocytes seem to undergo an oxidation-dependent maturation implying increased survival via Bcl-2 up-regulation, perhaps mimicking physiological activation.

HIV protease inhibitor therapy reverses neutrophil apoptosis in AIDS patients by direct calpain inhibition.

The reduction of neutrophils apoptosis is one of the main non-virological effects of protease inhibitor (PI) therapy. We explore here whether this may be due to the cross-inhibition of calpain, an important non-virological target of PI in vitro. We found that the high basal level of neutrophils apoptosis in AIDS patients is strictly related to an increased intracellular calpain activity. Both alterations disappear after PI treatment, with apoptosis and calpain going back to normal levels after 3 months of PI therapy, independently of a proficient antiviral effect. PI drugs exerted a similar antiapoptotic and anticalpain effects on neutrophils in ex vivo experiments: strikingly, the effects were mimicked by commercially available calpain inhibitors. This study shows, for the first time, that apoptosis of neutrophils in AIDS patients is mediated by calpain, and that neutrophil survival in PI treated AIDS patients is a non virological effect due to calpain inhibition.

NMR exposure sensitizes tumor cells to apoptosis.

NMR technology has dramatically contributed to the revolution of image diagnostic. NMR apparatuses use combinations of microwaves over a homogeneous strong (1 Tesla) static magnetic field. We had previously shown that low intensity (0.3-66 mT) static magnetic fields deeply affect apoptosis in a Ca2+ dependent fashion (Fanelli et al., 1999 FASEBJ., 13;95-102). The rationale of the present study is to examine whether exposure to the static magnetic fields of NMR can affect apoptosis induced on reporter tumor cells of haematopoietic origin. The impressive result was the strong increase (1.8-2.5 fold) of damage-induced apoptosis by NMR. This potentiation is due to cytosolic Ca2+ overload consequent to NMR-promoted Ca2+ influx, since it is prevented by intracellular (BAPTA-AM) and extracellular (EGTA) Ca2+ chelation or by inhibition of plasma membrane L-type Ca2+ channels. Three-days follow up of treated cultures shows that NMR decrease long term cell survival, thus increasing the efficiency of cytocidal treatments. Importantly, mononuclear white blood cells are not sensitised to apoptosis by NMR, showing that NMR may increase the differential cytotoxicity of antitumor drugs on tumor vs normal cells. This strong, differential potentiating effect of NMR on tumor cell apoptosis may have important implications, being in fact a possible adjuvant for antitumor therapies.

Magnetic fields protect from apoptosis via redox alteration.

Magnetic fields (MFs) are receiving much attention in basic research due to their emerging ability to alter intracellular signaling. We show here that static MFs with intensity of 6 mT significantly alter the intracellular redox balance of U937 cells. A strong increase of reactive oxygen species (ROS) and a decrease of glutathione (GSH) intracellular levels were found after 2 h of MF exposure and maintained thereafter. We found that also other types of MFs, such as extremely-low-frequency (ELF) MFs affect intracellular GSH starting from a threshold at 0.09 mT. We previously reported that static MFs in the intensity range of 0.3-60 mT reduce apoptosis induced by damaging agents (Fanelli et al., 1998). Here, we show that ELF-MFs are also able to protect U937 from apoptosis. Interestingly, this ability is limited to the ELF intensities able to alter redox equilibrium, indicating a link between MF's antiapoptotic effect and the MF alteration of intracellular redox balance. This suggests that MF-produced redox alterations may be part of the signaling pathway leading to apoptosis antagonism. Thus, we tested whether MFs may still exert an antiapoptotic action in cells where the redox state was artificially altered in both directions, that is, by creating an oxidative (via GSH depletion with BSO) or a reducing (with DTT) cellular environment. In both instances, MFs fail to affect apoptosis. Thus, a correct intracellular redox state is required in order for MFs to exert their antiapoptotic effect.

The cleavage mode of apoptotic nuclear vesiculation is related to plasma membrane blebbing and depends on actin reorganization.

