Integrins protect sensory neurons in models of paclitaxel-induced peripheral sensory neuropathy

We first sought to confirm and extend prior results on the effect of paclitaxel on Drosophila sensory dendrites using high-resolution analysis of terminal morphology. We administered paclitaxel (10, 20, and 30 μM) in food beginning from 24 to 28 h after egg laying (AEL; early first instar). By this stage, dendritic arborization (da) sensory neurons have completed axon pathfinding and formed major peripheral dendrite branches (33). All of these paclitaxel concentrations were previously used as models for CIPN (6, 7). We dissected treated larvae at the late third-instar stage. Survival of larvae treated with 30 μM was rare (<5% survival rate by third instar stages), but animals tolerated 10 and 20 μM well and often survived past third instar stages. The da neuron branches still broadly covered their territories in larvae fed 10 and 20 μM (Fig. 1 A–C), but neurons showed clear morphological changes, such as predegenerative varicosities and fragmentation resembling phenotypes reported in vertebrate models (9, 24) (Fig. 2 and SI Appendix, Fig. S1). Nociceptive neurons in animals treated with 10 μM did not show an overt change in branching pattern. In contrast, 20-μM treatments induced marked changes in branching patterns (Fig. 1 D, E, and G). Peripheral arbors were sporadically clumped and patchy in proximal regions of the arbor and depleted in more distal regions (Figs. 1 A–C, 2 A, C, and C', and SI Appendix). We developed an analysis to compare branch density and how density is distributed across the dendritic field. Density analysis revealed that control nociceptive arbors have a clear maximal peak density, indicating a relatively even and uniform field coverage (Fig. 1D and SI Appendix, Fig. S2A).

In contrast, paclitaxel dose-dependently increased overall local density of nociceptive arbors and the density distribution lacked a clear peak (Fig. 1D and SI Appendix, Fig. S2 B and C). Thus, paclitaxel treatment leads to a more heterogeneous distribution of dendrites. By plotting density according to the distance from soma, we found that paclitaxel caused an increase in proximal branch density (Fig. 1E), which was also supported by Sholl analysis (SI Appendix, Fig. S2E). Neither concentration of paclitaxel caused a change in total dendrite length compared to control (Fig. 1F). In contrast, we found that 20 μM caused a robust increase in the number of total dendrite branch points (P < 0.001), compared to a modest increase observed with 10-μM treatment (P = 0.106) (Fig. 1G). Thus, to explore the events that may precede degeneration of nociceptive arbors, we used both 10 and 20 μM, and used 20 μM of paclitaxel to further study the effects on nociceptive arbor branching pattern and degenerative changes.

In addition to causing degeneration in nociceptive neurons, paclitaxel feeding resulted in extensive branch crossing of cIV dendrites, indicating a failure of self-avoidance (Fig. 2 A, C, C', E, and F). Two parallel mechanisms impact self-avoidance in Drosophila sensory dendrites. Dscam1-mediated recognition between sister dendrites results in repulsion and avoidance (34–36). In parallel, integrin receptors for the ECM maintain dendrites in a mostly two-dimensional (2D) arrangement on the epidermis by mediating substrate attachment (19, 20). This placement ensures that dendrites come into contact with one another as they grow, rather than grow over or under each other in 3D (19, 20). The observed defect in self-avoidance caused by paclitaxel raised the possibility that paclitaxel disrupts one of these mechanisms that prevent dendrite crossing. Consistent with this hypothesis, cooverexpression of αPS1 and βPS integrins in nociceptive neurons lessened paclitaxel-induced dendrite crossing (Fig. 2 B, D, D', and E). These data suggest that paclitaxel compromises the link between dendrite processes and the ECM, which can be compensated by increased levels of integrins.

Remarkably, integrin overexpression also largely rescued paclitaxel-induced degenerative phenotypes (Fig. 2 D, D', and F). We compared this result with another cell adhesion receptor, N-cadherin, which is known to promote cellular growth through cell–cell interactions (37, 38). We found that N-cadherin overexpression in nociceptive neurons partially prevented paclitaxel-induced phenotypes but not to the same degree as integrins (SI Appendix, Fig. S3). These results raise the possibility that integrin-mediated attachment to the ECM substrate and, to a lesser degree, cell–cell interactions through N-cadherin, can counterbalance the effects of paclitaxel treatment.

