Sensory nerve release of CGRP increases tumor growth in HNSCC by suppressing TILs

Summary

Background:

Perineural invasion (PNI) and nerve density within the tumor microenvironment (TME) have long been associated with worse outcomes in head and neck squamous cell carcinoma (HNSCC). This prompted an investigation into how nerves within the tumor microenvironment impact the adaptive immune system and tumor growth.

Methods:

We used RNA sequencing analysis of human tumor tissue from a recent HNSCC clinical trial, proteomics of human nerves from HNSCC patients, and syngeneic orthotopic murine models of HPV-unrelated HNSCC to investigate how sensory nerves modulate the adaptive immune system.

Findings:

Calcitonin gene-related peptide (CGRP) directly inhibited CD8 T cell activity in vitro, and blocking sensory nerve function surgically, pharmacologically, or genetically increased CD8 and CD4 T cell activity in vivo.

Conclusions:

Our data supports sensory nerves playing a role in accelerating tumor growth by directly acting on the adaptive immune system to decrease Th1 CD4 T cells and activated CD8 T cells in the TME. These data support further investigation into the role of sensory nerves in the TME of HNSCC and points towards the possible treatment efficacy of blocking sensory nerve function, or specifically inhibiting CGRP release or activity within the TME to improve outcomes.

Graphical Abstract

eTOC blurb

Darragh et al. used orthotopic murine models of HNSCC and human RNA sequencing and proteomics to better understand the role of sensory nerves in the tumor microenvironment. They show that sensory nerve release of CGRP is likely decreasing the activation of CD4 and CD8 T cells in the TME.

Introduction

How intratumoral nerves play a role in tumorigenesis or tumor progression is relatively unknown. Patients with head and neck squamous cell carcinoma (HNSCC) who have perineural invasion (PNI) have a worse prognosis and are more likely to have local and metastatic recurrence ^1–5. Studies in prostate cancer ^6,7, gastric cancer ⁸, pancreatic cancer ⁹, and melanoma¹⁰, have shown that the interaction between nerves and the tumor microenvironment (TME) modulates tumor growth. In HNSCC, two studies have shown that cancer cell-nerve crosstalk contributed to a reduction in survival ^11,12, while in others, not in HNSCC, nerve-immune crosstalk was the driving factor^8,10. Although these studies have advanced our understanding of the underlying effects of nerves within the TME, questions remain about how nerves may be interacting with the immune system in HNSCC.

Pre-clinical studies have shown that denervation in gastric cancer increases immune cell infiltration of activated CD4 T cells⁸ and a recently published paper described the pro-tumorigenic role of sensory nerves and how these nerves can directly influence adaptive immune cells in murine models of melanoma¹⁰. Although there are only a few published studies on the interaction between nerves and immune cells in cancer, an infection model with Staphylococcus pyogenes also showed that denervation or sensory nerve block, whether pharmacologically or genetically, increased immune cell infiltration resulting in the host successfully fighting off the infection¹³. Together, these studies provide evidence that nerves can directly influence the immune system in a localized area, which may be contributing to how nerves may reduce an anti-tumor immune response.

Immunotherapy has quickly become a key addition to an oncologist’s toolbox for treating a variety of cancers, but its success in HPV-unrelated HNSCC has been limited^14,15. Current standard of care treatment for HNSCC includes radiation therapy (RT), a known immunomodulator, which has been shown to increase effector T cells in the TME, increasing the efficacy of immunotherapies^16–18. As nerve involvement within the TME is often an indication for adjuvant RT in oral cavity HNSCC¹⁹, targeting the neuronal cancer cross talk represents a possible new niche to improve immune response to RT. In the current study, we sought to determine if tumor growth delay by denervation was mediated by immune modulation, if denervation could synergize with RT, and if using pharmacological therapies targeting sensory nerve function can be repurposed as a treatment option to enhance therapeutic response to RT in HNSCC.

Results

Denervation reduces tumor growth

To determine if resistance to therapy in HNSCC may be driven by increased innervation of the TME we looked at gene signatures for axonogenesis, a known correlator with tumor growth²⁰, in responders and nonresponders to neoadjuvant radiation and immunotherapy combination in a recently published phase I/Ib clinical trial in HPV-unrelated HNSCC²¹. Using RNA sequencing of patients’ primary tumors pre-treatment, we found a significant increase in axonogenesis-related genes in non-responders (Figure 1A, Supplemental Table 1). The top 25 pathways driving the axonogenesis signature are shown in Figure 1B. The majority of these genes are associated with nerve function (GABRB1, MAPT, MAP2, SCN1B, NDNF, CDK5) and neurogenesis, sprouting or differentiation (EPHB3, NTRK2, EPHA4, NRP1, DISC1, SEMA3A, NRP2, EPHB2, LHX6, EPHB6), highlighted in Figure 1C and Supplemental Table 2.

Figure 1: Facial nerve axotomy reduces tumors growth.

A) Enrichment map displaying all responder versus nonresponder GO biological processes pathways calculated with gene set enrichment and showing NES < 0 and q < 0.05. Lines represent similarity in pathway gene membership. Select pathways chosen by manual inspection of pathways and groups in the map. B) Tree plot for top 25 pathways by q-value for select pathways chosen from panel A. Words in right column represent a 5-word word cloud of common themes in pathway titles. C) Heatmap for top 250 genes by responder vs. nonresponder p-value included in the leading edges of the select pathways in panel A. Select genes are manually annotated for commentary.NR=non-responder (n=5), R=Responder (n=8). D) Representative immunofluorescent images of nerve bundles present in both LY2 and MOC2 tumors (20x). E) BALB/c mice were implanted with 1×10^6 LY2 cells after a facial nerve axotomy (n=5) or sham surgery (n=4). F) C57BL/6 mice were implanted with 1×10^5 MOC2 cells after a facial nerve axotomy or sham surgery (n=6 facial nerve axotomy and n=5 sham surgery). G) Survival of C57BL/6 mice treated with or without a facial nerve axotomy (n=6) or sham surgery (n=5). To compare tumor growth differences a two-way analysis of variance (ANOVA) was used. Time-to-death by tumor-related symptoms was plotted using Kaplan-Meier (KM) curves and the survival difference between groups was compared using log-rank (Mantel-Cox) tests. The significance is denoted by asterisks, *p<0.05, **p<0.01, and ***p<.001. All data are reported with mean ± SEM (standard error of the mean).

Although RNA sequencing is an indirect measure of nerve abundance and functionality, we hypothesized that nerves within the TME were negatively regulating the TME to be pro-tumorigenic. Our aim was to investigate the impact of denervation on the growth of tumors in two syngeneic and immunocompetent orthotopic murine models of HNSCC. We confirmed that there was indeed nerve infiltration with PGP9.5 staining in both of our orthotopic mouse models of HNSCC (Figure 1D). Nerve bundles were present within the TME in both models, highlighted by the white arrows (Figure 1D). To study how nerves, specifically sensory nerves, may be influencing the tumor growth we performed either a previously developed facial nerve axotomy surgery²², as the facial nerve does not contain any adrenergic fibers, or a sham surgery before tumor implantation. After two weeks of recovery from the surgery, mice were implanted with LY2 or MOC2 tumor cells (Supplemental Figure 1A). Success of the facial nerve axotomy was determined by lack of whisker movement post-surgery as described previously²². We observed that both BALB/c and C57BL/6 mice that received facial nerve axotomy surgery had greater tumor control than the mice that received the sham surgery (Figure 1E, F and Supplemental Figure 1B, C). This reduction in tumor growth also conferred an increase in survival (Figure 1G). These data corroborate previous findings that denervation can lead to tumor growth reduction^8,10,11.

