Psilocybin prevents chemotherapy-induced peripheral neuropathy through mitochondrial trafficking preservation

INTRODUCTION

Chemotherapy-induced peripheral neuropathy (CIPN) is a common, disabling, and often irreversible complication of cancer treatment affecting many patients receiving platinum- or taxane-based regimens. Patients experience pain, cold sensitivity, numbness, and impaired touch sensation that can persist for years, frequently forcing dose reductions or discontinuation of otherwise life-saving therapy. The condition arises from injury to sensory nerves, mitochondrial dysfunction, loss of nerve endings in the skin, and maladaptive changes in central pain-processing circuits. Despite its enormous clinical burden, no proven preventive strategy exists.

RATIONALE

Psilocybin is a naturally occurring compound found in certain mushroom species that activates serotonin 2A receptors. These receptors have been linked to neuronal plasticity and mitochondrial regulation, which indicates that psilocybin could protect the sensory nervous system from chemotherapy-induced injury. Because nerve damage, once established, is often irreversible, prevention is more tractable than treatment. We therefore asked whether psilocybin administered before chemotherapy could prevent neuropathy from ever developing and used a combination of mouse behavioral models, electrophysiology, live-cell imaging, molecular profiling, and human tissue analysis to define the underlying mechanisms.

RESULTS

In mouse models, just two doses of psilocybin given before chemotherapy prevented the development of pain hypersensitivity caused by neurotoxic chemotherapies, with protection maintained across up to six consecutive chemotherapy cycles and over the course of 8 months of follow-up. Critically, psilocybin also protected tumor-bearing mice without reducing tumor growth or altering immune cytokine levels. Using highly selective pharmacological tools, we showed that serotonin 2A receptor activation is both necessary and sufficient for this neuroprotection and that the effect does not require the hallucinogenic properties of psilocybin—a nonhallucinogenic serotonin 2A agonist conferred equivalent protection. Mechanistically, psilocybin acted through two complementary pathways: Peripherally, it preserved the structure and energy supply of distal sensory nerve endings, protecting tactile function and maintaining skin nerve fiber density; centrally, it normalized cortical electrical activity disrupted by chemotherapy. Using live imaging of mitochondria in freshly isolated human peripheral nerves from surgical patients and in human stem cell–derived sensory neurons, we found that psilocybin preserved the active transport of mitochondria along axons, the process by which nerve fibers distribute energy to their most distant, vulnerable endings. Chemotherapy halted this transport and depleted energy at distal nerve endings; psilocybin prevented both effects. Molecular studies traced this protection to a signaling cascade, the TrkB-Akt-PAK5-MAP2-KIF5B pathway, that supports mitochondrial transport along sensory axons. Human donor sensory neurons and patient skin biopsies independently confirmed that this serotonin 2A–mitochondrial trafficking network is conserved in human tissue.

CONCLUSION

Psilocybin prevents CIPN by activating serotonin 2A receptors on peripheral sensory neurons, preserving axonal mitochondrial transport and the local energy supply that nerve endings require for survival. This peripheral neuroprotective mechanism, previously unrecognized for psilocybin, acts in concert with the preservation of central cortical function. Given that no effective preventive treatment currently exists and psilocybin has an established clinical safety profile, these findings position psilocybin and related nonhallucinogenic serotonin 2A agonists as a first-in-class prophylactic strategy warranting clinical evaluation to prevent one of the most common and undertreated toxicities of cancer therapy.

Psilocybin prevents mechanical hypersensitivity through 5HT2A receptors

CIPN is a multifaceted somatosensory dysfunction, including spontaneous pain, allodynia, hyperalgesia, sensory deficits, and central sensitization. To assess psilocybin effects on the mechanical hypersensitivity component of CIPN, we performed von Frey testing in a CIPN mouse model. Cisplatin induced a persistent increase in paw withdrawal response scores compared with controls, indicative of mechanical hypersensitivity, that lasted more than 4 months (Fig. 1A). A single prophylactic dose of psilocybin did not prevent immediate postchemotherapy hypersensitivity but shortened its duration, with scores normalized by day 95 (Fig. 1A). Two psilocybin doses completely prevented the onset of mechanical hypersensitivity (Fig. 1A). To model clinically relevant repeated chemotherapy exposure and cumulative neurotoxicity, mice underwent six consecutive cisplatin cycles with psilocybin pretreatment before each cycle (Fig. 1B). Psilocybin conferred robust and durable protection, maintaining reduced scores across all six cycles relative to cisplatin controls. Psilocybin’s protective effects also extended to taxane chemotherapy in paclitaxel and docetaxel models (fig. S1, A and B), which suggests that neuroprotection is not limited to platinum agents.

To control for systemic cancer effects, we used an established syngeneic oral squamous cell carcinoma (MOC1) model (17). Eight-week-old C57BL/6 mice were injected with MOC1 cells, and once tumors reached 5 mm, they were treated with cisplatin with or without psilocybin pretreatment. Psilocybin-pretreated mice maintained scores similar to those of controls and significantly lower than those of cisplatin-treated mice (Fig. 1C). We next modeled adjuvant chemotherapy after tumor resection. Tumors were resected on day 42 followed by a second cisplatin course without additional psilocybin pretreatment (Fig. 1D). Under these conditions, previously psilocybin-pretreated mice developed mechanical hypersensitivity comparable to that in cisplatin controls, which indicates that a single prophylactic psilocybin course does not protect against a subsequent unpretreated chemotherapy challenge (Fig. 1D). These findings underscore that psilocybin’s neuroprotection requires pretreatment before each cisplatin exposure, consistent with the cycle-matched pretreatment schedule used throughout this study.

We next asked whether psilocybin’s prophylactic effect scales with the number of pretreatments. A single psilocybin pretreatment modestly reduced cisplatin-induced mechanical hypersensitivity (fig. S2A; mixed-effects model, P = 0.048). Two pretreatments were sufficient for full protection; responses were similar across two-, four-, and eight-dose schedules, with no association between dose number and protection by linear regression [β = −0.26, F(1,26) = 0.66, P = 0.424; fig. S2B]. These findings were consistent with a threshold-dependent mechanism in which a two-dose pretreatment is sufficient to engage the full protective state. Psilocybin did not alter tumor growth or systemic cytokine levels at any tested dose (fig. S2C and table S2). Similarly, cytokine profiles in patients with active malignancy were unaffected by psilocybin treatment (fig. S3, A and B, and table S1).

To broadly characterize psilocybin’s protective effects across distinct components of CIPN, we assessed multiple behavioral outcomes. Psilocybin pretreatment significantly attenuated cold allodynia (Fig. 1E) and preserved nesting behavior—a spontaneous, goal-directed activity disrupted in chronic pain states—compared with cisplatin controls (P = 0.058, Cohen’s d = 1.65; fig. S4, A to C). The large effect size suggests biologically meaningful attenuation of affective-motivational dimension of CIPN, complementing the sensory protection observed in acetone and von Frey assays.

To assess the role of 5HT2A signaling in psilocybin-mediated neuroprotection, we confirmed its expression in peripheral sensory neurons. Immunofluorescence of human skin biopsies demonstrated 5HT2A-positive staining colocalizing with the pan-neuronal marker PGP9.5, establishing receptor expression in peripheral nerve fibers (Fig. 1F). RNAscope in situ hybridization of mouse lumbar (L4-L5) dorsal root ganglia (DRG) corroborated this at the transcript level, revealing robust Htr2a expression across sensory neurons (Fig. 1G). Immunofluorescence of mouse hindpaw skin further confirmed 5HT2A localization within peripheral nerve fibers (Fig. 1H). We then tested whether 5HT2A signaling is required for CIPN prevention. Mice received the 5HT2A antagonist ketanserin [intraperitoneally (i.p.)] before psilocybin or vehicle pretreatment. Ketanserin abolished psilocybin’s protection against mechanical hypersensitivity, consistent with a 5HT2A-mediated mechanism (Fig. 1I). To confirm that ketanserin’s blockade reflected 5HT2A antagonism rather than off-target interactions, we repeated the experiment using volinanserin (M100,907), a highly selective 5HT2A antagonist (18–20). Volinanserin fully abolished psilocybin’s protection against CIPN’s mechanical hypersensitivity (Fig. 1J), corroborating a receptor-specific mechanism. We next asked whether psilocybin’s hallucinogenic properties are required for neuroprotection. Mice were pretreated with tabernanthalog (TBG), a nonhallucinogenic 5HT2A agonist (21–23), using the same treatment schedule. TBG conferred protection against mechanical hypersensitivity comparable to psilocybin (Fig. 1J), demonstrating that 5HT2A agonism is sufficient to drive CIPN neuroprotection independent of hallucinogenic activity.

Central 5HT2A signaling preserves cortical function while peripheral nociceptor hyperexcitability remains unaffected

Previous studies have attributed psilocybin effects to central mechanisms, involving cortical and subcortical serotonergic circuits (24). Hence, we tested the potential role of central 5HT2A signaling in CIPN. To block 5HT2A centrally, we administered ketanserin intracerebroventricularly (ICV) (25) before psilocybin and assessed mechanical sensitivity (Fig. 1K and fig. S5) (26). ICV ketanserin abolished centrally mediated head twitch responses (fig. S5) (26). Mice pretreated with ICV ketanserin and psilocybin showed significant mechanical hypersensitivity after cisplatin exposure compared with mice pretreated with psilocybin alone, which suggests that psilocybin mitigates the central sensitization component of CIPN (Fig. 1K).