In U937 monocytic cells induced to apoptosis, plasma membrane blebbing of different intensities appears, before the development of nuclear alterations; this latter phenomenon can occur through two major pathways, namely the cleavage and the budding mode (Dini et al., 1996). Strongly blebbing cells develop deep nuclear constrictions leading to nuclear fragmentation according to the cleavage mode, while cells with milder forms of blebbing, or no blebbing at all, undergo nuclear fragmentation along the budding mode. Compounds interfering with different cytoskeletal components affect blebbing, which is completely inhibited by the actin polymerization inhibitors, cytochalasins, while disturbance of tubulin network with taxol limits blebbing to milder forms. At the same time, the cytoskeletal poisons affect the type of nuclear fragmentation, abolishing the cleavage mode, shifting all events into the budding pathway. Adherent cells, which possess a more structured cytoskeleton, do not develop strong blebs and undergo nuclear fragmentation via budding. These observations suggest that the deep cytoskeletal movements that cause the strongest forms of plasma membrane blebbing strangle the nucleus, leading to the constrictions that later evolve into nuclear fragmentation by cleavage. The trigger for the cytoskeletal movements, known to be redox-sensitive, is probably the apoptotic GSH extrusion.

Hyperpolarization of plasma membrane of tumor cells sensitive to antiapoptotic effects of magnetic fields.

Chemical/physical agents able to prevent apoptosis are receiving much attention for their potential health hazard as tumor promoters. Magnetic fields (MFs), which have been shown to increase the occurrence of some tumors, reduce damage-induced apoptosis by a mechanism involving Ca2+ entry into cells. In order to discover the mechanism of such effect of MFs, we investigated the interference of MFs on cell metabolism and analyzed cell parameters that are involved in apoptotic signaling and regulation of Ca2+ fluxes. Here we show that different types (static and extremely low-frequency, ELF pulsating) of MFs of different intensities alter plasma membrane potential. Interestingly, MFs induce plasma membrane hyperpolarization in cells sensitive to the antiapoptotic effect of MFs, whereas cells that are insensitive showed a plasma membrane depolarization. These opposite effects suggest that protection against apoptosis and membrane potential modulation are correlated, plasma membrane hyperpolarization possibly being part of the signal transduction chain determining MFs' antiapoptotic effect.

[Lack of direct cytotoxic effect of intracellular nanotubes]. / Impatto occupazionale ed ambientale dei nanotubi di carbonio, analisi della citotossicità su linee cellulari.

Nanotubes have a great therapeutic potential due to their astounding physico-chemical features, the possibility to be funtionalised for ad hoc uses, and the specific interaction of nanotubes as such with life molecules (DNA and proteins). These features recommend a thorough toxicological study before widespread pharmaceutic use. We provide evidence that culture cells with phagocytic potential internalise multi wall nanotubes (10-50 nm average size). This is not accompanied by cytotoxicity in terms of induction of &apoptosis or necrosis at the doses used (up to 125 microg/mI).

Oxidative Bax dimerization promotes its translocation to mitochondria independently of apoptosis.

Bax is a cytosolic protein, which in response to stressing apoptotic stimuli, is activated and translocates to mitochondria, thus initiating the intrinsic apoptotic pathway. In spite of many studies and the importance of the issue, the molecular mechanisms that trigger Bax translocation are still obscure. We show by computer simulation that the two cysteine residues of Bax may form disulfide bridges, producing conformational changes that favor Bax translocation. Oxidative, nonapoptogenic treatments produce an up-shift of Bax migration compatible with homodimerization, which is reverted by reducing agents; this is accompanied by translocation to mitochondria. Dimers also appear in pure cytosolic fractions of cell lysates treated with H2O2, showing that Bax dimerization may take place in the cytosol. Bax dimer-enriched lysates support Bax translocation to isolated mitochondria much more efficiently than untreated lysates, indicating that dimerization may promote Bax translocation. The absence of apoptosis in our system allows the demonstration that Bax moves because of oxidations, even in the absence of apoptosis. This provides the first evidence that Bax dimerization and translocation respond to oxidative stimuli, suggesting a novel role for Bax as a sensor of redox imbalance.

Hypoxic stress stably alters apoptotic parameters on U937 cells.

Tumor promonocytic U937 cells cultured under a low O(2)/high CO(2) atmosphere display altered characteristics after restoration of normal atmosphere: increased resistance to apoptosis induced by different treatments; apoptotic morphology; lack of glutathione (GSH) extrusion in apoptosis; lack of protection by antioxidants; and lack of Ca(2+) mobilization with thapsigargin. These alterations were stably maintained for many months of culture in normal conditions, originating the stable U937-HX variant. Since the hypoxic treatment did not produce a great selective pressure, the alterations are conceivably the result of stable adaptative response.