Denervation increases T cell activation in the TME

Recently, it has been reported that nerves, specifically sensory nerves, within the TME can directly modulate the adaptive immune response in melanoma¹⁰. Previous work in gastric cancer model similarly showed an increase in activated T cells in the TME after vagal nerve denervation⁸. Others suggest that in HNSCC nerves influence tumor growth by directly acting on cancer cells, not immune cells¹¹. To determine if the adaptive immune system was involved in mediating the differences in tumor growth observed in our models after denervation, we used a RAG −/− mouse model where there are no mature B cells or T cells. Using this RAG−/− model, we observed no differences in tumor growth between mice that received a facial nerve axotomy or a sham surgery (Figure 2A). This suggests that the effect of denervation on tumor growth is mediated by the adaptive immune system.

Figure 2: Denervation increases T cell activation in the TME.

A) RAG−/− mice that were treated with or without a facial nerve axotomy were implanted with 1×10^5 MOC2-OVA tumor cells in the buccal. Tumor growth is shown (n=7 for both groups). B) Populations of CD4 T cells (CD3+CD4+), CD8 T cells (CD3+CD8+), and C) Tregs (CD3+CD4+Foxp3+) cells within the DLNs of mice treated with or without a facial nerve axotomy (n=5 per group). D-E) CD4 and CD8 T cells in the TME expressing activation markers in mice treated with or without a facial nerve axotomy (n=5). Differences in T cell populations was determined using an unpaired student’s t-test. The significance is denoted by asterisks, *p<0.05, **p<0.01, and ***p<.001. All data are reported with mean ± SEM (standard error of the mean).

In our immunocompetent orthotopic model, we used flow cytometry to determine the differences in the TME after denervation. Specifically, we examined the activation status of T cells. As most of the T cell priming occurs in the draining lymph nodes (DLNs)^23,24, we asked if denervation impacted T cell priming in the DLNs. We found differences in the DLNs including an overall increase in CD8 T cells and CD4 T cells (Figure 2B), which coincided with a decrease in the immunosuppressive cell type regulatory T cells or Tregs (Figure 2C). Gating strategies for the DLNs are provided in Supplemental Figure 2A. Major differences, however, were observed in the TME. A significant increase was noted in activated CD4 T cells with a Th1 phenotype defined by Tbet, IL-2, and IFNγ expression (Figure 2D). There was also a significant increase in PD-1 expression on CD4 T cells within the TME (Figure 2E). Although PD-1 expression is often associated with an increase in exhaustion, it is also a sign of T cell activation as PD-1 is expressed after antigen-mediated activation²⁵. In addition to an increase in a CD4 Th1 phenotype, there was also an increase in activated CD8 T cells defined by IFNγ expression (Figure 2F). Gating strategies for the TME are provided in Supplemental Figure 2B. Collectively, these data suggest that denervation primarily confers a local effect on the TME and that intratumoral nerves are suppressing T cell activation.

Radiation improves response to denervation

As radiation therapy (RT) is used to treat over 50% of cancer patients²⁶ and PNI is an indication for adjuvant RT¹⁹, we sought to understand if denervation improves responses to RT. We have previously demonstrated that RT increases immune cell infiltration into the TME and improves response when combined with immunotherapies^16,17,27. To test if denervation would be an effective treatment when combined with RT, we combined our facial nerve axotomy surgery with one dose of 10 Gy, a dose which has been effective in the past at increasing immune cell infiltration^16,23, to test if this would improve outcomes. RT alone had a similar survival benefit as denervation alone, but when combined, we observed the greatest tumor growth reduction and survival benefit (Figure 3A–B). To explore the immunological changes that RT exerts on the TME in the setting of denervation, we conducted flow cytometry to examine changes in immune cell populations and T cell activation status. Mice treated with both facial nerve axotomy surgery and RT had the lowest tumor weight on day 12 post-implantation, the same day that tissues were collected for flow cytometry (Supplemental Figure 3A). Similar to what we observed without RT (Figure 2D), CD4 T cells’ expression of IFNγ was increased in mice receiving the facial nerve axotomy surgery (Figure 3C). Although we only observed a trend towards an increase in IL-2 expression in mice treated with RT and facial nerve axotomy surgery (Figure 3D), there was an increase in Tbet-expressing CD4 T cells (Figure 3E), suggesting an increase in Th1 CD4 T cell phenotype within the TME after denervation. Again, we observed an increase in PD-1 expression on CD4 T cells, further corroborating the fact that CD4 T cell activation is increased after denervation (Figure 3F). Finally, we observed an increase in activated CD8 T cells by IFNγ expression (Figure 3G). We also confirmed these findings in our MOC2 model (Supplemental Figure 3B–C). Together, these results suggest that when combined with RT, denervation enhances RT’s effect on effector T cell activation.

Figure 3: Radiation improves response to denervation.

A) C57BL/6 mice were implanted with 1×10^5 MOC2-OVA cells after a facial nerve axotomy or sham surgery. Mice were subsequently treated with RT (1×10Gy) when the tumors reached ~300mm³. Tumor growth over time is shown. Survival data is shown for the four groups of mice (n=7 per group). B) Time-to-death by tumor-related symptoms was plotted using Kaplan-Meier (KM) curves and the survival difference between groups was compared using log-rank (Mantel-Cox) tests. C-G) CD4 and CD8 T cells in the TME expressing activation markers in mice treated with or without a facial nerve axotomy (n=5). Differences in T cell populations was determined by an unpaired student’s t-test. The significance is denoted by asterisks, *p<0.05, **p<0.01, and ***p<.001. All data are reported with mean ± SEM (standard error of the mean).

Intratumoral nerves increase pathways associated with neuropeptide synthesis

To better understand the functionality of intratumoral nerves, HNSCC patients’ nerves were harvested from within the TME (intratumoral or involved), within 1.5 cm of the TME (peritumoral or adjacent), and normal nerves. Patient demographic information is shown in Table 1. We observed noticeable differences in protein expression between the three groups of nerves, with the intratumoral nerves having overall lower levels of protein expression (Figure 4A). We found that normal and adjacent nerves clustered closer together than involved nerves using a principal component analysis (Figure 4B). Clustering the samples by gene expression revealed two clusters that distinguished involved nerves from normal and adjacent samples (Figure 4A). Performing GO pathway analysis on cluster 1, proteins elevated in involved nerves, we observed increases in many pathways (Supplemental Figure 4), but specifically in pathways involved in the production, breakdown, and recycling of neuropeptides, the production of growth factors, and the production of proteins via cytoplasmic translation (Figure 4C).