Drosophila larvae show a stereotyped nocifensive escape behavior in response to thermal or mechanical noxious stimuli. The escape sequence consists of C-shaped body bending, lateral rolling, and fast escape crawl (17, 21, 39, 40). To examine whether paclitaxel disrupts nocifensive responses, we administered paclitaxel at 1, 10, and 20 μM and induced escape behavior using global heat stimulation. We observed a significant decrease in nocifensive responses at 10 and 20 μM, but not at 1 μM (Fig. 3A). To determine if paclitaxel exerted an effect primarily on sensory neurons or downstream of primary nociceptors, we manipulated the downstream circuitry (SI Appendix, Fig. S4). We fed larvae paclitaxel and activated downstream circuitry by expressing TrpA in Down and Back (DnB) interneurons, one of the main downstream targets of larval nociceptive neurons (21). Thermogenetic activation of DnB neurons elicits bending and rolling behavior (21), so we predicted that direct activation of DnB neurons would still be able to induce rolling behavior if the nociceptive neurons were the major target of paclitaxel. Indeed, activation of DnB neurons effectively induced nociceptive behaviors with and without paclitaxel feeding. Our results are consistent with paclitaxel disrupting nociceptive behavior through action primarily on sensory neurons; however, we cannot eliminate the possibility that our thermogenetic activation approach may overcome disruptions to the interneurons.

We next examined whether integrin overexpression in nociceptive neurons can restore nocifensive behavior in paclitaxel-treated larvae. Without paclitaxel administration, we found that changes in the levels of integrins or overexpression of N-cadherin did not strongly alter nociceptive responses relative to genotype controls (SI Appendix, Fig. S5) (integrin down-regulation P = 0.099; integrin overexpression P = 0.676; Ncad overexpression P = 0.971). We found some evidence for a dose-dependent relationship between integrin levels and nocifensive responses, insofar as integrin knockdown vs. overexpression led to differences in the number of rolls initiated per nocifensive bout (P = 0.013). This effect on nociceptive responses could conceivably be due to functional effects arising from changes in the 3D positioning of arbors relative to the epidermis, as has been described in a prior study (41). At both 10 and 20 μM of paclitaxel, overexpression of integrins in nociceptors led to a stronger rolling response to noxious heat compared to genotype controls (Fig. 3B). These results together suggest that integrin overexpression in nociceptive neurons prevents functional disruption in detecting noxious heat stimuli.

We next investigated the cellular changes in neurons caused by paclitaxel treatment that could be offset by augmenting integrin levels. Conceivably, these changes could include reduction in de novo integrin synthesis, reduced integrin recycling, and increased integrin degradation. We therefore investigated potential changes caused by paclitaxel in endosomal and lysosomal pathways in nociceptive neurons in Drosophila. To assess overall changes in the endocytic pathway, we chose two markers that label distinct populations of endo-lysosomal vesicles, the small GTPase Rab4 and the transporter protein Spinster (Spin) (42, 43). Rab4 mediates the early endosome to recycling endosome transition and is found in both early and recycling endosomes (43, 44). Spin plays a critical role in the autophagy-lysosome transition and is found in late endosomes, lysosomes, and autophagosomes (42, 45).

We expressed Rab4-RFP and Spin-RFP in nociceptive neurons and monitored their dynamics in primary neurites in late third-instar larvae. High-resolution time-lapse imaging (SI Appendix, SI Materials and Methods) revealed bidirectional trafficking of Rab4⁺ and Spin⁺ endosomes of various sizes and speeds (Fig. 4 A and B'). We found that effects on Rab4 and Spin trafficking were evident with both 10 and 20 μM chronic paclitaxel treatment (Fig. 4 C–I), indicating changes in both recycling and lysosomal pathways. Chronic paclitaxel treatment caused a higher portion of Rab4⁺ vesicles to become stationary at the expense of both anterograde and retrograde motility (Fig. 4 C, C', and E), with the motile pool of Rab4⁺ vesicles showing a reduced velocity and a higher frequency of direction switching (Fig. 4 F and G). Motile vesicles showed a significant reduction in density, whereas paclitaxel treatment modestly reduced the density of stationary vesicles (Fig. 4H) (VehCont vs. PTX10: P = 0.291 and VehCont vs. PTX20: P = 0.200). In addition, both 10 and 20 μM paclitaxel caused a marked enlargement of some nonmotile Spin⁺ vesicles (Fig. 4 D, D', and I). Together, our results suggest disruptions in both recycling and lysosomal pathways by paclitaxel.