Apoptotic GSH extrusion is associated with free radical generation.

Reactive oxygen species (ROS) are involved in many forms of apoptosis and mediate apoptosis in a number of cell types. In this paper, we use a variant of U937 monocytic cells (U937 HX) that show different biochemical features with respect to standard U937. Apoptotic standard U937 extrude reduced glutathione (GSH) and generate free radicals concomitantly with loss of mitochondria transmembrane potential (mt Deltapsi). These events are correlated with the extrusion of intracellular GSH. Conversely, apoptotic U937 HX cells retain GSH, and the loss of mt Deltapsi is not accompanied by generation of free radicals. The perfect inverse correlation between (a) ROS generation and (b) the presence of intracellular GSH during apoptosis suggests novel mechanisms to finely tune ROS generation in apoptosis.

H2O2-induced block of glycolysis as an active ADP-ribosylation reaction protecting cells from apoptosis.

H2O2 treatment on U937 cells leads to the block of glycolytic flux and the inactivation of glyceraldehyde-3-phosphate-dehydrogenase by a posttranslational modification (possibly ADP-ribosylation). Glycolysis spontaneously reactivates after 2 h of recovery from oxidative stress; thereafter cells begin to undergo apoptosis. The specific ADP-ribosylation inhibitor 3-aminobenzamide inhibits the stress-induced inactivation of glyceraldehyde-3-phosphate-dehydrogenase and the block of glycolysis; concomitantly, it anticipates and increases apoptosis. Exogenous block of glycolysis (i.e., by culture in glucose-free medium or with glucose analogs or after NAD depletion), turns the transient block into a stable one: this results in protection from apoptosis, even when downstream cell metabolism is kept active by the addition of pyruvate. All this evidence indicates that the stress-induced block of glycolysis is not the result of a passive oxidative damage, but rather an active cell reaction programmed via ADP-ribosylation for cell self-defense.

GSH extrusion and and the mitochondrial pathway of apoptotic signalling.

New evidence suggests that physiological and damaging agents activate two different pathways of apoptotic signalling, which are mediated by protein-protein interactions and mitochondrial alterations respectively. The two pathways converge at the activation of caspase 3, the key effector of the execution phase of apoptosis, thus giving similar final results. The knowledge that different biochemical routes exist allows us to re-evaluate previous apparently contradictory results concerning the events occurring during apoptosis, and their respective roles. In particular, this applies to the role of oxidative stress and redox imbalance in the signal transduction events of apoptosis. It now appears that oxidative alterations are absent, or at least unnecessary, for the development of the physiological pathway. Instead, clear indications are emerging showing that redox imbalance is required for the damage-induced mitochondrial pathway. This is suggested by the finding that the depletion of glutathione, a common event in damage-induced apoptosis, is necessary and sufficient to induce cytochrome c release, the key event of this pathway. A model is proposed with GSH efflux as the backbone of the damage-induced apoptotic pathway.

Glutathione depletion causes cytochrome c release even in the absence of cell commitment to apoptosis.

We demonstrate here that the release of mature cytochrome c from mitochondria is a cellular response to the depletion of glutathione, the main intracellular antioxidant, independently from the destiny of the cells, i.e., apoptosis or survival. On the one hand, cytosolic cytochrome c was detected in cells where the inhibition of glutathione synthesis led to glutathione depletion without impairing viability or in tight concomitance with glutathione depletion prior to puromycin-induced apoptosis. Removal of the apoptogenic agent prior to apoptosis, but after glutathione extrusion and cytochrome c release, led to recovery of preapoptotic cells, which resume healthy features, i.e., restoration of normal glutathione levels and disappearance of cytosolic cytochrome c. On the other hand, in an example of apoptosis occurring without glutathione depletion, no translocation of cytochrome c from mitochondria to cytosol was detected. Unlike the other instances of apoptosis, in this case caspase 3 was not activated, thus suggesting the following oxidant-related apoptotic pathway: glutathione depletion, cytochrome c release, and caspase 3 activation. These results show that cytochrome c release is not a terminal event leading cells to apoptosis, but rather is the consequence of a redox disequilibrium that, under some circumstances, may be associated with apoptosis.