Table 1:

Patient Demographic Data

	Normal	Normal adjacent	Tumor involved
Age at Diagnosis, mean (range)	64 (35–86)	72 (57–86)	71 (61–78)
Sex, N (%)
Male	4 (57.1%)	4 (57.1%)	3 (75.0%)
Female	3 (42.9%)	3 (42.9%)	1 (25.0%)
Race, N (%)
White	4 (57.1%)	4 (57.1%)	3 (75.0%)
Hispanic	1 (14.3%)	1 (14.3%)	0 (0%)
Asian	2 (28.6%)	2 (28.6%)	1 (25.0%)
Smoking History, N (%)
Positive	0 (0%)	0 (0%)	0 (0%)
Negative	7 (100%)	7 (100%)	3 (75.0%)
Unknown	0%	0%	1 (25.0%)
Alcohol Use, N (%)
Positive	2 (28.6%)	4 (57.1%)	2 (50.0%)
Negative	5 (71.4%)	2 (28.6%)	1 (25.0%)
Unknown	0 (0%)	1 (14.3%)	1 (25.0%)
Primary Tumor Site, N (%)
Oral tongue	2 (28.6%)	0 (0%)	0 (0%)
Parotid	2 (28.6%)	2 (28.6%)	1 (25.0%)
Tonsil	1 (14.3%)	0 (0%)	0 (0%)
Mandible	1 (14.3%)	1 (14.3%)	0 (0%)
Buccal mucosa	1 (14.3%)	1 (14.3%)	0 (0%)
Scalp	1 (14.3%)	2 (28.6%)	2 (50.0%)
Larynx	0 (0%)	1 (14.3%)	1 (25.0%)
Primary Tumor Type, N (%)
Pleomorphic adenoma	1 (14.3%)	1 (14.3%)	0 (0%)
Squamous cell carcinoma	6 (85.7%)	6 (85.7%)	4 (100%)
T Stage, N (%)
T1	1 (14.3%)	0 (0%)	0 (0%)
T2	0 (0%)	0 (0%)	0 (0%)
T3	4 (57.1%)	4 (57.1%)	2 (50.0%)
T4a	22 (18.5%)	1 (14.3%)	1 (25.0%)
T4b	1 (14.3%)	0 (0%)	0 (0%)
Tx	1 (14.3%)	2 (28.6%)	1 (25.0%)
N Stage, N (%)
N0	4 (57.1%)	4 (57.1%)	2 (50.0%)
N1	1 (14.3%)	0 (0%)	0 (0%)
N2a	0 (0%)	0 (0%)	0 (0%)
N2b	0 (0%)	1 (14.3%)	1 (25.0%)
N2c	0 (0%)	0 (0%)	0 (0%)
N3a	0 (0%)	0 (0%)	0 (0%)
N3b	1 (14.3%)	1 (14.3%)	1 (25.0%)
Nx	1 (14.3%)	1 (14.3%)	0 (0%)

Figure 4: Intratumoral nerves increase pathways associated with neuropeptide synthesis.

A) Protein expression data of nerves that were harvested from within the TME (involved), within 1.5cm of the TME (adjacent), or normal (not near the tumor). B) Principal component analysis or PCA plot of involved, adjacent, and normal nerves from our proteomics analysis. C) GO pathway analysis of cluster 1 highlighting the top hits from pathways elevated in involved nerves when compared to both adjacent and normal nerves from human proteomics. D) Tumor growth curves and survival plot of C57BL/6 mice implanted with 1×10^5 MOC2 cells treated with botulinum toxin A and 1×10Gy (n=7–8 per group).

Although both sympathetic efferent and sensory nerve function within the TME can mediate tumor growth, based on our surgical denervation data and on published data^9,11,12,23, we hypothesized that sensory nerve function in HNSCC may be promoting tumor growth and resistance to therapy. We found that nerves (defined by NF-H) within the TME were expressing CGRP, and that CGRP was co-localizing with TRPV1, suggesting sensory nerve involvement in our murine tumor model (Supplemental Figure 5A–B). Similarly, nerves in the tongue have been found to be NF-H⁺CGRP ⁺TRPV1⁺²⁸. To test if sensory nerves are driving tumor growth in our mouse models, we used botulinum toxin A, known to effect the release of CGRP²⁹, to target sensory nerves surrounding the TME. While Botulinum toxin A does not directly regulate CGRP release, it blocks SNARE proteins which are essential for the dense vesicle core release containing neuropeptides. Botulinum toxin A treated mice had decreased tumor growth and when also treated with radiation therapy, tumor growth was further decreased in our syngeneic mouse model in vivo (Figure 4D). This supports our hypothesis that sensory nerves are promoting immunosuppression in the TME.

CGRP increases tumor growth by inhibiting CD8 T cell cytotoxicity

Recent evidence suggests that CGRP secretion by sensory nerves in the TME can directly inhibit CD8 T cells^10,30,31. We investigated whether sensory nerves were directly inhibiting CD8 T cells through CGRP release. Using an in vitro calcein release assay to measure cancer cell death by CD8 T cells, increasing concentrations of CGRP resulted in increased cancer cell survival (Figure 5A). Increased concentrations of CGRP without CD8 T cells present did not affect cancer cell survival or cause cancer cell death (Supplemental Figure 6A). To confirm that this was a result of inactivation of CD8 T cells, we used an ELISpot assay to show that CD8 T cells incubated with increasing concentrations of CGRP showed reduced activation through IFNγ expression (Figure 5B). CGRP has multiple receptors and BIBN4096 only blocks ramp1/calcrl. Since our in vitro data suggested that CGRP directly deactivated CD8 T cells, we utilized BIBN4096 in vivo as a CGRP antagonist. A single dose of BIBN4096 made little difference to tumor growth and survival (Supplemental Figure 6B and 6C). However, daily dosing was effective at reducing tumor growth (Figure 5C). These data suggest that CGRP is being continuously released into the TME and continual dosing will be necessary to mitigate neuronal release of CGRP into the TME.

Figure 5: CGRP increases tumor growth by inhibiting CD8 T cell cytotoxicity.

A) Dose escalation cancer cell death assay. MOC2-OVA cancer cells were co-incubated with OTI CD8 T cells and varying concentrations of CGRP (0nM, 10nM, 100nM, 1000nM, 10000nM). Cancer cell death was measured by release of fluorescent calcein into the supernatant (n=4 technical replicates). Representative experiment of three repetitions. B) ELISpot of CD8 T cell activation in the presence of varying concentrations of CGRP (0nM, 10nM, 100nM, 1000nM, 10000nM). Activation was measured by the amount of IFNγ spots detected (n=3 technical replicates). C) C57BL/6 mice implanted with 1×10^5 MOC2 cells were treated with 1×10Gy RT and CGRP receptor antagonist BIBN4096 (25pg) per intratumoral injection daily. Tumor growth and survival was monitored (n=7–8 per group. D) C57BL/6 mice and TRPV1 KO mice implanted with 1×10^5 MOC2 cells were treated with 1×10Gy RT. Tumor growth and survival were monitored (n=7–8 per group). E) C57BL/6 mice and TRPV1 KO mice implanted with 5×10^4 P029 cells were treated with 3×8Gy RT. Tumor growth and survival were monitored (n=6–7 per group). F) Quantification of serum CGRP levels in TRPV1 KO mice or WT mice treated with RT after tumor implantation on DPI 17 (n=6 per group). G) C57BL/6 mice implanted with 1×10^5 MOC2 cells were treated with 1×10Gy RT and Gabapentin (2mg) via IP injection daily. Tumor growth and survival were monitored (n=7–8 per group). To compare tumor growth differences, a one-way analysis of variance (ANOVA) was used with Tukey’s post hoc test for multiple comparisons. Time-to-death by tumor-related symptoms was plotted using Kaplan-Meier (KM) curves and the survival difference between groups was compared using log-rank (Mantel-Cox) tests. To compare cancer cell death, CD8 T cell activation and differences in CGRP in the TME an unpaired student’s t-test was used. The significance is denoted by asterisks, *p<0.05, **p<0.01, and ***p<.001. All data are reported with mean ± SEM (standard error of the mean).