CIPN is characterized by pathological nociceptor alterations, including increased spontaneous firing and neuronal hyperexcitability within the DRG (27, 28). To determine whether psilocybin prevents cisplatin-induced nociceptor sensitization, we performed whole-cell patch-clamp recordings from L4-L5 DRG neurons (primary afferents innervating the hindpaw) harvested from experimental mice. Subacute recordings, 1 day after cisplatin, revealed an increased proportion of spontaneously active neurons in cisplatin-treated mice compared with both vehicle- and psilocybin-only controls (Fig. 2A). Cisplatin also depolarized resting membrane potential (RMP), decreased rheobase, and elevated action potential firing rates (Fig. 2, B to F), indicating increased neuronal excitability and a reduced firing threshold. Prophylactic psilocybin did not mitigate any of these changes. To confirm these findings upstream of the DRG somas, we measured spinal-level activation in the dorsal horn superficial laminae using c-Fos immunostaining (Fig. 2G). Spinal cords were collected 90 min after the last cisplatin or vehicle dose. Total neuron counts and dorsal horn area were comparable across groups (Fig. 2, H to K). Consistent with our ex vivo recordings, c-Fos was significantly elevated in cisplatin-treated mice compared with controls and was not suppressed by psilocybin pretreatment (Fig. 2, H and I). Notably, no group differences were detected in c-Fos expression within higher-order CIPN-related brain regions, specifically the anterior cingulate cortex, medial prefrontal cortex (mPFC), basolateral amygdala, and insula (fig. S6) (29–31).

These findings led us to hypothesize that rather than suppressing chemotherapy-induced neuronal hyperactivation, psilocybin preserves basal neuronal activity and prevents loss of neuronal function. Clinical electroencephalography (EEG) studies in chronic pain identify reduced lower-frequency power as a biomarker of chronic pain (32, 33). Recordings from acute ex vivo mPFC slices revealed a significant reduction in spontaneous activity in cisplatin-treated mice (Fig. 2L). To evaluate the chronic cortical activity that might underlie the clinical phenotype, we conducted EEG recordings in freely moving mice. EEG electrodes were implanted in mice after cisplatin treatment, with or without psilocybin pretreatment. Cisplatin altered cortical activity in alpha- and beta-band power versus vehicle, consistent with neuropathic pain EEG signatures (34). Psilocybin pretreatment significantly preserved beta power, with a similar trend in alpha (Fig. 2, M and N). Notably, psilocybin also prevented cisplatin-induced sigma power suppression, consistent with normalization of thalamocortical sleep-spindle activity linked to reduced allodynia in rodent chronic pain models (Fig. 2O) (35). Together, these findings support a model in which central network protection acts alongside peripheral mechanisms to underlie psilocybin’s protective effects against CIPN.

Psilocybin preserves tactile function and protects peripheral sensory neurons

Peripherally, a loss-of-function defining feature in long-term CIPN is hypoesthesia, clinically characterized by reduced tactile sensation, which often results from loss of distal mechanosensitive afferent terminals (36, 37). To define sensory neuron subtypes expressing 5HT2A, we colocalized Htr2a mRNA with NF200 and TRPV1 in mouse DRG. Htr2a was expressed across both populations, with a significantly greater coexpression in NF200⁺ myelinated mechanoreceptors (~48%) compared with that in TRPV1⁺ unmyelinated nociceptors (~36%; Fig. 3A), which indicates 5HT2A expression in large-fiber sensory neurons relevant to the mechanosensory deficits. To specifically interrogate tactile hypoesthesia, we used the adhesive tape removal test 2 weeks after cisplatin treatment (38). Cisplatin significantly increased adhesive removal latencies, indicating reduced tactile acuity, which was prevented by psilocybin pretreatment (Fig. 3B). Systemic, but not ICV, ketanserin abolished this protection, indicating that peripheral rather than central 5HT2A signaling preserves tactile acuity (Fig. 3B). This peripheral 5HT2A dependence was further corroborated using volinanserin, which similarly attenuated psilocybin’s protective effect (fig. S7). Together, these findings indicate that psilocybin’s prophylactic effects on CIPN are mediated peripherally, consistent with direct protection of afferent terminals.

To assess structural injury to cutaneous sensory terminals, we quantified intraepidermal nerve fiber (IENF) density in plantar hindpaw skin, at 1 and 4 weeks after cisplatin (Fig. 3, C to F). By 1 week at a standard cisplatin dose (Fig. 3D), we did not see a significant loss of IENF; however, at a higher cumulative dose, cisplatin significantly reduced IENF density compared with vehicle controls. Psilocybin pretreatment significantly preserved IENF density relative to cisplatin-only mice, indicating a sustained structural protection against peripheral sensory fiber loss even at 6 weeks after exposure to cisplatin (Fig. 3F and fig. S8).

To investigate psilocybin’s neuroprotective effects at the cellular level, we examined neurite architecture in cisplatin-treated human induced sensory neurons (iSNs) (Fig. 3G). Neurite complexity was quantified by Sholl analysis (Fig. 3, I to K). Psilocybin pretreatment attenuated cisplatin-induced reductions in neurite length (Fig. 3H) and preserved neurite branching complexity (Fig. 3, J and K). Ketanserin pretreatment abolished psilocybin’s structural protection (Fig. 3, H to K), confirming 5HT2A receptor dependence; these findings were independently replicated in iSNs and in mouse trigeminal ganglia neurons treated with docetaxel (fig. S9, A to F) and in iSNs using volinanserin (fig. S9, G to I).

To define the molecular mechanisms underlying psilocybin’s protective effects, we performed RNA sequencing (RNA-seq) analysis on DRGs harvested from cisplatin-treated mice pretreated with psilocybin or vehicle. Unbiased clustering revealed that DRGs from psilocybin-pretreated mice were transcriptionally similar to untreated DRGs (Fig. 3L). Gene ontology analysis revealed distinct transcriptional programs between groups (Fig. 3L). Cisplatin treatment activated multiple pathways associated with neuronal injury and degeneration, including significant up-regulation of genes linked to neurogenic lesions, abnormal morphology of the nervous system, apoptosis, progressive neurological disorders, and neurological signs. By contrast, psilocybin pretreatment shifted the transcriptional profile toward prosurvival and regenerative programs. The psilocybin-pretreated group showed enrichment of pathways promoting proliferation of neural progenitors while dampening the expression of genes associated with a delay in outgrowth of dendrites and degenerative brain disorders. This transcriptomic signature suggests that psilocybin activates endogenous neuroprotective mechanisms that counteract cisplatin’s neurotoxic effects by promoting neuronal survival pathways and facilitating neural recovery programs.

Psilocybin preserves mitochondrial trafficking and axonal energy distribution

Neuronal mitochondrial dysfunction is central to CIPN pathophysiology (39). To assess psilocybin effects on mitochondrial homeostasis, we measured mitochondrial abundance in human iSNs using MitoBright staining (Fig. 4A). Psilocybin, either alone or as a pretreatment before cisplatin, markedly increased mitochondrial abundance, whereas cisplatin alone had no effect (Fig. 4B). We therefore tested whether psilocybin-mediated neuroprotection is associated with the preservation of mitochondrial motility, distribution, and adenosine 5′-triphosphate (ATP) availability in chemotherapy-affected axons. To investigate mitochondrial trafficking and distribution, we analyzed mitochondrial colocalization with microtubules (TUBB3) in iSNs (Fig. 4C). Given that mitochondrial-microtubule interactions are transient, our analyses were designed to capture the acute effects of 5HT2A receptor activation by psilocin, the active metabolite of psilocybin. Psilocin was detected in iSNs (fig. S10, A and B) and in the DRGs of mice treated with psilocybin (fig. S10C). Cisplatin significantly decreased mitochondrial colocalization with TUBB3, and this effect was blocked by psilocin (Fig. 4D).

Functional assessment of mitochondria in cisplatin-treated iSNs revealed markedly reduced maximal respiratory capacity, measured by mitochondrial uncoupler-stimulated oxygen consumption rate (OCR), indicating impaired reserve capacity (Fig. 4E). Cisplatin also decreased mitochondrial membrane potential (fig. S11, A and B). Psilocybin pretreatment did not restore either deficit (Fig. 4E and fig. S11, A and B), which indicates that psilocybin does not protect from chemotherapy-induced energy impairment at the mitochondria or whole-neuron level. Rather, psilocybin may confer neuroprotection by preserving mitochondrial trafficking, compartmentalization, and energy distribution, thereby preventing focal energy deficits and downstream effects on neuronal signaling and survival.

To evaluate compartment-specific energy homeostasis, we measured ATP separately in the soma and neurites of iSNs cultured in custom-fabricated linear micropattern dishes, enabling live-cell spatial resolution of axonal and somatic compartments (Fig. 4F). Somatic ATP was unaffected across groups. Cisplatin markedly depleted neurite ATP, indicating a compartment-specific energy deficit, with psilocybin pretreatment preserving neurite ATP levels (Fig. 4G). 5HT2A blockade with volinanserin, abolished psilocybin’s preservation of neurite ATP levels, confirming a receptor-dependent energy homeostasis (fig. S11, C and D).