We next tested whether changes to the endosomal-lysosomal pathway could affect trafficking of integrins following chronic paclitaxel treatment. To examine whether the intracellular localization of integrins is changed by paclitaxel administration, we expressed YFP-Rab4, and coexpressed the α- and β-integrin subunits, αPS1 and βPS, in nociceptive neurons (Fig. 4J). We labeled YFP-Rab4 and βPS integrins, and used superresolution microscopy (46) to quantify their colocalization in soma, dendrites, and proximal axon. Soma volume was comparable between control and paclitaxel-treated cells (Fig. 4K). Rab4 density showed a modest decrease (P = 0.062), whereas the number of Rab4 vesicles that colocalized with integrins showed a stronger decrease upon paclitaxel treatment, particularly in dendrites (Fig. 4 L–P). These results suggest that paclitaxel disrupts trafficking of Rab4-mediated recycling endosomes carrying integrins in nociceptive arbors. We cannot exclude the possibility that changes in levels of protein synthesis or degradation contribute to these observations. However, our results suggest that these intracellular changes could affect the amount of integrins available on the surface to support nociceptive arbor maintenance.

Next, we examined whether changes in intracellular trafficking precede or follow morphological degeneration of nociceptive terminals. Instead of chronically treating larvae with paclitaxel (Fig. 4), we administered paclitaxel at 20 μM for either 48 h or 72 h and quantified changes in morphology (Fig. 5 A–D and SI Appendix, Figs. S6 and S7). For Rab4, 72-h, but not 48-h, administration of paclitaxel frequently induced varicosities in higher-order branches, but only rarely induced severe degeneration in either genetic control or Rab4-expressing nociceptive neurons (Fig. 5 B–D and SI Appendix, Figs. S6 and S7). Thus, we consider 48 h as a stage prior to paclitaxel-induced emergence of morphological degeneration and studied paclitaxel effects on Rab4 trafficking at this time point. In contrast, Spin-expressing neurons showed degeneration phenotypes in vehicle control (Fig. 5 B and D compared to SI Appendix, Fig. S8 A and B) and showed a modest increase in both degeneration and branch crossing phenotypes (SI Appendix, Fig. S8) at 48 h. Due to disruptions caused by Spin alone, we were not able to dissociate degeneration and trafficking defects and compared effects of paclitaxel to baseline degeneration phenotypes at 48 h.

To investigate potential changes preceding morphological degeneration caused by paclitaxel, we monitored the trafficking of Rab4 and Spin in primary neurites treated with paclitaxel for 48 h. Acute changes in Rab4 and Spin were similar to the phenotypes observed in animals chronically treated with paclitaxel (Figs. 4 and 5 and SI Appendix, Fig. S9). We found that the total number of Rab4 puncta was reduced upon paclitaxel administration and a higher portion of Rab4⁺ vesicles became stationary (Fig. 5 E–G). For Spin, paclitaxel caused a significant enlargement of Spin⁺ vesicles by 48 h (SI Appendix, Fig. S9 D and F).

Overexpression of integrins affected these vesicles differently following 48-h paclitaxel administration. Cooverexpression of integrins with Rab4 prevented paclitaxel-induced changes in Rab4 (SI Appendix, Fig. S9 A–C), whereas coexpression of integrins and Spin did not rescue the enlargement phenotype (SI Appendix, Fig. S9 E and F). Together, our results suggest that paclitaxel-induced intracellular trafficking changes precede morphological degeneration rather than emerge as by-products of the morphological changes caused by chronic paclitaxel administration. Furthermore, integrins appear to impact trafficking of Rab4⁺ vesicles, providing a possible mechanism of integrin-mediated protection in a CIPN model.