Knock-out or reduction of the cation channel TRPV1 reduces CGRP and tumor growth

Previous studies have shown that there is a high density of sensory nerves within the TME of HNSCC patients release CGRP³¹. The release of CGRP from sensory nerves is often stimulated by an influx of cations through the TRPV1 channel^10,32. Another possible route of reducing CGRP in the TME would be to prevent its initial release by sensory nerves. Using a global knock out (KO) mouse model of TRPV1, we hypothesized that we could further reduce tumor growth by preventing CGRP release into the TME. The TRPV1 KO mice had a reduction in tumor growth which was further decreased when combined with RT (Figure 5D). A decrease in tumor growth was also associated with an increase in survival (Figure 5D). Using a third model of HNSCC, the recently developed P029 cell line²³, we again observed a reduction in tumor growth and an increase in survival of TRPV1 KO mice treated with RT (Figure 5E). To confirm that CGRP was indeed reduced in our TRPV1 KO mice, we collected serum from both TRPV1 KO mice and control mice and measured that the TRPV1 KO mice had a reduction in CGRP in the serum (Figure 5F).

As a global KO or TRPV1 embryonically might lead to compensatory changes, we used gabapentin to reduce, but not abolish, cation influx through TRPV1 channels after tumor implantation. Gabapentin is a commonly used drug given to cancer patients to help manage nerve pain and chemotherapy or radiation induced mucositis. It is postulated to act by reducing cation channels on the presynaptic cleft³³. Treating mice with both gabapentin and RT resulted in a marked reduction of tumor growth and an increase in survival (Figure 5G). Altogether these data support out hypothesis that CGRP release by sensory neurons within the TME results in increased tumor growth and a decrease in survival.

Blocking CGRP function in the TME increases activated Th1 CD4 T cells and CD8 T cells

To determine if blocking CGRP through genomic or pharmacological methods would have similar effects on the TME as denervation, we performed flow cytometry to look at CD4 and CD8 T cell activation. We observed similar rates of tumor growth reduction using combination RT with BIBN4096, Gabapentin, or TRPV1 KO mice compared to mice treated with RT alone (Figure 6A). Overall, we observed similar trends in activation of Th1 CD4 T cells and CD8 T cells in the TME, concordant with our surgical denervation findings (Figure 6B–D). The gating strategy is showing in Supplemental Figure 7. Specifically, IFNγ, IL-2 and Tbet expression was increased by CD4 T cells in all groups targeting CGRP release or activity, supporting our previous findings that denervation increases Th1 CD4 T cells (Figure 6B). Also consistent with an activated T cell phenotype in the TME was the increased PD-1 expression on CD4 T cells in all groups (Figure 6C) and the increase in CD8 T cell activation as determined by IFNγ expression (Figure 6D). Collectively, and in concordance with our surgical denervation and in vitro data, these findings collectively implicate nerves in direct inhibition of CD8 T cell activation and cancer cell killing. In summary, we believe that our pharmacological inhibitors are decreasing CGRP release or activity resulting in an increase in immune cell activation Figure 6E.

Figure 6: Blocking CGRP function in the TME increases activated Th1 CD4 T cells and CD8 T cells.

A) Tumor growth curves for C57BL/6 mice or TRPV1 KO mice implanted with 1×10^5 MOC2 cells and treated with 1×10Gy. Mice were also treated with Gabapentin or the CGRP receptor antagonist BIBN4096 daily. B-C) CD4 T cells in the TME expressing activation markers (n=5 RT, n=7 Gabapentin, n=5 TRPV1 KO, n=7 BIBN4096). D) CD8 T cells in the TME and those expressing activation markers (n=5 RT, n=7 RT + Gabapentin, n=5 RT + TRPV1 KO, n=7 RT +BIBN4096). E) Cartoon of how different pharmacological inhibitors might inhibit CGRP release or activity upon release into the TME. Created using BioRender.com. Differences in T cell populations was determined by an unpaired student’s t-test. The significance is denoted by asterisks, *p<0.05, **p<0.01, and ***p<.001. All data are reported with mean ± SEM (standard error of the mean).

Discussion

As the addition of immunotherapies has lacked success in HPV-unrelated HNSCC, other forms of TME modulators are needed to improve outcomes. Perineural infiltration (PNI) and an increase in nerve density has long been associated with worse outcomes in HNSCC ^1,2,21,34. Although this correlation with outcome is widely known, it is not clear if intratumoral nerves are directly interacting with the immune cells within the TME. Previous research focused on intratumoral nerves in HNSCC have primarily focused on sympathetic efferent nerves and direct interactions with cancer cells¹¹. Contrarily, there has been an inclination that sensory nerves may play a role in promoting tumor growth through immune cell suppression as a vagal nerve axotomy increased CD44+ CD4 T cells in the TME in a gastric cancer model and a mouse CGRP KO model resulted in increased immune cell infiltration in HNSCC tumor model^8,31. More recently an elegant study was published showing that sensory nerve release of CGRP directly inhibited CD8 T cell activity leading to increased tumor growth in melanoma¹⁰.

Our data and previously published research support that nerves within the TME may be a poor prognostic factor in HNSCC^4,34. PNI has often been used in the clinical setting for evaluating prognosis and adjuvant treatment decisions¹⁹, but there are no clear standards for how to evaluate PNI in patient samples, grading often differs greatly between pathologists, and it is not always evaluted¹. Using more molecular approaches to evaluate not just presence of PNI, but also the functionality of the nerves within the TME might lead to greater consistency in evaluation and better biomarkers for treatment with BIBN4096 treatments or other treatments targeting sensory nerve immunosuppression.

In this study, we aimed to determine if sensory nerves were promoting tumor growth and whether this was through a direct interaction with adaptive immune cells in the TME. Our results, using two immunocompetent orthotopic models of HNSCC, support the hypothesis that sensory nerves can drive tumor growth and directly inhibit CD8 T cell activity through release of CGRP¹⁰. While recently shown in melanoma models¹⁰, to our knowledge, this is the first study to show in depth how sensory nerves are suppressing adaptive immune cells within the HNSCC TME using surgical, genetic, and pharmacologic methods in the context of RT. Both our in vivo/in vitro mouse data indicate that sensory nerves are likely promoting tumor growth, by directly influencing the adaptive immune system. This was expected as our surgical denervation model involved cutting the facial nerve which does not contain any sympathetic efferent fibers. It is likely that both sympathetic efferent and sensory nerves are playing significant roles in tumor growth and that differences in the TME and tumor type are driving the differences observed between models³⁵. Our study further corroborates that CGRP may be a biomarker for prognosis or a predictor of response to treatment and has been suggested as a biomarker of lymph node metastasis in HNSCC³⁶. Although we did not observe many differences in the lymph nodes between mice treated with sham or facial nerve axotomy surgery, limiting T cell activation may increase the metastatic potential of cancer cells within the TME.

Both CD8 and CD4 T cells are known to express the CGRP receptor, RAMP1^10,31. The direct effects of CGRP on CD8 T cells has now been reported in multiple tumor types including HNSCC^10,31 and our results corroborate these findings. Although other cell types including cancer cells can express RAMP1^37,38, our RAG−/− mouse model showed no differences between mice that received a sham surgery or a facial nerve axotomy surgery, suggesting that the majority of the effect of CGRP in the TME primarily acts on adaptive immune cells. One limitation of our study is that we did not look at how CGRP may be impacting B cell function. B cells are known to play an important role in anti-tumor immunity in HPV-positive HNSCC, but the role of B cells in HPV-unrelated HNSCC is less clear. However, in melanoma, the benefit of decreasing CGRP in the TME went away when CD8 T cells were depleted, suggesting sensory nerve modulation is primarily acting through a T cell-mediated immune response¹⁰.

When interpreting the human proteomics results from involved, adjacent, and normal nerves we observed increases in pathways associated with the production and recycling of neuropeptides and larger macromolecules. Increases in the peptide biosynthetic process could suggest increases in the production of neuropeptides, like CGRP, which act as neurotransmitters³⁹. Similarly, increases in the cellular macromolecule biosynthetic process could suggest that increased production of proteins and macromolecules essential for nerve growth and development like nerve growth factor (NGF) and brain-derived neurotrophic factor (BDNF), which promote nerve growth and survival^40,41. Additionally, increases in cytoplasmic translation are indicative that these nerves are increasing protein production from mRNA, which would also support increases in nerve growth and function along with the production of proteins involved in neurotransmission like ion channels and receptors⁴².