To directly assess subcellular mitochondrial distribution, we analyzed VDAC⁺ mitochondria in iSNs using linear micropattern dishes (Fig. 4H). Cisplatin caused near-complete loss of axonal mitochondria, whereas psilocybin pretreatment preserved the axonal mitochondrial content (Fig. 4I). These findings suggest that psilocybin influences subcellular positioning—a critical factor for sustaining energy supply at distal nerve endings under stress. Transmission electron microscopy (TEM) analysis of iSN axons confirmed that psilocybin attenuated cisplatin-induced reductions in axonal mitochondrial abundance and prevented loss of cristae complexity (Fig. 4, J to L).

Next, we performed live-cell imaging to analyze mitochondrial trafficking in freshly harvested human peripheral nerves, obtained from 29 patients undergoing surgery at MD Anderson Cancer Center (Fig. 4M). Kymography revealed robust bidirectional transport in untreated nerves, with diagonal traces indicating continuous anterograde and retrograde movement (Fig. 4, N and P, and movie S1). Ex vivo cisplatin markedly impaired mitochondrial motility, increasing stationary organelles and vertical kymograph traces, consistent with disrupted axonal transport (Fig. 4P). By contrast, psilocybin pretreatment before cisplatin exposure preserved mitochondrial trafficking, increasing diagonal traces and reducing stationary profiles compared with cisplatin alone (Fig. 4P). Spatial particle tracking analysis using live mitochondria imaging (Fig. 4O) revealed preserved mitochondrial velocity in psilocybin-pretreated nerves after exposure to cisplatin compared with nerves treated with cisplatin alone (Fig. 4Q; also replicated in mouse trigeminal ganglia neurons in fig. S12). These human and mice nerve data indicate that psilocybin mitigates cisplatin-induced deficits in mitochondrial axonal transport and preserves mitochondrial distribution along the nerve.

Psilocybin activates TrkB-Akt-PAK5 signaling to restore mitochondrial motor function

To define the molecular mechanisms underlying psilocybin-induced mitochondrial trafficking and sensory neuron survival, we performed reverse phase protein array (RPPA) profiling in cisplatin-treated iSNs with or without psilocybin. Cisplatin induced neuronal insult, marked by cleaved Caspase-8 induction, and perturbed proteins central to mitochondrial positioning and cytoskeletal integrity (Fig. 5A). Psilocybin pretreatment abrogated these maladaptive changes induced by cisplatin, with lower expression of apoptosis initiation signaling molecules, including γH2AX_pS139, cleaved Caspase-8, pChk1, BCL2A1, interleukin-6 (IL-6), and LCN2. Psilocybin maintained signaling critical for kinesin- and actin-dependent organelle transport machinery, such as Cdc42, LRP6_pS1490, FAK_pY397, and Src_pY416 and preserved expression of proteins central to mitochondrial positioning and cytoskeletal integrity in psilocybin-pretreated iSNs exposed to cisplatin compared with neurons treated with cisplatin alone (Fig. 5A). This led us to hypothesize that psilocybin engages the TrkB-Akt-PAK5-MAP2 signaling axis together with KIF5B, the kinesin-1 family motor protein responsible for microtubule-based forward transport of mitochondria (40)—a well-established mechanism of distal axonal mitochondrial positioning (40–46). We performed Western blot to assess whether psilocybin and its active metabolite, psilocin, activate this axis (Fig. 5, B to S). Consistent with a previous study (47), psilocybin and psilocin produced time-dependent increases in total and phosphorylated TrkB (Fig. 5, C and H) accompanied by increased phosphorylation of Akt (Fig. 5, D and I) and the downstream effector PAK5 (Fig. 5, E and J). PAK5 activation was associated with increased MAP2 (Fig. 5, N and Q), a microtubule-stabilizing protein necessary for efficient long-range transport. Psilocin also increased KIF5B phosphorylation (Fig. 5R), directly linking psilocybin to engagement of the anterograde mitochondrial motor. By contrast, cisplatin alone increased both KIF5B phosphorylation (Fig. 5R) and decreased MAP2 (Fig. 5Q), a maladaptive state consistent with arrested mitochondrial motility observed in our live imaging studies (41). Psilocybin pretreatment counterbalanced this effect, shifting the network toward a functional distribution of mitochondria along axons.

These findings were further validated by RNA-seq of human iSNs showing that psilocybin pretreatment produced a robust protective effect against cisplatin-induced transcriptional disruptions of mitochondrial and trafficking-related pathways (Fig. 5T). Psilocybin exerted significant protective effects by attenuating the cisplatin-induced activation of apoptotic cascades while preserving neurotrophic and regenerative signaling. The differential pathway signatures between cisplatin alone versus cisplatin plus psilocybin treatments revealed psilocybin’s capacity to normalize gene expression in mitochondrial metabolism pathways, including ketogenesis, and restoration of vesicle trafficking pathways, including coat protein complex II (COPII) transport and ESCRT complexes (46, 48–51). Psilocybin further stabilized cytoskeletal organization by maintaining Rho guanosine triphosphatase (GTPase) and CIT signaling (46, 48, 49, 51). Together, these effects support preservation of axonal integrity and mitochondrial distribution, both disrupted by cisplatin exposure. Next, we performed immunostaining of cutaneous nerves in human skin biopsies. This analysis confirmed robust expression of these upstream regulatory molecules, Akt and PAK5, and downstream trafficking–related molecules SNPH and KIF5B (Fig. 5U). Collectively, these data demonstrate that psilocybin counteracted cisplatin-induced disruption by engaging the TrkB-Akt-PAK5-MAP2-KIF5B axis to preserve directional mitochondrial motility. Transcriptome analysis confirmed normalization of mitochondrial function and trafficking networks—COPII transport, ESCRT, Rho GTPase, and cytoskeletal signaling—whereas RPPA showed that other transport regulators (e.g., dynein and myosin) remained unaffected.

We next tested whether TrkB signaling is required for psilocybin-mediated neuroprotection. In vivo, inhibition of TrkB with the selective antagonist ANA-12 (52, 53), abolished psilocybin’s protection against cisplatin-induced mechanical hypersensitivity (fig. S13A). In vitro, ANA-12 negated psilocybin-dependent preservation of neurite ATP distribution (fig. S13, B and C) and neurite complexity (fig. S13, D to F) in cisplatin-exposed iSNs. To determine whether TrkB-dependent neuroprotection extends to mitochondrial integrity within sensory axons in vivo, we used a cell type–specific mitochondrial reporter strategy using MitoTag mice crossed with Nav1.8-Cre mice. MitoTag mice [B6.Cg-Gt(ROSA)26Sor^{tm1(CAG-EGFP)Brsy/J}] harbor a Cre-dependent, loxP-STOP-loxP–gated construct knocked into the Rosa26 locus, encoding a hemagglutinin-tagged enhanced green fluorescent protein (EGFP) fused to the outer mitochondrial membrane targeting sequence (54–56). Crossing with Nav1.8-Cre mice restricted EGFP-tagged mitochondria to Nav1.8⁺ nociceptors and low-threshold C-fiber mechanoreceptors (57–59), enabling in vivo visualization and quantification of mitochondrial abundance and distribution within the sensory fibers most affected by CIPN. This approach enables assessment of mitochondrial trafficking and abundance within intact peripheral sensory axons in vivo. TrkB inhibition with ANA-12 significantly attenuated psilocybin-mediated preservation of mitochondrial abundance, quantified by mitochondrial DNA copy number, and disrupted axonal mitochondrial distribution in Nav1.8⁺ sensory fibers of the hindpaw sensory fibers, indicating that psilocybin’s maintenance of mitochondrial homeostasis in vivo is TrkB dependent (Fig. 5V).

Thus, psilocybin’s neuroprotective action in CIPN is mediated not by a global stabilization of all trafficking machinery, but by a targeted reinforcement of an anterograde signaling-motor network essential for axon-directed mitochondrial transport. This precise, psilocybin-dependent signaling cascade underlies enhanced organelle distribution and sensory neuron survival despite chemotherapy insult.

5HT2A-mitochondrial signaling is conserved in human peripheral sensory neurons

To determine whether these preclinical mechanistic data are recapitulated in humans, we took two independent complementary approaches: transcriptomic profiling of freshly harvested organ donor DRG neurons and multiplex immunofluorescence of skin biopsies from a separate patient cohort. Reactome pathway enrichment analysis of human DRG neurons revealed that psilocybin+cisplatin–treated samples were significantly enriched for mitochondrial bioenergetic pathways relative to cisplatin alone, including respiratory electron transport, complex I and IV biogenesis, mitochondrial translation, and aerobic respiration [false discover rate (FDR) < 0.05; Fig. 6A]. Translational stress-response pathways, including EIF2AK4/GCN2-mediated integrated stress response and ribosome quality control, were similarly enriched, indicating that psilocybin sustains proteostatic surveillance capacity under cisplatin stress. Hierarchical clustering of the 809 leading-edge genes confirmed separation between conditions, with a subset of key nodes in mitochondrial integrity, bioenergetics, and cytoskeletal transport independently validated by RPPA (highlighted in Fig. 6B).