We next asked whether effects on endosomal-lysosomal pathways and integrin-mediated protection from paclitaxel are also observed in vertebrate neurons. First, we examined the effect of paclitaxel treatment on Rab4 endosomes and Lamp1 lysosomes. Consistent with our findings in Drosophila, we found that the motility of Rab4⁺ and Lamp1⁺ vesicles is altered prior to morphological degeneration of mouse DRG neurons (SI Appendix, Figs. S10 and S11). We next asked whether integrin-mediated protection from paclitaxel is also observed in mouse DRG neurons. Since augmentation of a single integrin subunit can recruit endogenous subunit partners (47, 48), we chose to transduce the human integrin β1 subunit 1 (ITGB1), a major β-subunit in integrin heteromeric dimers in adult rodent DRG neurons (49–51). Lentivirus-mediated delivery of ITGB1 was carried out in adult DRG neurons at 5 d in vitro (DIV) prior to treatment with either vehicle or 50 nM paclitaxel for 72 h starting at 12 DIV, an experimental paradigm that had been previously shown to induce early signs of axon degeneration and reduce axon growth (8). To better capture the changes in the cells in culture and to avoid selection bias, we sampled large areas (2 × 2 mm²) for quantification. We found that treating DRG neurons with paclitaxel resulted in axon loss and axon fragmentation (Fig. 6 A and B'), as shown by a significant decrease in total axon area and an increased degeneration index (the ratio between fragmented axon and total axon areas) compared to control DRG neurons (Fig. 6 E and F and SI Appendix, Fig. S12). Consistent with the protective capacity observed in Drosophila nociceptive neurons, DRG neurons transduced with ITGB1 prior to paclitaxel treatment showed no change in total axon area and a reduced degeneration index compared to paclitaxel-treated control neurons (Fig. 6 E and F).

To examine whether integrin membrane trafficking is affected by paclitaxel treatment in DRG neurons, we next performed a surface biotinylation assay to measure total and surface levels of ITGB1. Cell lysates were collected after 24, 48, and 72 h following either vehicle or 50 nM paclitaxel treatment starting at 12 DIV. We found that treating DRG neurons with paclitaxel did not change total levels of ITGB1 arguing against an effect of paclitaxel on integrin de novo synthesis or turnover, but caused a significant reduction of ITGB1 localization at the cell surface at all observed time points in vitro (Fig. 6 G–I). These results suggest a conserved protective role for integrins in Drosophila and vertebrate CIPN models.

Because of their small caliber and location adjacent to a highly dynamic epidermis, IENFs are highly sensitive peripheral sensory structures (9). Our data show that integrin trafficking is impacted by paclitaxel and that augmenting integrin levels may help to restore neuronal interactions with the extracellular environment. A valuable next step would be assessing the effect of integrin augmentation in preclinical models in vivo. Given the accessibility of the peripheral environment, identification of the ligands that are involved in this protection is an important goal for future studies. Alternative signaling pathways could also contribute to integrin-mediated protection. For example, during the development of larval nociceptive neurons integrins colocalize with the conserved receptor tyrosine kinase Ret that is required for regular patterning, dynamic growth, and adhesion of dendritic branches (52, 53). Since adhesive roles for Ret and integrins are conserved in vertebrates (54), it is possible that similar receptors and ligands may interact with integrins to mediate interactions between IENFs and their extracellular environment.

There are important distinctions between Drosophila sensory dendrites and DRG axons, and indeed, we report distinct and shared phenotypes for these processes upon treatment with paclitaxel. We propose that the ability of integrins to counter degeneration in both systems is consistent with roles in both axons and dendrites, and our findings argue for some generality in these two models. Similarities could conceivably arise because peripheral sensory terminals share anatomical features such as naked endings and a close relationship with a substrate comprised of ECM. These features could render thin terminal sensory processes more susceptible to toxins that disrupt the connection to the ECM. The peculiarities of each system, as well as the in vivo vs. in vitro execution of the experiments, likely underlie the distinct paclitaxel-induced phenotypes we observed, such as branching and self-avoidance in vivo in sensory dendrites. As with their role in da neuron dendrites, integrins may play a critical role in maintaining the peripheral processes of mature sensory neurons. Supporting this hypothesis, the localization of tagged virally expressed or endogenous integrins is limited to the somatodendritic compartment in the mature CNS neurons, whereas integrins also traffic to axons of adult DRG sensory neurons and retinal ganglion cells (55–59).