Targeting sensory nerves surrounding and within the TME is translationally quite relevant as drugs already exist that are FDA-approved for other uses and that have minimal side effects and toxicity profiles compared to many medications prescribed for cancer treatment. We found that gabapentin, a drug often used in the treatment, with debated efficacy^43,44, of nerve pain and RT or chemotherapy related mucositis, reduced tumor growth when combined with RT. This was surprising as others have studied whether gabapentin affects tumor growth and have had mixed results ^45,46. When combining treatment modalities, toxicity can have a significant impact on what treatments can be combined. Combining multiple treatments that target the same or similar pathways can result in adverse effects that are not well tolerated as has been observed with combining multiple immunotherapies. As we observed that the effects of surgical, genetic, or pharmacological depletion of CGRP were improved when combined with RT, we expect that denervation, through pharmacological means, will improve the effects of immune checkpoint inhibition as we observed an increase in PD-1 expression on T cells and have limited side effects. We anticipate that with the recent surge in research on the role of nerves within the TME that there is potential for translating these therapies to patients in a timely manner.

Limitations of Study

A limitation of this study is our use of a global TRPV1 KO model. Although we are not aware of any reports of a global TRPV1 KO altering development of neurons, besides the absence of TRPV1, or the immune system, it does change the mouse in important ways. TRPV1 KO mice have been reported to respond less to pain, heat, acidification of their environment, and they live longer than WT mice with their metabolism appearing younger and with improved spacial memory⁴⁷. Additionally, TRPV1 expression is not limited to neurons and is also known to be expressed on immune cells⁴⁸. As a result, a global TRPV1 KO could be directly affecting neuronal function and/or immune function. For these reasons, we tested whether the release of CGRP into the TME impacts tumor growth and immune cell activation in a variety of ways, with TRPV1 KO being only one of the methods we used. All pharmacological methods have drawbacks, but by using a CGRP receptor antagonist, gabapentin, botulinum toxin A, and surgical denervation in vivo and in vitro co-culture assays we believe that we have shown that sensory neuron release of CGRP is inhibiting CD8 T cells directly and leads to an increase in tumor growth. It is still unclear what in the TME might be activating TRPV1 as it is classically activated by capsaicin, high temperatures, or LPS, but it is increased in multiple cancer types and appears to play a role in inflammation independent of stimulation with these classical ligands suggesting other ligands may be playing a role in the TME⁴⁸. One possible mechanism by which TRPV1 is activated in the TME is through a decrease in the pH⁴⁹, but more research needs to be conducted to determine what factors activate TRPV1 in the TME.

STAR Methods

Resource Availability

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Dr. Sana Karam (sana.karam@cuanschutz.edu).

Materials Availability

Requests for the MOC2-OVA cell line should be directed to and will be fulfilled by the lead contact, Dr. Sana Karam (sana.karam@cuanschutz.edu).

Data and Code Availability

The RNA sequencing used in this study can be accessed online though the GEO accession number GSE210287. The proteomics data is available as a supplemental file. This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental Model and Study Participant Details

Human Nerve Collection

Informed consent was obtained for the collection and research use of human nerve tissues according to the guidelines of the World Health Organization and was approved by the university Institutional Review Board at the University of California of Los Angeles (IRB# 11–002858). Human nerves that were to be removed due to clinical indication standard of care were collected during head and neck cancer resection surgery, and included branches of the facial nerve, the spinal accessory nerve, and the greater auricular nerve. Collections comprised of nerves adjacent to tumor and tumor involved nerves. Within 60 minutes of acquisition, nerve tissues were cryopreserved using CryoStor^® CS10 medium at −80°C in order to maximize post-thaw cell recovery and viability prior to proteomics analysis.

Cell Lines and Cell Culture

Murine MOC2, LY2, and P029 HNSCC cell lines were used for in vivo studies. The MOC2 cell line was isolated from a C57BL6 mouse that developed squamous cell carcinoma in the oral cavity after exposure DMBA for 25 weeks. We are thankful to have obtained the MOC2 cell line from Ravindra Uppaluri from the Dana Farber Cancer Institute. The LY2 cell line was isolated from the draining lymph node of a BALB/c mice that had been inoculated with the PAM 212 squamous cell carcinoma cell line. We are thankful to have obtained this cell line from the lab of Nadarajah Vigneswaran at the University of Texas Health Science Center. P029 cell line was provided in collaboration with XJ Wang at University of Colorado Anschutz Medical Campus, Department of Pathology. P029 cell line was generated as described previously²³. All cell lines were cultured at 37°C and 5% CO₂. LY2 and P029 cells were cultured in DMEM-F12 media supplemented with 10% fetal bovine serum and 2% Primocin and 1% Fungin (InvivoGen, San Diego CA). MOC2 cells were cultured in DMEM-F12:IMDM (2:1) supplemented with 2% Primocin and 1% Fungin (InvivoGen, San Diego CA), and 1.75μg EGF and 20μg hydrocortisone per 500mL of media, and 0.1% insulin (Sigma Aldrich, St. Louis, Missouri, USA).

To generate MOC2-OVA cells: Five hundred thousand HEK293-FT cells were transfected with 2ug of pLVX-puro-cOVA-IRES-BFP (Addgene plasmid #135074) and 2ug of packaging viral mix (1:2 ratio of psPAX2 (Addgene plasmid #12260) and (Addgene plasmid #12259)) to generate lentiviral (LV) particles in a well of a six-well plate. Two ml of LV was collected 3 days post-transfection. 1mL LV was used to transduce 500,000 MOC2 cells. Media was changed ~24 hours post transduction. Transduced cells were selected with puromycin (1ug/mL) for 5–10 days.

Animal Tumor Models

BALB/c mice were purchased from Charles River, Wilmington, MA, C57BL6 and TRPV1 KO mice were purchased from the Jax Labs, Bar Harbor, Maine, USA. In vivo orthotopic HNSCC tumor models were established as previously described. Briefly, 1×10^6 (LY2), 1×10^5 (MOC2), or 5×10^4 (P029) per 50μl of non-FBS containing media of either LY2/P029 or MOC2 cells, respectively was prepared. A 1:1 mixture of cells and Matrigel (10mg/mL, BD Biosciences, San Jose, California, USA) at a volume of 100μl was injected into the right buccal mucosa of the mice. Radiation treatment was given when tumors reached ~100–200mm³. Tumor size was measured using digital calipers and tumor volume was estimated using the equation V=A×B²/2, where A is the longest dimension of the tumor and B is the shortest. Mice were euthanized when the mice reached study end qualifications approved by the Institutional Animal Care and Use Committee (IACUC). All animal protocols used in this study were approved by IACUC of the University of Colorado, Denver.

Methods Details

Irradiation

Before treatment with radiation, the mice were anesthetized using isoflurane. The mice were then placed in an X-RAD image guided small animal irradiator (Precision X-Ray, Bradford CT). The irradiation field was determined using fluoroscopy for each mouse individually. After the field was established, the mice were irradiated with 225kVp/20mA with a Copper filter, which corresponds to 5.6 Gy/min for 109 seconds.