To assess whether the molecular architecture linking 5HT2A receptor expression to the neuroprotective signaling axis is present in human peripheral sensory nerve fibers, we performed multiplex immunofluorescence on skin biopsies from an independent cohort. 5HT2A expression in cutaneous nerve bundles showed significant strong positive correlations with markers of mitochondrial structure and function [TOMM20: correlation coefficient (r) = 0.71; VDAC1-3: r = 0.64], neuronal structural integrity (GAP43: r = 0.67; TUBB3: r = 0.65), and the AKT-SNPH signaling axis (AKT: r = 0.72; SNPH: r = 0.65; SIRT1: r = 0.64; KIF5B+KIF5C: r = 0.31), with all P < 0.001 (Fig. 6C). Notably, KIF5B, the anterograde motor identified as a key downstream effector of psilocybin-mediated mitochondrial trafficking in our preclinical model, significantly colocalized with 5HT2A in human sensory nerve bundles (Fig. 6E). Critically, this coexpression pattern was receptor subtype specific; parallel analysis revealed markedly weaker associations for 5HT2C, most notably for TOMM20 (5HT2A: r = 0.71 versus 5HT2C: r = 0.41; P < 0.001) and UCHL1 (5HT2A: r = 0.47 versus 5-HT2C: r = 0.20; Fig. 6D). This receptor subtype dissociation supports specificity of the 5HT2A mitochondrial neuroprotective axis and argues against a generalized serotonergic receptor effect.

Collectively, these data establish that the molecular network linking 5HT2A signaling to mitochondrial anchoring proteins, anterograde motor machinery, and axonal structural integrity is conserved in human sensory neurons at both the transcriptomic and protein expression levels. The convergence of transcriptomic enrichment in organ donor DRG neurons with protein-level colocalization in an independent patient cohort provides cross-modality, cross-cohort translational validation of the neuroprotective axis identified in our preclinical model and supports the mechanistic plausibility of 5HT2A-targeted intervention for CIPN in the clinical setting.

Animals and study design

All animal experiments were approved by the Institutional Animal Care and Use Committees (IACUC) of University of Texas MD Anderson Cancer Center, and in accordance with the International Association for the Study of Pain (IASP) guidelines. All efforts were taken throughout the experiment to limit the number of animals used and minimize discomfort. C57BL6/J mice were obtained from the Jackson Laboratory (stock no. 000664) and housed in a 12 hour–12 hour light-dark cycle with food and water provided ad libitum (two to five mice per cage). All behavioral experiments were performed using both male and female mice. Mice were 8 to 10 weeks old at the start of the experiment and randomly assigned to treatment groups. All behavioral testing was conducted by experimenters fully blinded to treatment conditions. MD Anderson Cancer Center is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC). For each experiment, the total sample size (n) and the number of males and females per group are specified in the corresponding figure legends. Sex distribution was balanced across treatment groups whenever possible.

CIPN was induced in mice by systemic administration of cisplatin, paclitaxel, or docetaxel. Cisplatin-treated mice received daily i.p. injections of cisplatin (Sigma-Aldrich, MO) using two dosing regimens: a low-dose protocol consisting of 1 mg/kg daily for 7 consecutive days (cumulative dose 7 mg/kg), and a higher cumulative-dose protocol consisting of 2.3 mg/kg daily for 5 days, followed by a 5 day rest period and an additional 5 days of treatment (cumulative dose 23 mg/kg). Paclitaxel-treated mice received i.p. injections of paclitaxel (4 mg/kg) (Teva Pharmaceuticals, CA) every other day for a total of four doses (cumulative dose: 16 mg/kg). Docetaxel-treated mice received i.p. injections of docetaxel (7.5 mg/kg) (Sagent Pharmaceuticals, IL) every other day for a total of three doses (cumulative dose: 22.5 mg/kg). Psilocybin (Usona, WI) was diluted in phosphate-buffered saline (PBS) to 1 mg/ml and administered through i.p. injections of 1 mg/kg for in vivo experiments and to 50 mM concentration for in vitro assays and stored at −20°C. The 1 mg/kg dose of psilocybin was selected after a volume distribution calculation to match the in vitro concentration. Using a simplified total-volume distribution model (Vd ≈ total body water, ~0.7 liters/kg), a 1 mg/kg systemic dose yields an estimated mean tissue concentration of ~1.4 μg/ml. Assuming uniform distribution across well-perfused tissues, this predicts DRG concentrations that correspond to 3 to 7 μM psilocybin. Furthermore, 1 mg/kg is frequently used in rodent studies using psilocybin to achieve robust serotonergic signaling without inducing overt behavioral disruption and elicits the head twitch response, typical phenotype evoked by centrally acting psilocybin (72–77). Ketanserin (+)-tartrate salt (Sigma-Aldrich, MO) was administered through i.p. injections of 4 mg/kg 45 min before injection with either vehicle or psilocybin. ICV ketanserin (100 μM) injections were performed as described by Horton et al. (78), with adjustments. Mice were placed on a flat surface, and the skin over the skull was gently tightened. A 27-gauge needle attached to a microsyringe was aligned with bregma and inserted 1 mm posterior and 1 mm lateral to bregma. A 10 μl drug solution was infused into the lateral ventricle over 30 s. TBG (AChemBlock, CA) was administered through i.p. injections of 10 mg/kg. Volinanserin (Sigma-Aldrich, MO) was administered through i.p. injections of 0.3 mg/kg 20 min before injection with either vehicle or psilocybin. ANA12 (Sigma-Aldrich, MO) was administered through i.p. injections of 1 mg/kg 30 min before injection with either vehicle or psilocybin. To induce oral tumor, MOC1 cells (mouse squamous cell carcinoma cells, accession number CVCL_ZD32, 5 × 10⁵ cells suspended in 20 μl PBS) were injected under brief isoflurane anesthesia into C57BL/6 mice. Tumor growth was monitored every 3 days by caliper measurements.

Blinding and randomization

Animals were randomly allocated to treatment groups, and all behavioral assays, including von Frey, acetone evaporation, nesting, and adhesive removal test, were performed by investigators blinded to group assignments. Tissue processing, imaging, and quantitative analyses, including immunohistochemistry, RNAscope, and morphometric assessments, were conducted on coded samples, with group identity concealed until completion of analysis. These procedures are detailed within the corresponding sections of the Materials and methods.

Von Frey test for mechanical hypersensitivity

Mechanical sensitivity to punctate mechanical stimulation was evaluated using calibrated von Frey filaments (Bioseb, FL). Mice were placed in transparent, bottomless enclosures positioned on a raised wire mesh platform. After an initial habituation period, animals were acclimated for at least 30 min on test days before assessment. A 0.4 g von Frey filament was applied perpendicularly to the plantar region of each hindpaw 10 times, and the number of positive withdrawal responses was recorded. Swift paw withdrawal was considered a positive response. Baseline measurements were taken before treatment initiation (day 0), with subsequent assessments performed at designated time points as described in the relevant sections.

Adhesive removal test for tactile hyposensitivity

To examine hypoesthesia, we used a modification of the adhesive removal test (38). Briefly, an adhesive patch (3 to 4 mm in diameter) was placed on the plantar region of the mouse hindpaw and mouse was immediately placed in a cage with empty flat bottom and transparent walls. The time to first positive reaction to the patch (e.g., paw shaking or attempted removal) was recorded as a measure of the latency to first reaction.

Acetone evaporation test

The acetone-evoked cold allodynia assay was performed following the protocol of Vissers and Meert (79), with slight procedural adjustments. Mice were placed individually in a transparent plastic chamber equipped with a wire mesh floor and allowed a 30-min habituation period. Subsequently, 50 μl of acetone was gently applied to the mid-plantar surface of the left hindpaw using a syringe. Behavioral responses were monitored for 30 s after the application and scored on a four-point scale: 0 = no reaction; 1 = brief paw withdrawal or flick; 2 = sustained flicking or paw stomping; 3 = repetitive flicking accompanied by licking of the stimulated paw. Each animal underwent three consecutive trials separated by 5-min intervals, and the mean value of the three trials was calculated as the final acetone response score.

Nesting behavior assay

Nesting behavior was evaluated using a zone-based nestlet clearance paradigm. Mice were placed in individual cages and provided with six equal pieces of nesting material, each ~1 inch in size, placed in six predefined zones of the cage. After 100 min, nesting behavior was scored based on the number of zones cleared of nesting material, with higher scores indicating increased nest-building activity. Mice that underwent the testing were randomly assigned to treatment groups, and nesting assay was repeated before treatments, 1 day after psilocybin/vehicle treatment, and 5 days after cisplatin/vehicle treatment.

Cytokine profiling in tumor-bearing mice

Circulating cytokine levels were quantified longitudinally in tumor-bearing mice using a multiplex bead-based immunoassay (MILLIPLEX Mouse Cytokine Expansion Panel 1, MCYT1-190K-SPX, Millipore Sigma, MO). Tumors were established by subcutaneous flank injection of MOC1 cells. Mice received two pretreatments with psilocybin (1 mg/kg, i.p.) followed by cisplatin administration (1 mg/kg, i.p., daily for 7 days). Blood samples were collected at baseline (before psilocybin treatment) and 1 week after completion of cisplatin treatment. Plasma cytokine levels were measured according to the manufacturer’s instructions, allowing longitudinal assessment of tumor-associated systemic inflammation and treatment-related changes.