Paclitaxel is a microtubule binding agent and interferes with dynamic instability of microtubules required for mitosis in cell division, leading to apoptosis of cancer cells (60). However, we currently have a limited understanding of how paclitaxel-induced changes disrupt intracellular transport in neurons and whether such disruption contributes to peripheral neuropathy (8, 61–65). Different studies consistently report that paclitaxel treatment results in changes to intracellular transport—including mitochondria, microtubule motor proteins, and lysosomes—but there is no strong consensus on the extent to which such changes contribute to peripheral sensory neuropathy. Prior studies also indicate that paclitaxel disrupts trafficking of specific intracellular cargos from the nucleus, including RNA granules required for maintenance of sensory terminals and nuclear encoded mRNA required for mitochondrial function (26, 66).

In our present study, we provide evidence that paclitaxel disrupts trafficking of recycling endosomes in Drosophila peripheral nociceptive arbors and adult mouse neurons in vitro. Specifically, we found a decrease in small GTPase Rab4 motility in both Drosophila and mouse, which precedes morphological degeneration in both systems. We also report a decrease in integrin colocalization with Rab4 vesicles in Drosophila sensory arbors in vivo after chronic paclitaxel treatment. Rab4 acts at the interface between early/sorting endosomes and recycling endosomes, and is a key mediator of integrin recycling (30, 67, 68). Perturbation of Rab4 motility and availability would likely lead to reduced integrin recycling, as further corroborated by our results indicating a reduction of ITGB1 surface expression levels in DRG neurons in vitro. Reciprocally, we found that integrin augmentation prevents paclitaxel-induced Rab4 changes in Drosophila following acute treatment, suggesting an intracellular mechanism of integrin-mediated protection that may ameliorate pathological progress by influencing vesicular trafficking. Given the disruption of integrin delivery to the cell surface in DRG neurons, augmentation of integrins may increase the probability that integrins reside in the membrane, facilitating the maintenance of the neuron–ECM link. Additionally, our data suggest that increased integrins in endosomes may affect their trafficking or downstream signaling events, ultimately changing integrin delivery and function (69). Using two complementary models, our study supports a conserved mechanism of pathology and protection by integrins. Strikingly, in both Drosophila larval sensory neurons and adult DRG neurons, paclitaxel treatment leads to disruption of integrin membrane trafficking, and augmentation of integrins was sufficient to protect sensory neurons from paclitaxel. In addition to other studies showing selective intracellular trafficking changes (26, 66), our results support a contribution of distinct endocytic changes to paclitaxel toxicity.

We also found enlargement of Spin⁺ vesicles in larval nociceptive neurons upon paclitaxel treatment. The significance of lysosome enlargement to neuropathic changes is not currently known, and our results indicate that they are unlikely to be directly linked with integrin pathways. Prior studies have linked perturbation of lysosomal degradation, lysosome enlargement, and progressive neurodegeneration (45, 70–74). Several mechanisms could explain the lysosome enlargement phenotype, including defects in lysosomal degradation machinery (72), failure in autophagic lysosome reformation (45, 70), and defective endosomal-to-lysosomal transport (71, 73, 74). Furthermore, while a prior study showed that paclitaxel-induced neurotoxicity is independent of lysosome trafficking in a mouse CIPN model in vitro (8), we found a decreased lysosome motility in DRG neurons starting from 24-h paclitaxel treatment, a time point prior to morphological degeneration of DRG axons. This difference may reflect the timeline of emergence of toxicity, different scoring methods, or dose-dependent toxicity of paclitaxel: 25 nM (8) vs. 50 nM (present study). Our current study does not prove causality between lysosomal changes and degeneration; however, it suggests that lysosomal changes might be a hallmark of CIPN pathology. Thus, examination of whether paclitaxel impacts integrin recycling by affecting lysosomal activity in chronic neuropathic conditions will be an important goal for future studies.

A number of survival and maintenance factors are modulated upon paclitaxel treatment. Paclitaxel reduces axonal localization of the prosurvival factor Bclw, but not Bcl2 or Bclx, in embryonic sensory neuron culture (26). MMP13 was selectively activated in epidermal cells but not in neurons upon paclitaxel treatment (25). Similarly, integrins may be one of several cargos that are reduced in the plasma membrane as a consequence of paclitaxel-induced disruption in endo-lysosomal compartments. Integrin-mediated pathways may be highly susceptible to these changes because integrin surface abundance relies heavily on recycling and integrins may be continuously required in dynamic cellular compartments, such as nociceptive terminals or the leading edge of cancer cells (19, 20, 67, 68, 75). Surface proteomic analysis may address the broader impact of paclitaxel on the levels of membrane proteins required for maintenance of sensory neurons. Our study provides motivation for further studies of membrane proteins in CIPN.

IENF density correlates with the severity of sensory peripheral neuropathy in patients, and the change in morphology serves as a strong predictor of symptoms and prognosis in clinics (2, 11, 12, 76–80). In patients with chronic peripheral neuropathy, axon terminals are present in the subepidermal layer but absent in the epidermis (77). It is therefore critical to understand the factors that help maintain terminal nerve fibers and that could prevent their degeneration in CIPN. Our results show that integrin supplementation promotes the maintenance of sensory terminals upon paclitaxel treatment in Drosophila and in DRG neurons cultured in vitro, suggesting that integrins could be a key factor in somatosensory terminal maintenance to counteract CIPN. It will be essential to know whether these effects translate to vertebrate axons in vivo targeting the epidermis. Notably, other studies have shown that integrin levels can increase after neuron injury, and levels correlate with regenerative ability (81). Differences in integrin levels between peripheral DRG neurons and neurons in the CNS may correlate with different abilities of these neurons to regenerate (55). Together, the present study and previous results suggest that integrins could be major determinants of neuronal maintenance in different types of injuries in the nervous system.

Currently, there is no effective method of preventing or treating CIPN other than stopping chemotherapeutic treatment or changing the chemotherapy regimen. Up to 80% of CIPN patients treated with paclitaxel still report symptoms in the long term after the cessation of chemotherapeutic treatment (22, 23). Limited symptomatic relief is provided by opioid analgesics, antidepressants, or anticonvulsants (23, 82). The in vivo Drosophila model to study chemotherapeutic-induced changes in neurons has led to the identification of integrins as a protective pathway effective also in vertebrate sensory neurons in vitro. Accumulating evidence underscores the importance of intrinsic mechanisms in preventing neural degeneration, yet the efficacy of these approaches in relevant in vivo contexts is poorly understood. The present study provides evidence that changes in the ability of neurons to link to, and interact with, the extracellular environment results in neuropathic changes in multiple models. Further studies of substrate interactions in these models might provide important insights into the mechanisms, etiology, and treatment of CIPN.

Paclitaxel (Tocris Bioscience) was diluted to 5 mM in ethanol, and fresh aliquots of paclitaxel were diluted to a final concentration of 1, 10, 20, or 30 μM in 1× PBS. A matching concentration of ethanol was added as a vehicle control. This solution was used to make food using instant Drosophila medium (Formula 4-24, Carolina Biological Supply Company) immediately before starting the larval assay. Embryos were collected on grape plates with yeast paste made with 0.5% propionic acid for chronic treatment and second instar or early third-instar larvae were collected manually (Fig. 5A) directly from standard molasses Drosophila media for acute treatment.

Global activation heat nocifensive assays were performed as previously described (21). The number of 360° rolls was scored by using trachea as a reference for the dorsal region of the larva.

All protocols and procedures used in this study to prepare primary culture of DRG neurons were approved by the Institutional Animal Care and Use Committee at Columbia University and according to the Guide for the Care and Use of Laboratory Animals distributed by the National Institutes of Health (83). DRG from 8- to 12-wk-old C57BL/6J mice of both sexes were dissected, dissociated, and plated in a 12-well plate. Paclitaxel was treated starting on 12 DIV at a final concentration of 50 nM; 1 μL of DMSO was added as a vehicle control.