Flow Cytometry

Both Tumors and draining lymph nodes (DLNs) were harvested from mice and immediately put on ice in Hank’s Balanced Salt Solution (HBSS). To facilitate single cell suspension of lymphocytes from tumors, the tumors were minced and then incubated for 30 minutes at 37°C with 200U of Collagenase III (Worthington, Lakewood, New Jersey, USA). Both tumors and DLNs were then filtered through 70μm nylon filters. The tumors were then spun down at 400g for 5 minutes and resuspended in 3 mL of Red Blood Cell Lysis Buffer for 3 minutes. After 3 minutes, 6 mL of HBSS was added and all the samples were spun down again at 400g for 5 minutes. The samples were aspirated and then were plated in 24 well plates in 1 mL of RPMI with .1% monensin to prevent cytokine release by the Golgi apparatus and .2% brefeldin A with cell stim cocktail of PMA and ionomycin to simulate cytokine production and transport in the cells. The cells were incubated in the stim media for 4 hours at 37°C. The plates were then spun down and the samples were resuspended in 200μl of Fc Block (anti-CD16/CD32 antibody, Tonbo Biosciences, San Diego, CA) for 20 minutes at room temperature. The samples were then plated into a 96 well plate and spun down at 400g for 5 minutes. The samples were then resuspended in 100μl of PBS containing .5μl of live/dead aqua viability stain (Invitrogen, Carlsbad, California). The samples were incubated away from light for 15 minutes at room temperature. The samples were then spun down and resuspended in a mixture of the extracellular antibodies in brilliant stain buffer (BD Biosciences). The cells were incubated for 30 minutes at room temperature. After incubated the cells were washed twice with HBSS. The cells were then incubated overnight with the Foxp3 transcription factor staining kit (eBioscience). The next day, the cells were spun down and washed twice as per the kit instructions. The cells were then stained with the intracellular antibodies in brilliant stain buffer for 30 minutes at room temperature. The cells were then washed twice with PBS, re-suspended, and then run on an Aurora Spectral Flow Cytometer (Cytek Biosciences) at the Barbara Davis Center Diabetes Research Center Cell and Tissue Analysis Core.

The following antibodies were used in this study: CD45-PerCP (Clone: 30-F11 BioLegend), CD3-BUV805 (Clone: 17A2 BD Biosciences), CD4-BUV496 (Clone: GK1.5 BD Biosciences), CD8-BB515 (Clone: 53–6.7 BD Biosciences), Tbet-BV421 (Clone: 4B10 BioLegend), IL-2-Alexa Flour 700 (Clone: JES6–5H4), IFNγ-BUV737 (Clone: XMG1.2 BD Biosciences), Foxp3-Alexa Flour 532 (Clone: FJK-16s ThermoFischer), and CD279 (PD-1)-BUV615 (Clone: J43 BD Biosciences). FlowJo (10.8.1) analysis software was used for data analysis.

RNA Processing and Sequencing

For the human RNA sequencing library preparation, a TempO-Seq Human Full Transcriptome FFPE Assay 96 sample Kit was used (BioSpyder, Carlsbad, CA). FFPE samples from HPV-unrelated HNSCC tumors pre- and post-radiation therapy was obtained from the University of Colorado biorepository. Patients had histologically or cytologically confirmed nonmetastatic stage II-IVB oral cavity, stage III-IVB larynx, stage III-IVB hypopharynx, or stage III-IVB HPV and/or p16 negative intermediate-high risk oropharynx head and neck cancer (AJCC 8th edition). Patients were treated with 6 Gy × 2, 6 Gy × 3, or 8 Gy × 3 within a two-week period. H&E stains of the tumor samples were reviewed by a pathologist and areas of tumor cellularity were identified and marked. Only the areas of tumor cellularity that were marked by the pathologist were scraped and processed per BioSpyder kit instructions. Samples were pooled and run in two sequencing lanes using a NextSeq high throughput sequencing instrument at the Next Generation Sequencing Core at the University of Colorado Boulder. Reads were aligned and counts were generated using the BioSpyder TempoSeqr Platform. Genes with less than 1 mean raw count or less than 1 mean counts per million (CPM) were removed from the dataset.

RNA-Seq analysis of patient samples

RNA-Seq expression and differential gene expression for pre-treatment patient samples from our previous publication²¹ was subjected to gene set enrichment analysis, with only pathways downregulated in responders versus nonresponders (NES < 0) kept for downstream clustering of pathways. R (v.4.2.3) package clusterProfiler⁵⁰ (v.4.6.2) was used to perform gene set enrichment analysis with function GSEA() using ranks as the full list of log2 fold changes, minGSSize = 15, maxGSSize = 2000, and probing GO biological processes from MSigDB⁵¹. An enrichment map was created with pathways with q < 0.05 with enrichplot::emapplot() and otherwise default options. From this map, select pathways were extracted and the top 25 of these pathways by p-value were displayed with enrichplot::treeplot() with cluster.params n = 5. Among all pathways circled in Fig. 1A, the top 250 genes from the leading edges of these pathways by responder vs nonresponder p-value were displayed in a heatmap using R package ComplexHeatmap⁵² (v.2.14.0) and select genes annotated.

Human Nerve Processing for Proteomics

Proteomics was conducted on the nerves as previously described⁵³. Briefly, the nerves were digested using the FASP protocol with a 10kDa molecular weight cutoff. A timsTOF SCP instrument (Bruker Daltonics, Bremen) and an Evosep One system (Evosep Biosciences) were coupled to perform the assay using a ~58-minute gradient Whiper 20SPD method with a flow rate of 100nL/min. Fragpipe was used to convert raw files to peak lists in the MGF format, for downstream identification, validation, filtering, and quantification.

Proteomics Analysis

Intensity values were filtered with the following criteria. Proteins were removed if expression was constant within all groups, or if >50% of all samples had zero expression. In order to assess pathways, proteins were converted to gene symbol. Six duplicated gene symbols were removed by keeping the protein isoform with the highest average expression. Basic statistics (fold change, ANOVA, z-score normalization) were performed with Metaboanalyst 5.01⁵⁴ with the Stastical Analysis [one factor] module. A heatmap was created with the R package ComplexHeatmap2⁵⁵ (version 2.12.1) using expression values z-score scaled across each row for the top 400 proteins by ANOVA p-value and default clustering settings across rows: heirarchical clustering by Euclidean distance by R function dist and agglomeration method “complete” by R function hclust. K-means clustering was set within the Heatmap function by row_km = 3 and row_km_repeats = 100. Genes within specified k-means clusters were assessed by over-representation analysis using R package clusterProfiler3⁵⁰ (version 4.4.4) using the enrichGO function for GO biological pathways in size range 10–500 and otherwise default settings. Treeplots were generated with R package enrichplot4⁵⁰ (version 1.16.2) for the top 25 pathways by qvalue with setting nCluster = 5, and clusters manually labeled. R software5 version 4.2.1 was used for all proteomics R analyses.

Facial nerve axotomy surgery

Facial nerve axotomy surgery was conducted as previously described. Briefly, mice were anesthetized with isoflurane and a toe pinch was used to ensure that the mouse was under anesthesia. The area behind the right ear was shaved and a small 5mm incision was made horizontal to the blood vessel dividing the ear. Blunt dissection was used to find the white cartilaginous rings of the ear canal. Further blunt dissection was used to find the facial nerve as it ran across the bottom of the cartilaginous ear canal. Once the right facial nerve was identified, the right facial nerve was cut, and a small segment of the facial nerve was removed to ensure that the nerve could not heal together post-surgery. The structures of the dissection area were repositioned as best as possible before bacitracin was used to prevent infection. Surgical clips were used to seal the wound and the mice were given buprenorphine for pain management. When awoken from anesthesia, mice were examined for lack of vibrissae movement on the surgical side to confirm facial nerve axotomy. The full procedure can be viewed in the JOVE paper https://www.jove.com/v/52382/facial-nerve-axotomy-mice-model-to-study-motoneuron-response-to²². For sham treatment, the above procedures were similarly performed; however, the facial nerve was not dissected out, cut, or removed. The mice were given two weeks to recover from the surgery before tumor implantation. Mice were monitored for wound dehiscence and signs of excessive bleeding, hematoma, infection, or pain.