Multielectrode array recordings in acute mouse mPFC slices

Acute mPFC slices from mice (250 μm thick) were prepared using Compresstome VF-510-OZ vibrating microtome (Precisionary Instruments, MA) in cutting ACSF buffer (150 mM sucrose, 40 mM NaCl, 4 mM KCl, 1.25 mM NaH₂PO₄, 26 mM NaHCO₃, 10 mM glucose, 7 mM MgSO₄ and 0.5 mM CaCl₂). The slices were placed in oxygenated recovery buffer (119 mM NaCl, 2.5 mM KCl, 1 mM NaH₂PO₄, 26.2 mM NaHCO₃, 11 mM glucose, 1.3 mM MgSO₄ and 2.5 mM CaCl₂) for 40 min. Slices were then positioned on 6-well MEA plate (Cytoview MEA, Axion Biosystems, GA), aligned with recording electrodes, and maintained at 36°C in oxygenated artificial cerebrospinal fluid. To enhance adherence and maintain close contact, a nylon mesh with a platinum ring was gently placed over the slice. After a stabilization period, baseline recording of spontaneous electrophysiological activity was recorded for 10 min. Signals were acquired at 12.5 kHz per electrode, bandpass filtered (200 to 3000 Hz), and thresholded at 5.5× the root mean square (RMS) noise level for spike detection using AxIS software (Axion Biosystems, GA). Spike detection and raster plots were generated using AxIS Navigator and Neural Metric Tool. Active electrodes were defined as those exhibiting ≥1 spikes/min.

EEG recordings

Video-EEG assay was performed as previously published with a few modifications (80). Animals at 3 to 4 months of age were secured on a stereotaxic frame (David Kopf, CA) under 1 to 2% isoflurane anesthesia. Each mouse was implanted with electrodes (Teflon-coated silver wire, bare diameter 127 μm, A-M systems) under aseptic condition for a total of 4 channels of field potential recordings (81): CH1 (right parietal cortex, 1.7 mm posterior, 1.7 mm lateral to the bregma), CH2 (left frontal cortex, 1.7 mm anterior, 1.7 mm lateral to the bregma), CH3 (right frontal cortex, 1.7 mm anterior, 1.7 mm lateral to the bregma). These three cortical channels shared the reference at the midline over the cerebellum. The fourth channel recorded the electromyography (EMG) in the neck muscles. All the electrode wires together with the attached miniature connector sockets were fixed on the skull by dental cement. After a week of postsurgical recovery, each mouse was habituated in the recording chamber for 48 hours and then received three sessions of 24-hours video recordings separated by 2 days, using the Sirenia data acquisition system (Pinnacle Tech, KS). EEG signals were amplified (10x), filtered (bandpass, 0.5 Hz to 2.5 kHz), and digitized at 5 kHz. All EEG/EMG data together with synchronized mouse behavior were saved on hard drive for offline data analysis. For the analyses presented in this study, data from CH1 were used.

Mouse primary neuronal culture

Two enzyme solutions were prepared, sterile filtered, and warmed to 37°C before DRG harvesting. Solution 1 consisted of 3 μl saturated NaHCO₃, 1 mg L-Cysteine (Sigma-Aldrich, MO) and 60 U papain (Roche, IN) in 1.5 ml Ca²⁺/Mg²⁺-free Hank’s Balanced Salt Solution (HBSS, Corning, NY). Solution 2 consisted of 12 mg collagenase type 2 (Gibco, MA) and 14 mg dispase II (Sigma-Aldrich, MO) in 3 ml of HBSS. Mice were euthanized using isoflurane overdose followed by decapitation. After laminectomy lumbar DRGs were harvested and transferred to Ca²⁺/Mg²⁺-free HBSS for a brief wash and incubated for 10 min at 37°C in enzyme solution 1, followed by 1 min spin at 100g and 10 min incubation at 37°C in enzyme solution 2. DRGs were washed with Ham’s F-12 Medium (Corning, NY) and carefully dissociated in Ham’s F-12 medium containing 10% FBS (Sigma-Aldrich, MO), 1% MEM vitamins (Corning, NY), penicillin and streptomycin (Gibco, MA). The cells were plated on 6 glass coverslips coated with Poly-L-lysine (Sigma-Aldrich, MO) in a 24 well plates and maintained in modified F-12. Cultures were maintained in the incubator at 37°C with a 5% CO₂ atmosphere until use within 8 hours after plating. This dissociation and culturing protocol was used for trigeminal ganglia (TG) cultures. TG cells were kept in primary cell culture in Ibidi 8 well glass bottom plates.

Whole-cell recording in acutely dissociated mouse DRG neurons

Four hours after plating, DRG neurons were transferred to a recording chamber perfused with oxygenated (95% O2 and 5% CO2) extracellular solution at flow rate of 2 ml/min containing 140 mM NaCl, 5 mM KCl, 2 mM CaCl2, 2 mM MgCl2, 10 mM HEPES and 11 mM glucose adjusted to pH 7.4 with NaOH. Glass-micropipettes were filled with an internal solution with 125 mM KCl, 15 mM K-gluconate, 5 mM Mg-ATP, 0.5 mM Na2GTP, 5 mM HEPES, 2 mM MgCl2, 5 mM EGTA, and 0.5 mM CaCl2 adjusted to pH 7.4 with KOH. After reaching whole-cell access the DRG neurons were held at 0 pA to record spontaneous action potentials for 4 min. Neurons were then held at 0 pA, and action potentials were evoked using a series of 300 ms depolarizing current injections in 10-pA steps from −50 pA. The current that induced the first action potential was defined as 1X rheobase. In the input/output recordings neurons were held at 0 pA, and action potentials were evoked using a series of 1000 ms depolarizing current injections in 1x rheobase steps. Only neurons with a RMP of at least −35 mV, stable baseline recordings, and evoked action potentials that overshot 0 mV were used for further experiments and analysis. Series resistance (Rs) was compensated to above 70%. All recordings were made at room temperature.

Immunofluorescence and confocal imaging of mouse tissue

Immunohistochemical analysis of c-Fos was performed on spinal cords and brains collected 90 min after last treatment injection, indicated in respective Results section. Tissue processing, imaging, and quantitative analyses (including immunohistochemistry, RNAscope, and morphometric measurements) were performed using coded samples, with the investigator blind to group identity until data analysis was completed. Tissue was fixed using freshly prepared 4% paraformaldehyde (PFA) delivered through intracardial perfusion. Spinal cords and brains were dissected and placed in 4% PFA at 4°C for 2 hours, followed by 30% sucrose overnight and sectioned to 50 μm thick sections on Leica CM3050S cryostat (Leica Biosystems, IL). Sections were blocked in 0.1 M Phosphate buffer with 10% Normal Goat Serum and 0.3% Triton-X and incubated in c-Fos (225-308, Synaptic Systems, Göttingen) and NeuN (MAB377, Millipore, MO) primary antibodies for 48 hours at 4°C. Tissues were incubated in Alexa Fluor secondary antibodies (A-21450 and A-21127, Thermo Fisher, MA) for 2 hours at room temperature. Sections were slide mounted and cover slipped with Prolong Diamond Antifade Mountant (Thermo Fisher, MA). Spinal cord slice images were captured using Olympus Fluoview FV4000 confocal microscope at 20X magnification. Brain slice images were captured using Andor Revolution XDI WD Spinning Disk Confocal microscope (Oxford Instruments, MA). Total neuronal counts and c-Fos positive neurons were counted using IMARIS Image Analysis and Tracking Software (Bitplane AG, Switzerland).

Skin biopsies collected from plantar hindpaw regions for IENF density analysis were fixed using Zamboni fixative overnight, followed by 30% sucrose overnight and cut to 50 μm thick sections on Leica CM3050S cryostat. Samples were immunostained using a free-floating staining method for collagen IV (1:50, 1340-01, SouthernBiotech, AL) and PGP9.5 (1:100, ab108986, Abcam, MA) and images were captured using Nikon Eclipse Ti2 A1R confocal microscope. Secondary antibodies used were Cy3 AffiniPure (1:400, 711-165-152, Jackson ImmunoResearch, PA) and DyLight 488 (1:200, ab96931, Abcam, MA). Collagen IV staining was used to identify basement membrane and nerve fibers crossing were manually counted and normalized to a basement membrane length.

Mouse peripheral nerve samples were fixed in formalin and embedded in paraffin. Multiplex immunofluorescence staining was performed on formalin-fixed, paraffin-embedded tissue sections using primary antibodies against NFH (ab8135, Abcam; 1:1000; Opal 520, 1:100) and 5-HT2A receptor (PA5-120747, Thermo Fisher Scientific; 1:1000; Opal 690, 1:100), followed by 4′,6-diamidino-2-phenylindole (DAPI) nuclear counterstaining. Staining was performed using Opal-based detection on the Xmatrx platform according to the manufacturer’s protocol.

Immunohistochemistry and RNAscope of mouse DRG tissue

Combined immunohistochemistry and RNAScope were carried out on mouse DRGs. Adult mice lumbar DRGs were dissected and immediately embedded in OCT (Thermo Fisher, MA) and frozen on dry ice. Frozen tissues were cryostat sectioned at 20 μm onto SuperFrost Plus charged slides. mRNA transcripts were detected using the RNAscope Fluorescent Multiplex Assay (Advanced Cell Diagnostics, CA) and RNAscope Fluorescent Multiplex Reagent Kit (no. 320851 and no. 323100). The RNAscope catalog probes used to detect were Htr2a (no. 401291) and Trpv1 (no. 313331-C2). Counterstaining for NF200 (no. MAB5266, Millipore, MO) was done with immunohistochemistry on the same tissues. Tissues were blocked in 0.1M Phosphate Buffer with 10% Normal Goat Serum and 0.3% Triton-X. Primary antibody was incubated overnight at 4°C. Secondary antibodies (Goat anti-Mouse IgG2a AlexaFluor-647 Thermo Fisher, cat. no. A21241) were incubated for 1 hour at room temperature. Coverslips were mounted with Prolong Gold Antifade (Thermo Fisher, MA).

Mitochondrial DNA isolation from peripheral sensory neurons

To enable cell type–specific interrogation of mitochondrial content in peripheral sensory neurons, we used a genetic mitochondrial labeling strategy using MitoTag reporter mice crossed with Nav1.8-Cre driver mice. MitoTag mice [B6.Cg-Gt(ROSA)26Sor ^{tm1(CAG-EGFP)Brsy/J}; The Jackson Laboratory no. 032290] harbor a Cre-dependent loxP-STOP-loxP cassette knocked into the Rosa26 locus, driving expression of an EGFP fused to an outer mitochondrial membrane targeting sequence upon Cre-mediated recombination sequence (54–56). When crossed with Nav1.8-Cre mice (Scn10a-Cre; The Jackson Laboratory), this system results in selective labeling of mitochondria within Nav1.8⁺ peripheral sensory neurons (57–59). This approach enables enrichment and isolation of mitochondria specifically from nociceptive neuron populations without the need for cell sorting, preserving mitochondrial integrity and minimizing contamination from nonneuronal cell types. Experimental cohorts were generated by crossing Nav1.8-Cre mice with MitoTag reporter mice, and genotyping was performed by tail biopsy PCR at weaning (postnatal day 21).

For mitochondrial isolation, plantar skin tissue from mouse paws was collected and pooled per sample. Tissue was dissociated and GFP-labeled mitochondria were isolated using anti-GFP magnetic microbeads (uMACS GFP Isolation Kit, Miltenyi Biotec, Germany) after magnetic column separation. Custom buffers [PBS containing 0.5% bovine serum albumin (BSA) and 5 mM EDTA] were used for column equilibration, separation, and elution, and columns were cleaned using sucrose-Tris buffer (250 mM sucrose, 10 mM Tris).

DNA was extracted from isolated mitochondria using the DNeasy Blood & Tissue Kit (no. 69504, Qiagen, MD). Mitochondrial DNA (mtDNA) copy number was quantified by quantitative PCR using the Mouse Absolute Mitochondrial DNA Copy Number Quantification Kit (MTM-CNQ-100, RayBiotech, GA). Reactions contained 5 ng DNA per sample and were performed in triplicate according to the manufacturer’s instructions.

Human neurons in vitro

Human neurons used in this study were obtained from human donors or were obtained as commercially available human induced pluripotent stem cells derived sensory neurons (iSNs). Donors or next of kin provided written consent to donate tissue samples for the study before inclusion, for which the protocol was reviewed and approved by The University of Texas MD Anderson Cancer Center Institutional Review Board (IRB no. 2013-0871). Human iSNs were cultured under standard conditions according to supplier protocols (RealDRG cell line, catj. no. 1020, Anatomic, MN). Cells were seeded into multiwell plates coated with iMatrix 511-SILK (Anatomic, MN) and allowed to mature in Chrono Senso-MM maturation media (Anatomic, MN) before drug treatment. Psilocybin (10 μM), psilocin (Sigma-Aldrich, MO), cisplatin (5-10 μM), docetaxel (10 nM), ketanserin (10 μM), volinanserin (10 nM), and ANA-12 (10 μM) were used at the indicated concentrations according to the experimental design. In selected experiments, cells were pretreated with ketanserin, volinanserin, or ANA-12 to inhibit psilocybin-mediated signaling. Ketanserin, volinanserin, and ANA-12 were applied 45, 15, and 15 min before psilocybin exposure, respectively.

For the organ donor DRG neurons preparation, a section of the thoracic and lumbar spine was removed upon completion of organ collection and DRGs were dissected free (82, 83). All excised DRGs were immediately immersed in cold (~4°C) sterile balanced salt solution in 50 ml centrifuge tubes, which were sealed and placed on ice for transport to the laboratory. DRG was cut into several small ~1 mm² sections using a surgical scalpel and placed in a Petri-dish with 2 ml DMEM/F12 containing trypsin, collagenase, and DNase and shaken for 20 min in a heated (37°C) orbital shaking incubator. The digestion solution was then collected and put into 15 ml centrifuge tubes with 10 ml blocking solution (DMEM/F12 with 10% horse serum) while the remaining tissue was again submerged in 2 ml of fresh digestion solution. This process was repeated until the sections were fully digested. Tubes with combined digestion and blocking solution were centrifuged at 23°C and 180g for 5 min. The supernatant was removed, and the cell pellet resuspended with 3 ml of culture media (DMEM/F12 with 10% horse serum) and then filtered through a 100-μm cell strainer. The resulting cell suspension was collected and centrifuged at 23°C and 180g for 5 min. The supernatant was again removed, and the final cell pellet resuspended with culture media. Cells were plated on poly-L-lysine–coated glass sheets and held in culture dishes with culture media until used.

Immunofluorescent analysis of iSNs and TGs

iSNs or TG cultures were fixed with 4% PFA for 30 min at 4°C. After washing in PBS and blocking using 5% normal donkey serum, cells were incubated with respective primary antibodies: Mitobright lt green (1:1000, NC1879569, Thermo Fisher, MA), NFH (1:2000, AB4680, Abcam, MA), VDAC (1:200, AB14734, Abcam, MA), TUBB3 (1:1000, AB18207, Abcam, MA), Tomm20 (1:500, MABT166, Sigma-Aldrich, MO), followed by secondary antibodies Alexa Fluor 594 (1:500, AB150176, Abcam, MA), Alexa Fluor 488 (1:500, AB150113, Abcam, MA), and Alexa Fluor 647 (1:500, AB150075, Abcam, MA). Imaging was performed using the Cytation 7 imaging platform and analyzed with Gen5 software (v3.17). Morphometric parameters—including total neurite length, branch count, and soma number—were quantified using automated tracemasks calibrated for fluorescence images immunostained for NFH. Thresholding, edge detection, and mask smoothing were optimized for each batch and applied uniformly across replicates. Measurements were normalized to area analyzed to account for sparse areas of wells, and only areas of wells with neurons present were analyzed. Colocalization of Tomm20 with TUBB3 was used to compare axonal mitochondrial distribution across treatment groups. Relative fluorescent intensity of MitoBright lt green staining was used to compare mitochondrial abundance in iSNs.

To compare the ATP distribution in iSNs, cells were stained using Bio Tracker ATP-Red live-cell dye (40 μM, Sigma Aldrich, MO) for 20 min in dark at 37°C and washed twice using the Chrono Senso-MM medium (Anatomic, MN). Images were captured using the Cytation 7 imaging platform at 40x magnification and analyzed in ImageJ software (National Institute of Health).

Mitochondrial membrane potential was assessed using Dojindo Molecular Technologies MT-1 MitoMP Detection Kit (NC1933275) where the cells were incubated with MT-1 MitoMP dye according to the manufacturer’s instructions, washed with culture medium, and imaged live using the Cytation 7 imaging platform at 40× magnification. Relative fluorescence intensity was quantified in ImageJ using identical acquisition and analysis settings across treatment groups.

For immunofluorescent mitochondrial distribution imaging iSNs were cultured on custom IBIDI plates containing micropatterned grooves to facilitate analysis of intracellular mitochondrial distribution and enable clear delineation between somatic and axonal compartments. Fluorescence imaging was performed using a Cytation 7 Cell Imaging Multi-Mode Reader at 40× magnification, acquiring images in three channels: DAPI (nuclear staining), Texas Red (β3-tubulin immunofluorescence to show cellular architecture/morphology), and GFP (VDAC immunofluorescence to detect mitochondria). All image acquisition was conducted by an investigator blinded to experimental conditions. Quantitative analysis of axonal mitochondrial content was performed using Gen5 software, where soma regions were identified using DAPI staining to define nuclear boundaries with surrounding TUBB3-positive cytoplasm delineating the cell body perimeter, while axonal regions were defined as TUBB3-positive areas extending beyond the soma. An initial mask was used to detect the cells from background using the TUBB3 signal to guide recognition with the Cytation 7 software. A binary mask was generated to identify TUBB3-positive axonal regions, serving as the analytical region of interest, within which VDAC-positive areas were quantified using consistent upper and lower fluorescence intensity thresholds applied across all experimental conditions. Axonal mitochondrial content was expressed as the total VDAC-positive area within TUBB3-defined axonal regions, with fluorescence intensity thresholds for VDAC detection remaining constant across all conditions and imaging sessions to ensure reproducibility and minimize inter-experimental variability.