Botulinum Toxin A Studies

For the botulinum toxin A study, botulinum toxin A was injected at 6 days post-implantation. 80 μL of 1 unit Botox diluted in PBS was injected intratumorally at 4 injection sites. 80 μL of PBS was injected similarly for the sham treatment.

Gabapentin Studies

For the gabapentin study, mice were given 2mg of gabapentin daily. Gabapentin in powder form was dissolved in DPBS (100mg of gabapentin powder in 5mL of DPBS). Before injection, the gabapentin was heated at 37°C for complete dissolution in the DPBS. The mice were given gabapentin via 100μL IP injections. Control mice were given DPBS via 100μL IP injection.

CGRP Studies

For our first CGRP study, the CGRP receptor antagonist (BIBN4096, Tocris #4561) was given (25pg in 100ul, subcutaneously) once during the study before RT. We chose this dose as a previous publication that treated mice both IP and subQ with BIBN4096 found that this dose when given subQ was effective¹³. Sham treatment was DPBS injection subcutaneously. For our second CGRP study, the same treatment was given, but BIBN4096 was given every day after the initial injection. A sham injection of DPBS was also given every day.

CGRP ELISA

Blood was collected from mice at time of death via heart stick upon sacrifice and placed into a BD microtainer (BD Biosciences, San Jose CA). Following incubated for 30 minutes at room temperature, the blood was centrifuged at 6000 RPM for 2 minutes. Serum was collected and was stored at −80°C until used. The ELISA was conducted by the manufacturer’s instructions (mouse CGRP ELISA: EIAM-CGRP-1, RayBiotech).

Cancer Cell Death Assay

Spleens were harvested from OTI mice. Spleens were passed through a 70μm nylon strainer to create a single cell suspension. CD8 T cells were isolated using the STEMCELL Mouse EasyStep (#19853) magnet separation kit using the manufacturers protocol. MOC2-OVA cells were stained with calcein, AM (Invitrogen, C3099) by adding 1 × 10^6 MOC2-OVA cells per mL of 2μg/mL of Calcein in 10% FBS RPMI and then incubated at 37°C for 30 minutes. The MOC2-OVA cells were then washed twice to remove excess Calcein. CD8 T cells and MOC2-OVA cells were then added to a 96 well plate at a 2:1 ratio in 10% FBS RPMI. To activate the CD8 T cells anti-CD3 (Clone: 17A2, BioXCell BE0002) (3μg/mL) and anti-CD28 (Clone: 37.51, BioXCell BE0015–1) (5μg/mL) were added to the wells. CGRP (Pheonix Pharmaceuticals Inc., #015–09) was added at various concentrations (0nM, 10nM, 100nM, 1,000nM, and 10,000nM) to the wells. The plates were then spun down at 100g for 1 minute and then incubated at 37°C for 4 hours. After the incubation the plate was spun down at 400g for 5 minutes. The supernatant was then collected and calcein measured at excitation: 485nm and emission: 530nm. To calculate release the following formula was used: [(Test Release-Spontaneous Release)/(Maximum Release-Spontaneous Release)] × 100. To get maximum release add Triton x-100 at 2% concentration in the media with cancer cells alone.

CD8 T cell Activation ELISPOT Assay

Anterior cervical lymph nodes were harvested from OTI mice where the CD8 T cells are specific for the protein Ovalbumin. 3 × 10^5 cells from the lymph nodes were used per well in the ELISPOT. To activate the CD8 T cells ovalbumin peptide (OVA_257–264, InvivoGen) was added to the wells at 10μg/mL. CGRP (Pheonix Pharmaceuticals Inc., #015–09) was added at various concentrations (0nM, 10nM, 100nM, 1,000nM, and 10,000nM) to the wells. The kit instructions were followed for the assay (mouse IFNγ ImmunoSpot, Cleveland, USA).

Immunofluorescence

Immunofluorescence staining (IF) on formalin-fixed paraffin-embedded (FFPE) 4-μm tumor sections was performed to visualize CGRP expression and nerve type in the tumor microenvironment. Slides were deparaffinized using two 2-minute washes in histological xylenes (Thermo Fisher Scientific, REF#X3P-1GAL, LOT#215369). The slides were sequentially rehydrated twice for 5 minutes each in 100%, 90%, 80%, and 70% ethanol in deionized water, followed by two additional 5-minute washes in deionized water. Heat-induced antigen retrieval was performed using a 1% (v/v) citric acid-based antigen unmasking solution (Vector Laboratories, REF#H-3300, LOT#ZJ0516) diluted in deionized water in an electric pressure cooker (Cuisinart, MODEL#CPC-600) on the “high pressure” setting for 15 minutes. Following two 5-minute washes in 1X Dulbecco’s phosphate buffered saline (Gibco, REF#14190–144), the slides were permeabilized twice in PBS containing 0.025% (v/v) TWEEN^® 20 (Sigma-Aldrich, LOT#SLCB6677) for 10-minutes each. The following blocking, primary antibody, and secondary antibody incubations were performed in a dark, humidified chamber. The slides were blocked in 1% (w/v) bovine albumin serum fraction V (Sigma-Aldrich, REF#03116956001, LOT#10042752) diluted in 0.025% PBS-T for 1 hour at room temperature. After tapping off the blocking buffer, the slides were incubated in primary antibodies for 2 hours at room temperature. Combinations of anti-CGRP (1:500, Thermo Fisher Scientific, REF#PA5–114929, LOT#WI3370826A), anti-NF-H (1:500, Thermo Fisher Scientific, REF#PA1–10002, LOT#YG3998631), anti-TRPV1 (1:200, Alomone Labs, REF#ACC-030–200UL, LOT#YH4006981), and anti-ChAT (1:100, Thermo Fisher Scientific, REF#MA5–31383, LOT#YD3882051) were diluted in 1% BSA blocking buffer. After two 5-minute washes of 0.025% PBS-T, the slides were sequentially incubated in secondary antibodies in the dark for 1 hour at room temperature with two 5-minute washes of 0.025% PBS-T between each incubation. The second secondary incubation included combinations of either goat anti-mouse IgG (H+L) highly cross-adsorbed Alexa Fluor^™ Plus 488 (1:1000, Thermo Fisher Scientific, REF#A32723, LOT#VH309036), goat anti-rabbit IgG (H+L) cross-adsorbed Alexa Fluor^™ Plus 647 (1:1000, Thermo Fisher Scientific, REF#A21244, LOT#1741783), and or goat anti-chicken IgY (H+L) Alexa Fluor^™ 647 (1:300, Thermo Fisher Scientific, REF#A21449, LOT#2480082) diluted in 1% BSA blocking buffer. After two 5-minute washes of 0.025% PBS-T, the slides were mounted with DAPI Fluoromount-G^® (SouthernBiotech, REF#0100–20, LOT#L0116-Z056), coverslipped, and sealed with clear gel nail polish. After storing them overnight in the dark at 4°C, the slides were imaged using an inverted fluorescence phase contrast microscope (KEYENCE, MODEL#BZ-X810) at 20X magnification with automatic black balance settings and level correction.

Statistical Analysis

When comparing two groups for tumor growth curves a two-way ANOVA was used. For survival studies, log-rank Mantel-Cox test was used. When comparing two groups, unpaired t-tests were used unless otherwise stated. All statistical analysis was done in Prism Software (v9.1.0).