Sholl analysis

Neurite complexity was assessed by Sholl analysis in ImageJ using the Neuroanatomy plugin (SNT). Immunofluorescence images of NFH-stained neurons (NFH, 1:2000, AB4680, Abcam, MA) were thresholded to preserve neurite continuity while minimizing background signal. Cleanly isolated neurons were manually selected using the polygon tool and analyzed by an investigator blinded to condition. At least 16 neurons per condition were analyzed across multiple images per image set. Soma centers were defined using the single-point tool, and Sholl analysis was performed using the “Sholl: From Image” function. Neurite intersections were quantified at increasing radial distances from the soma using identical end radii and step sizes across groups. Total intersections, critical value, dendrite maximum, and Schoenen Ramification Index were exported for downstream statistical analysis.

Maximal mitochondrial respiration capacity

Mitochondrial respiration was analyzed using Agilent Seahorse XFp Flux analyzer with a Seahorse XFp Mito Stress Test Kit (Agilent Technologies, CA) according to manufacturer’s recommendations. OCRs were measured in human iSN culture seeded at 15,000 cells per well in a 96 well Agilent Seahorse XF Cell Culture Microplate with sequential injection of 1.5 μM oligomycin, 3.5 μM FCCP, and each 0.5 μM rotenone/antimycin. Maximal respiratory capacity was identified as the OCR after FCCP. Respiration measurements were performed on at least three biological replicates.

TEM

Human iSN cells were fixed with a solution containing 3% glutaraldehyde and 2% PFA in 0.1 M cacodylate buffer, pH 7.3, washed in 0.1 M sodium cacodylate buffer, treated with 0.1% Millipore-filtered cacodylate buffered tannic acid, and postfixed with 1% buffered osmium tetroxide. Staining solution contained 1% Millipore-filtered uranyl acetate. Cells were dehydrated in graded concentrations of alcohol, infiltrated and embedded in LX-112 medium. Samples were polymerized in oven for 3 days at 60°C. Samples were examined using a JEM 1010 transmission electron microscope (JEOL Inc., MA) at an accelerating voltage of 80 kV. Digital images were obtained using AMT Imaging System (Advanced Microscopy Techniques Corp., MA).

Mitochondrial morphology was quantified from TEM images by an investigator blinded to treatment condition. Mitochondria within neurite/axonal profiles were manually identified in ImageJ, and morphological features including mitochondrial area, aspect ratio, circularity, and cristae integrity were assessed using identical analysis settings across groups. Quantified values were exported for downstream statistical analysis

Live-cell kymography imaging

Mitochondrial motility in live mouse trigeminal ganglion neurons and ex vivo cultured human peripheral nerve samples obtained from 29 patients after informed consent under an IRB-approved protocol (UTMDACC LAB08-0848). Human nerve samples were maintained in culture medium and were subsequently treated using conditions similar to those described for human iSNs. Mitochondria were visualized using MitoBright Light Green fluorescent labeling. Time-lapse microscopy captured 5-min sequences at 10 pixels/μm resolution. Mitochondrial tracking used TrackMate v7.14.0 with Laplacian of Gaussian detection (5-pixel radius) and Simple Linear Assignment Problem linking (15-pixel maximum distance, 2-frame gap tolerance). Mean velocity and total distance parameters were extracted from semi-automated analyses of ≥10 regions of interest containing single axon sections per condition.

Western blot

Human iSN cells were washed with cold PBS and scraped into RIPA lysis buffer with 1× Halt Protease and Phosphatase Inhibitor Cocktail (78440, Thermo Fisher, MA). Protein assays were performed using a DC Protein Assay kit (5000111, Bio-Rad, CA) and were used to load equivalent total protein on 10% (SDS)-polyacrylamide gel. Protein was transferred onto a polyvinylidene-difluoride membrane. Membranes were blocked with 2% BSA and incubated with primary antibody at 4°C overnight. Primary antibodies used included TrkB polyclonal antibody (1:500, 13129-1-AP, Thermo Fisher, MA), p-TrkB (Tyr516) polyclonal antibody (1:500, PA5-36695, Thermo Fisher, MA), Akt monoclonal antibody (1:1000, 4691S, Cell Signaling, MA), p-Akt (Ser473) monoclonal antibody (1:1000, 4690S, Cell Signaling, MA), p-PAK5 antibody (1:1000, 3241, Cell Signaling, MA), SNPH monoclonal antibody (1:1000, Ab192605, Abcam, MA), anti-β-actin monoclonal antibody (1:3000, A5441, Sigma Aldrich, MO), p-Kinesin 5B (Ser154) polyclonal antibody (1:500, PA5-105705, Thermo Fisher, MA), MAP2 polyclonal antibody(1:2000, 17490-1-AP, Thermo Fisher, MA), KIF5B polyclonal antibody (1:1000, 21632-1-AP, Thermo Fisher, MA), p-MAP2 (Ser136) antibody (1:1000, 4541S, Cell Signaling, MA), Membranes were washed with TBS + 0.1% Tween and incubated with secondary antibodies anti-rabbit IgG, HRP-linked antibody (1:3000, 7074P2, Cell Signaling, MA) and m-IgGκ BP-HRP (1:3000, sc-516102, Santa Cruz, TX). Western blot quantitation was done by densitometry measurements using ImageJ software (National Institute of Health). Ratios were calculated relative to actin.

RNA-seq

RNA of iSNs and of human and mouse DRGs was extracted using the RecoverALL Total Nucleic Acid Isolation kit (Thermo Fisher, MA) according to the manufacturer’s instructions. RNA libraries were prepared and sequenced at the University of Houston Sequencing and Editing Core according to standard protocols. Enrichment for mRNA from total RNA was performed by coding-region-specific biotinylated capture probes, which were selected using streptavidin magnetic beads. mRNA-enriched libraries were prepared using the QIAseq Stranded Total RNA Kit (Qiagen, MD) with 200 ng of input RNA. RNA was fragmented, reverse-transcribed into cDNA, ligated with sequence adaptors and amplified by PCR. Size selection for libraries was performed using SPRIselect beads (Beckman Coulter, CA). The library purity was analyzed using the DNA HS1000 tape with a 4200 TapeStation system (Agilent Technologies, CA) and the Qubit fluorometer (Thermo Fisher, MA). The prepared libraries were pooled and sequenced on the NovaSeq 6000 system (Illumina, CA), using an S4 flow cell at paired-end 150 base pair configuration at 20 to 30 million read depth for each sample.

Neuron transcriptome analyses

The raw sequencing data (FASTQ files) were imported into Qiagen CLC Genomics Workbench 20.0.4. Initial quality control of the sequencing reads was performed using the built-in tools, including the Quality Control function, to evaluate read quality metrics, such as read length distribution, quality scores and potential adapter contamination. Low-quality reads and contaminants were trimmed or filtered out. The cleaned sequencing reads were mapped to the GRCh38.p14 reference genome using the Map Reads to Reference tool in the CLC Genomics Workbench. Gene expression levels were quantified using the Count Features tool. This step involved counting the number of reads mapped to each gene based on the annotation file corresponding to the reference genome. The output generated a count table indicating the raw read counts for each gene across all samples. Read counts were normalized using the trimmed mean of M method implemented in the R Bioconductor package edgeR (v.4.2.1) 71 to determine the abundance of each gene. Next, we fit a negative binomial generalized log-linear model to estimate the quasi-dispersions with empirical Bayes moderation (glmQLFit function), followed by hypothesis testing with empirical Bayes quasi-likelihood F-tests. P-value adjustment was performed using the Bonferroni correction method.

RPPAs

iSNs were seeded in a density of 500,000 cells per well on a 6 well plate, 3 wells were pooled per sample (1.5 × 10⁶ cells per sample) with four replicates per treatment group. Plates were coated with Poly L lysin (Sigma Aldrich, MO) for 1 hour at 37°C followed by iMatrix 511-silk (Anatomic, MN), 1:50 dilution in dPBS (without calcium and magnesium) for 2 hours at 37°C. Cells were cultured in Chrono Senso-MM media (Anatomic, MN). Cells were treated 24 hours after seeding with psilocybin, ketanserin (45 min before psilocybin), cisplatin, combination of psilocybin and cisplatin, and combination of ketanserin, psilocybin and cisplatin. Cells were treated with psilocybin every 24 hours for the first 48 hours (two doses total). For groups receiving ketanserin, the antagonist was applied 45 min before each psilocybin dose. After this pretreatment phase, the drug-containing medium was replaced with either fresh Chrono Senso-MM medium (Anatomic, MN) or medium containing cisplatin for 24 hours. After the first cisplatin exposure, cells were given a 24-hour recovery period in fresh Chrono Senso-MM medium (Anatomic, MN) before receiving a second 24-hour cisplatin treatment. All drugs were applied at 10 μM. On day 6, cells were washed with calcium- and magnesium-free dPBS and lysed in RIPA buffer supplemented with protease inhibitor for 20 min on ice. Then lysate was collected by scraper and centrifuged for 10 min, 14,000 rpm, at 4°C. Supernatant was collected and 80 μg total protein per sample was processed for RPPA. Primary data analysis used custom software to determine signal intensity, curve construction, and relative protein expression and modification.