Supplementary Material

1

Supplemental Table 1: Genes in Figure 1A, related to Figure 1.

2

Supplemental Table 2: Genes in Figure 1C, related to Figure 1.

3

Supplemental Table 3: Data used as input for Proteomics analysis, related to Star Methods.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rat Anti-mouse CD45-PercP (30-F11)	BioLegend	Cat# 103130; RRID AB_893339
Rat Anti-mouse CD3-BUV805 (17A2)	BD Biosciences	Cat# 569192; RRID AB_2871285
Rat Anti-mouse CD4-BUV496 (GK1.5)	BD Biosciences	Cat# 612952; RRID AB_2813886
Rat Anti-mouse CD8-BB515 (53–6.7)	BD Biosciences	Cat# 564422; RRID AB_2738801
Mouse Anti-Mouse Tbet-BUV421 (4B10)	BioLegend	Cat# 644816; RRID AB_10959653
Rat Anti-mouse IL-2-AF700 (JES6-5H4)	BioLegend	Cat# 503818; RRID AB_528931
Rat Anti-mouse IFNγ-BUV737 (XMG1.2)	BD Biosciences	Cat# 612769; RRID AB_2870098
Rat Anti-mouse Foxp3-AF532 (FJK-16s)	ThermoFischer	Cat# 58-5773-82; RRID AB_11218870
Hamster Anti-mouse CD279(PD-1)-BUV615 (J43)	BD Biosciences	Cat# 752299; RRID AB_2875816
Purified anti-mouse CD16/CD32 FC Shield (clone 2.4G2)	Tonbo Biosciences	Cat#: 70-0161-M001, RRID: N/A
Rabbit Anti-mouse CGRP	Invitrogen	Cat# PA5-114929; RRID AB_2899565
Chicken Anti-mouse NF-H	Invitrogen	Cat# PA1-10002; RRID AB_1077155
Rabbit Anti-mouse TRPV1	Alomone Labs	Cat# ACC-030-200UL
Mouse Anti-mouse ChAT	Invitrogen	Cat# MA5-31383; RRID AB_2787020
Goat Anti-mouse IgG (H+L) highly cross-adsorbed AF Plus 488	Invitrogen	Cat# A-11055; RRID AB_2534102
Goat Anti-rabbit IgG (H+L) cross-adsorbed AF Plus 647	Invitrogen	Cat# A32733; RRID AB_2633282
Goat Anti-chicken IgY (H+L) AF 647	Invitrogen	Cat# A-21449; RRID AB_2535866
InVivoMab Anti-mouse CD3	BioXCell	Cat# BE0002
InVivoMab Anti-mouse CD28	BioXCell	Cat# BE0015-1
Biological samples
Human HNSCC – isolated nerves	University of California Los Angeles	Collected per COMIRB-11-002858
Human HNSCC – tumor tissue	University of Colorado	Clinical Trials.gov NCT03635164
Chemicals, peptides, and recombinant proteins
Fetal Bovine Serum	Gibco	Cat# 10437028
RPMI 160 Medium	Gibco	Cat# 11875-093
DMEM/F-12	Gibco	Cat# 10565042
Primocin	InvivoGen	Cat# ant-pm-05
Fungin	InvivoGen	Cat# ant-fn-1
BIBN4096	Tocris	Cat# 4561
pLVX-puro-cOVA-IRES-BFP plasmid	Addgene	Plasmid #135074
psPAX2 plasmid	Addgene	Plasmid #12260
pMD2.G plasmid	Addgene	Plasmid #12259
Puromycin	Invivogen	Cat# ant-pr-1
Matrigel	BD Biosciences	Cat# 05850
Brilliant Stain Buffer	BD Biosciences	Cat# 563794
Trypsin Inhibitor, Soybean Purified, AOF	Worthington Biochemical Corporation	Cat# LS003571
Collagenase Type III	Worthington Biochemical Corporation	Cat# LS004182
eBioscience 1X RBC Lysis Buffer	Invitrogen	Ref# 00-4333-57
eBioscience FOXP3/Transcription Factor Fixation/Permeabilization	Invitrogen	Ref# 00-5521-00
Live/Dead Aqua Fixable Viability Dye	Invitrogen	Cat#: L34966
Calcein, AM	Invitrogen	Cat#: C3099
DAPI Fluoromount-G®	SouthernBiotech	Cat# 0100-20
CryoStor CS10	STEMCELL	Cat# 07931
Xylene	Fischer Scientific	REF#X3P-1GAL
Vector Unmasking Solution	Vector Laboratories	REF#H-3300
TWEEN	Sigma-Aldrich	LOT#SLCB6677
Bovine Albumin Serum Fraction V	Sigma-Aldrich	REF#03116956001
Critical commercial assays
EasySep Mouse CD8+ T cell isolation kit	StemCell Technologies	Cat#: 19853
Mouse CGRP-I ELISA EIA Kit	RayBiotech	Code: EIAM-CGRP-1
CGRP	Phenix Pharmaceuticals	Cat# 015-09
Mouse IFNγ Single-Color ELISPOT	ImmunoSpot
Deposited data
Human RNA Sequencing Data	Darragh et al. https://doi.org/10.1038/s43018-022-00450-6	GEO: GSE210287
Human Nerve Proteomics Data	This paper	Supplemental File.
Experimental models: Cell lines
MOC2	Provided by R. Uppaluri	RRID: CVCL_ZD38
LY2	Provided by Nadarajah Vigneswaran	RRID: CVCL_Z594
P029	Provided by X.J. Wang	N/A
HEK293-FT cells	Provided by Molishree Joshi
Experimental models: Organisms/strains
Mouse: C57BL/6J	Jackson Laboratories	Strain Code: 000664
Mouse: Balb/C	Charles River	Strain Code: 028
Mouse: TRPV1 KO	Jackson Laboratories	Strian Code: 003770
Software and algorithms
Flowjo (v10.8.1)	BD Bioscience	https://www.flowjo.com
Grahpad Prism (v9.1.0)	Dotmatics	https://www.graphpad.com
Biorender	Biorender.com
R (v.4.2.3 and v. 4.2.1)	R Foundation for Statistical Computing
clusterProfiler (v.4.6.2 and 4.4.4)	Wu, T. et al. clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innovation (Camb) 2, 100141 (2021).
ComplexHeatmap (v.2.14.0 and v.2.12.1)	Gu, Z., Eils, R. & Schlesner, M. Complex heatmaps reveal patterns and correlations in multidimensional genomic data. Bioinformatics (Oxford, England) 32, 2847-2849 (2016).
Metaboanalyst 5.01		https://www.metaboanalyst.ca/

Highlights.

• Facial nerve denervation reduces tumor growth in orthotopic murine HNSCC

• Denervation reduces activation of CD4 and CD8 T cell in vivo

• Reduction in calcitonin gene-related peptide reduces tumor growth

• Calcitonin gene-related peptide decreased CD8 T cell activation in vitro\]]

Context and Significance.

Subsets of head and neck cancer still have high morbidity and mortality rates. Although the presence of nerves within the tumor has been known to be a poor prognosis marker for patients with this disease, how nerves impact tumor growth and our immune system is relatively unknown. Darragh et al. used genetically engineered mouse models, pharmacological inhibitors, and surgical models to test how nerves impact tumor growth. Darragh et al. show that sensory nerves release calcitonin gene-related peptide into the tumor and its surroundings and that this molecule inhibits the immune cells capable of killing cancer cells. This data suggests that targeting nerves in the tumor, with drugs already available, may reduce tumor growth.