Canonical pathway integrative analysis

Canonical pathway activation and down-regulation were predicted in QIAGEN Ingenuity pathway analysis. Ingenuity pathway analysis was used to identify the cascade of upstream and downstream regulators of the core gene set. Ingenuity pathway analysis uses a priori knowledge of expected interactions between transcriptional regulators and their target genes and proteins stored in Ingenuity Knowledge Base, a scientific literature-based database, https://www.g6g-softwaredirectory.com/bio/cross-omics/dbs-kbs/20018U-Ingenuity-Knowledge-Base.php.

Liquid chromatography–mass spectrometry (LC-MS) analysis of conversion of psilocybin to psilocin

Human iSNs were seeded at a density of 35,000 cells per well on 24 well plate and allowed 2 weeks to mature. Coating and culture procedures followed vendor’s protocol and is described in previous sections. Cells were treated with 10 μM psilocybin and incubated in dark at 37°C. Supernatant was collected at 0, 2, 4, 8, 16, and 32 hours. Four replicates per time point were used to measure concentration of psilocybin and converted psilocin using LC-MS. As a control we used media with psilocybin, without cells, after 2 and 16 hours of incubation. Calibration curve was prepared from blank media and media with psilocybin at 1, 10, 100, and 1000 nM.

To measure psilocin levels directly in DRGs, mouse DRGs were harvested at multiple time points after systemic psilocybin administration (1 mg/kg, i.p.). DRG tissue was collected, processed for LC-MS analysis, and psilocin concentrations were quantified.

Multiplex immunofluorescence staining of human skin samples

Multiplex immunofluorescence staining was performed on human skin samples obtained from 14 patients after informed consent under an IRB-approved protocol (UTMDACC LAB08-0848) using the Lunaphore COMET platform according to the manufacturer’s protocol. A panel of markers relevant to neuronal structure and mitochondrial function was included. Neuronal bundles and filaments were identified from multiplexed tissue images using the GAP43 and UCHL1 channels, and segmentation with axonal annotation was performed using the AxonFinder framework (84). For each segmented region, morphological features and marker intensity values were extracted using the regionprops function from scikit-image. The analysis focused on markers relevant to neuronal integrity and mitochondrial motility, including 5HT2A, AKT, GAP43, SIRT1, SNPH, TUBB3, VDAC1 TOMM20, and KIF5B/KIF5C. Marker intensity values were normalized at the feature level using the UniFORM framework to align intensity distributions while preserving biologically meaningful signal variation. For each marker and structural category (bundles and filaments), normalized intensity distributions were modeled using a two-component Gaussian mixture model to distinguish negative and positive populations, with positivity thresholds defined at the intersection of distributions and visually validated. Each bundle or filament was then classified as marker-positive or -negative. Marker-centered correlation analyses were then performed, restricting each analysis to marker-positive structures and computing correlations between the normalized intensity of the index marker and all other markers within the same subset. Pearson and Spearman correlation coefficients were used to assess linear and monotonic relationships, respectively, and the coefficient of determination (R²) was used to evaluate the strength of linear associations.

Patient plasma cytokine measurements

Blood samples were obtained from patients after informed consent under an IRB-approved protocol (NCT05398484, NCT04593563). Blood was collected and processed according to standard hospital procedures, and plasma cytokine levels of IL-6 and tumor necrosis factor–α (TNF-α), were measured by the clinical laboratory using validated diagnostic assays. Cytokine concentrations were reported in picograms per milliliter and used for downstream analysis.

Statistical analysis

Sample sizes (n) are indicated in the corresponding figure legends and represent biological replicates, including individual animals, independent cell cultures, or cells per fields of view, as appropriate. Experiments were performed at least two independent times and/or included sufficient biological replicates to support reproducibility and statistical analysis.

Comparisons between two independent groups were performed using two-tailed unpaired Student’s t tests. Comparisons among more than two groups were analyzed using one-way analysis of variance (ANOVA). Experiments with two independent variables were analyzed using two-way ANOVA or mixed-effects models fitted with restricted maximum likelihood (REML) to account for repeated measures and missing values. When overall effects were significant (P < 0.05), pairwise comparisons were performed using Tukey’s multiple-comparison test.

Effect sizes were calculated to quantify the magnitude of observed effects. For t tests, effect sizes were estimated using Cohen’s d. For ANOVA and mixed-effects models, effect sizes for fixed effects were estimated using partial eta squared (η²p), derived from the corresponding F statistics and degrees of freedom. For mixed-effects models, variance components were also used to estimate the intraclass correlation coefficient (ICC) to assess the contribution of subject-level variability. All tests were two-sided. Statistical analyses were performed using GraphPad Prism.

Funding:

This work was supported by the following funding sources: M.A. was supported by the National Institutes of Health (NIH), National Cancer Institute grant R37 1R37CA242006-01A1; a Stiefel family Discovery award; Institutional Research Grant, R01 DE032018 (NIDCR) funding; and the Disruptive Science Moonshot award, MDACC. P.M.D. was supported by NIH grants 5U19NS130608 and R01NS132483 and holds the H.E.B. Endowed Professorship in Basic Science. S.E.E. was supported by the Cancer Early Detection Advanced Research Center (CEDAR; Project ID no. 2023-1768). J.C.B. was supported by NIH, NCI R01CA286651; a Breast Cancer Research Foundation NextGen Grant for Transformative Cancer Research (20-20-26-BORN); a Department of Defense Idea Development Award (W81XWH-22-1-0871); and the NIH Cancer Center Support Grant 5P30CA045508-36. Data were generated in part using the Innovation Nexus and Flow Cytometry and Cellular Imaging Facility and Research Animal Support Facility shared resources, which receive partial support from the National Cancer Institute under grant P30 CA016672 to MD Anderson Cancer Center. The research reported in this publication did not directly arise from the grant P30 CA016672 to MD Anderson Cancer Center and is not within the scope of such grant. The FV4000 was generously made available on loan from Evident Scientific. CCSG grant NIH P30CA016672 supports the High-Resolution Electron Microscopy Facility. J.C. is supported by the CPRIT Research Training Program (RP260817) and was an MD Anderson CATALYST trainee. A.L. was an MD Anderson CATALYST trainee.

Author contributions:

Conceptualization: M.H., L.P., H.S., P.M.D., M.A.; Data curation: M.H., L.P., H.S., S.N., Y.V., C.L.S., W.M., T.X., F.O.G.-N., S.S.K., H.E., P.Z., J.C. A.L., S.T., P.H.G.; Formal analysis: M.H., L.P., S.N., H.S., W.M., T.X., F.O.G.-N., Y.V., H.E., S.S.K., P.Z., J.C., S.T., P.H.G., K.A.A., M.J.K., S.E.E., J.C.B., E.G., G.H.J.; Funding acquisition: P.M.D., M.A.; Investigation: M.H., L.P., H.S., S.N., Y.V., C.L.S., W.M., T.X., H.E., P.Z., J.C., A.L., D.L.S., M.R.M., Z.-H.L., S.T., P.H.G., K.A.A., M.J.K., S.E.E., J.C.B., G.H.J.; Patient nerve tissue procurement: M.R.M., Z.-H.L.; Methodology: M.H., L.P., S.N., H.S., W.M., T.X., F.O.G.-N., Y.V., H.E., P.Z., J.C., A.L., D.L.S., L.C., J.C.B., E.G., P.M.D., M.A.; Project administration: P.M.D., M.A.; Resources: M.R.M., Z.-H.L., K.A.A., M.J.K., S.E.E., L.C., J.C.B., P.M.D., M.A.; Software: H.S., W.M., T.X., P.Z.; Supervision: L.C., J.C.B., P.M.D., M.A.; Visualization: M.H., L.P., S.N., H.S., F.O.G.-N., P.Z.; Writing – original draft: M.H., L.P., G.H.J., P.M.D., M.A.; Writing – review & editing: M.H., L.P., S.N., H.S., W.M., D.Y., T.X., Y.V., H.E., P.Z., M.L.U., J.C., A.L., M.R.M., Z.-H.L., S.E.E., M.J.K., L.C., J.C.B., E.G., G.H.J., P.M.D., M.A.

Competing interests:

The contributions of the NIH authors are considered works of the United States government. The findings and conclusions presented in this paper are those of the authors and do not necessarily reflect the views of the NIH or the US Department of Health and Human Services. M.A. reports filing US provisional patent application no. 64/106,659 related to the use of 5-HT2A receptor agonists for toxicity mitigation. The remaining authors have no competing interests to disclose.

Data, code, and materials availability:

RNA-seq datasets generated during this study have been deposited in the Gene Expression Omnibus (GEO) under accession no. GSE333441 (“Psilocybin prevents chemotherapy-induced peripheral neuropathy via mitochondrial trafficking preservation”) and will be made available in accordance with the journal’s requirements.

License information:

The PDF file includes:

Other Supplementary Material for this manuscript includes the following: