Sensitization of Cutaneous Primary Afferents in Bone Cancer Revealed by In Vivo Calcium Imaging

Simple Summary Cancer-induced bone pain severely impairs the quality of life of cancer patients, many of whom suffer from inadequate pain relief. The development of new analgesic therapies depends on the identification of the cells and mechanisms involved in cancer-induced bone pain. Bone marrow innervating sensory neurons have been proposed to contribute to this debilitating disease, but their role remains unexplored. Here we used in vivo calcium imaging to determine the functional role of bone innervating and skin innervating neurons in contributing to pain at an advanced stage of bone cancer. Our results indicate increased excitability of skin innervating neurons, while those innervating bone are unaffected. Our data suggests skin-innervating neurons become hyperexcitable in cancer-induced bone pain and are a potential target for pain relief. Abstract Cancer-induced bone pain (CIBP) is a complex condition, comprising components of inflammatory and neuropathic processes, but changes in the physiological response profiles of bone-innervating and cutaneous afferents remain poorly understood. We used a combination of retrograde labelling and in vivo calcium imaging of bone marrow-innervating dorsal root ganglia (DRG) neurons to determine the contribution of these cells in the maintenance of CIBP. We found a majority of femoral bone afferent cell bodies in L3 dorsal root ganglia (DRG) that also express the sodium channel subtype Nav1.8—a marker of nociceptive neurons—and lack expression of parvalbumin—a marker for proprioceptive primary afferents. Surprisingly, the response properties of bone marrow afferents to both increased intraosseous pressure and acid were unchanged by the presence of cancer. On the other hand, we found increased excitability and polymodality of cutaneous afferents innervating the ipsilateral paw in cancer bearing animals, as well as a behavioural phenotype that suggests changes at the level of the DRG contribute to secondary hypersensitivity. This study demonstrates that cutaneous afferents at distant sites from the tumour bearing tissue contribute to mechanical hypersensitivity, highlighting these cells as targets for analgesia.

Cancers 2020, 12   To determine if bone afferents could be defined based on size within each subpopulation of sensory neurons, we compared the average sizes within each population. The mean size of the Na v 1.8+ DRG population is significantly lower than the mean size of all bone afferents and Na v 1.8-cre expressing bone afferents (Kolmogorov-Smirnov test: **** p < 0.0001, Figure S1A). Likewise, the mean size of the Tmem233+ population is smaller than the mean size of all retrogradely labelled bone afferents ( Figure S1B; Kolmogorov-Smirnov test: **** p < 0.0001) and Tmem233+ bone afferents (Kolmogorov-Smirnov test: ** p = 0.0019) respectively. Size distribution of Pvalb+ neurons and bone afferents did not significantly differ ( Figure S1C; Kolmogorov-Smirnov test: p = 0.0940).

GCaMP3-Expressing Mice Show a Moderate Pain Phenotype in a Model of CIBP
The endpoint of the study was defined as animals showing clear signs of limping (limb use score = 2 or 1), which is also when mice with CIBP show altered gene expression at the level of the DRG [33]. Cancer bearing mice started to limp between days 10 and 18 post-surgery, as outlined in the survival curve ( Figure 2A) (log-rank test, **** p < 0.0001). Static weight bearing on the affected limb was markedly reduced at the endpoint in cancer animals compared to both baseline and their sham counterparts ( Figure 2B) (two-way ANOVA with Bonferroni post-hoc, **** p < 0.0001). Secondary mechanical hypersensitivity was assessed with von Frey filaments. We found a significant difference in 50% withdrawal thresholds at the endpoint in cancer animals compared to baseline (two-way ANOVA with Bonferroni post-hoc, *** p = 0.0002) and to sham animals (two-way ANOVA with Bonferroni post-hoc, ** p = 0.0092) ( Figure 2C). Primary hypersensitivity was assessed using non-noxious palpation of the distal femur head, which produced a marked increase in nocifensive responses (including guarding, flinching, and licking) in cancer animals compared to baseline (two-way ANOVA with Bonferroni post-hoc, * p = 0.0378) and compared to sham animals at the endpoint (two-way ANOVA with Bonferroni post-hoc, * p = 0.0159) ( Figure 2D). On the other hand, pain pressure thresholds of the paw ipsilateral to the affected femur were comparable in sham and cancer bearing animals at baseline and endpoint ( Figure 2E). Similarly, thermal pain thresholds of cancer animals measured using the hot-plate test at 50 • C also indicated no differences between groups ( Figure 2F).
way ANOVA with Bonferroni post-hoc, * p = 0.0378) and compared to sham animals at the endpoint (two-way ANOVA with Bonferroni post-hoc, * p = 0.0159) ( Figure 2D). On the other hand, pain pressure thresholds of the paw ipsilateral to the affected femur were comparable in sham and cancer bearing animals at baseline and endpoint ( Figure 2E). Similarly, thermal pain thresholds of cancer animals measured using the hot-plate test at 50 °C also indicated no differences between groups ( Figure 2F). (B) Ongoing pain was measured by percentage weight bearing on the affected limb. (C) Mechanical withdrawal threshold to von Frey filaments in cancer (n = 12) and sham animals (n = 21). (D) Number of nocifensive responses (guarding, licking, flinching) observed during the 2 min period after palpation of the distal femur head in cancer bearing (n = 5) and sham (n = 11) mice. (E) Mechanical withdrawal thresholds to application of the Randall-Selitto apparatus to the paw in cancer (n = 6) and sham animals (n = 7). (F). Thermosensation measured by response threshold to 50 • C hot-plate in cancer (n = 6) and sham mice (n = 4). * p < 0.05; ** p < 0.01; *** p < 0.001.

Properties of Cutaneous Afferents in CIBP Animals
To investigate the role of cutaneous afferents in secondary hyperalgesia of the hindpaw, a number of different stimuli were applied to the glabrous skin of the plantar surface for 3 s at 30 s intervals during in vivo calcium imaging. First, tweezers were used to apply pressure across the dermatome covering L3-L4, followed by hot water at 55 • C, and ice-cold water. Cancer bearing animals presented a significantly larger proportion of neurons sensitive to pinch compared to sham animals, but a lower proportion of heat sensitive neurons ( Figure 3A) (two-way ANOVA with Bonferroni post-hoc, **** p < 0.0001) (Videos S1 and S2). This shift may depend on increased polymodality in response to tissue injury or inflammation [34]. Polymodality was significantly increased in cancer bearing animals from 2.56% to 5.56% ( Figure 3B) (Welch's t-test, * p = 0.0463). We also investigated whether cross-activation of sensory neurons contributes to the observed shift by determining the coupling response of DRG neurons [35]. The percentage of coupled responses within all mechanosensitive neurons, although higher in cancer bearing animals, did not significantly differ from sham mice (6.59 ± 1.73% vs. 2.42 ± 1.33% respectively; Welch's t-test, p = 0.0717). On the other hand, cross-activation of heat sensitive neurons was observed in very few cases both in sham and cancer animals ( Figure 3C) (3.54 ± 1.27% and 1.60 ± 0.92%, respectively; Welch's t-test, p = 0.1909). Changes in fluorescence of GCaMP-expressing DRG neurons during stimulus application (compared to baseline unevoked fluorescence) serves as a surrogate for the strength of the calcium transient. We found that the maximum response intensity of both mechanosensitive ( Figure 3D) (1.32 ± 0.07 in sham vs. 1.62 ± 0.08 in cancer, Welch's t-test, ** p = 0.0077) and heat sensitive ( Figure 3E) (1.31 ± 0.04 in sham vs. 1.62 ± 0.07 in cancer, Welch's t-test, *** p = 0.0005) L3 cutaneous afferents were increased in cancer bearing compared to sham animals. To determine if previously silent nociceptors are recruited in response to noxious pinch, we compared size distributions. A total of 170 mechanosensitive neurons in sham animals and 147 mechanosensitive neurons in cancer animals showed the same size distribution ( Figure 3F) (Kolmogorov-Smirnov test, p = 0.8275). Likewise, 366 heat sensitive neurons in sham and 147 in cancer animals did not show any difference in size distribution ( Figure 3G) (Kolmogorov-Smirnov test, p = 0.6651).   coupled responses over total responses of the same stimulus modality for pinch and heat in sham and cancer bearing mice. (D,E) Response intensity of mechanosensitive (D) and heat sensitive (E) cutaneous afferents expressed as maximum fluorescence intensity during stimulus application vs. baseline. (F,G) Size distribution of mechanosensitive (n = 147 neurons in cancer, n = 170 neurons in sham) (F) and heat sensitive (n = 147 in cancer, n = 366 in sham) cutaneous afferents (G). * p < 0.05; ** p < 0.01; *** p < 0.001, **** p < 0.0001

Activation of Femoral Bone Marrow Afferents
To investigate the contribution of intraosseous pressure and the acidic microenvironment in CIBP, we used intrafemoral injections of saline and citric acid, respectively, during in vivo DRG imaging of sham and cancer animals. In a subset of mice, we used a pump-controlled syringe connected to a pressure gauge to deliver 10 µL saline into the femoral marrow of sham and cancer bearing animals to serve as an indirect measure of intraosseous pressure ( Figure 4). This initial pressure prior to injection of saline did not significantly differ between cancer and sham animals (125.60 ± 18.96 and 86.67 ± 13.33 mmHg respectively, Welch's t-test, p = 0.1440) ( Figure 5A). The time required for delivering 10 µL saline in cancer animals (52.2 ± 4.41 s) was nearly threefold of that observed in sham animals (12.67 ± 6.22 s) (Welch's t-test, ** p = 0.0020) ( Figure 5B). To visualize bone marrow afferents during in vivo calcium imaging, mice were injected with Fast Blue retrograde tracer during tumour infiltration (Video S3). Bone afferents were deemed to be activated (or 'responders') if a calcium transient was detected either during or after intraosseous injection (representative frames and traces taken from a recording are shown Figure 5C and Figure 5D, respectively). We found that only a small proportion (8.5%) of Fast Blue retrogradely labelled neurons responded to intraosseous stimulation. However, 15% of all DRG neurons that were activated by intraosseous injection were also retrogradely labelled with Fast Blue ( Figure 5E). The percentage of bone marrow afferents responding to stimulation of the paw following intrafemoral saline or acid injection did not significantly vary between cancer and sham animals (one-way ANOVA, p = 0.6144) ( Figure 6A). The response intensity of bone afferents also appeared to be unaffected by the presence of cancer (one-way ANOVA, p = 0.4152) ( Figure 6B). Although size distribution of responding bone afferents did not significantly differ between cancer and sham animals (one-way ANOVA, p = 0.1762), a shift to the right was apparent in cancer bearing animals, suggesting that more medium-to-large sized bone afferents are recruited in nociceptive processing in CIBP ( Figure 6C).

Activation of Femoral Bone Marrow Afferents
To investigate the contribution of intraosseous pressure and the acidic microenvironment in CIBP, we used intrafemoral injections of saline and citric acid, respectively, during in vivo DRG imaging of sham and cancer animals. In a subset of mice, we used a pump-controlled syringe connected to a pressure gauge to deliver 10 µL saline into the femoral marrow of sham and cancer bearing animals to serve as an indirect measure of intraosseous pressure ( Figure 4). This initial pressure prior to injection of saline did not significantly differ between cancer and sham animals (125.60 ± 18.96 and 86.67 ± 13.33 mmHg respectively, Welch's t-test, p = 0.1440) ( Figure 5A). The time required for delivering 10 µL saline in cancer animals (52.2 ± 4.41 s) was nearly threefold of that observed in sham animals (12.67 ± 6.22 s) (Welch's t-test, ** p = 0.0020) ( Figure 5B). To visualize bone marrow afferents during in vivo calcium imaging, mice were injected with Fast Blue retrograde tracer during tumour infiltration (Video S3). Bone afferents were deemed to be activated (or 'responders') if a calcium transient was detected either during or after intraosseous injection (representative frames and traces taken from a recording are shown Figure 5C and 5D, respectively). We found that only a small proportion (8.5%) of Fast Blue retrogradely labelled neurons responded to intraosseous stimulation. However, 15% of all DRG neurons that were activated by intraosseous injection were also retrogradely labelled with Fast Blue ( Figure 5E). The percentage of bone marrow afferents responding to stimulation of the paw following intrafemoral saline or acid injection did not significantly vary between cancer and sham animals (one-way ANOVA, p = 0.6144) ( Figure 6A). The response intensity of bone afferents also appeared to be unaffected by the presence of cancer (oneway ANOVA, p = 0.4152) ( Figure 6B). Although size distribution of responding bone afferents did not significantly differ between cancer and sham animals (one-way ANOVA, p = 0.1762), a shift to the right was apparent in cancer bearing animals, suggesting that more medium-to-large sized bone afferents are recruited in nociceptive processing in CIBP ( Figure 6C).  (1) To measure secondary hypersensitivity, noxious pinch, hot water (55 • C), and ice-cold water (0 • C) were applied to the plantar surface of the affected paw. (2) To activate bone afferents 10 µL solution were delivered to the femoral marrow through a syringe and pressure was recorded. (3) Step 1 was repeated to determine if activation of bone afferents resulted in changes in excitability of distal cutaneous afferents.

Response of Plantar Cutaneous Afferents is Unchanged after Intraosseous Stimulation
To investigate if acute stimulation of the bone marrow drives cross-activation of bone and cutaneous DRG neurons, the paw was stimulated again following injection of saline or acid into the femur. We found that the proportion of neurons responding to each modality was unchanged ( Figure  6D) (two-way ANOVA within each group for each stimulus; cancer saline: p = 0.0815 for pinch, p = 0.1269 for heat; for all other groups and conditions: p > 0.9999).

Femoral Bone Marrow Afferents are Largely Nociceptive and not Proprioceptive
We first analysed the rostrocaudal distribution of mouse femoral bone afferent along lumbar DRG using retrograde labelling with Fast Blue. Consistent with sciatic and femoral nerve anatomy in the mouse [36], we found the majority of somata of retrogradely labelled bone afferents in L3 DRG. Similar to previous studies in the rat, we also observed a range of soma sizes of bone afferents [22,37], with most neurons classified as large-and medium-sized. Using three different reporter lines, we identified the nociceptive neuronal marker Nav1.8 to be expressed in over three-quarters of bone

Response of Plantar Cutaneous Afferents Is Unchanged after Intraosseous Stimulation
To investigate if acute stimulation of the bone marrow drives cross-activation of bone and cutaneous DRG neurons, the paw was stimulated again following injection of saline or acid into the femur. We found that the proportion of neurons responding to each modality was unchanged ( Figure 6D) (two-way ANOVA within each group for each stimulus; cancer saline: p = 0.0815 for pinch, p = 0.1269 for heat; for all other groups and conditions: p > 0.9999).

Femoral Bone Marrow Afferents Are Largely Nociceptive and Not Proprioceptive
We first analysed the rostrocaudal distribution of mouse femoral bone afferent along lumbar DRG using retrograde labelling with Fast Blue. Consistent with sciatic and femoral nerve anatomy in the mouse [36], we found the majority of somata of retrogradely labelled bone afferents in L3 DRG. Similar to previous studies in the rat, we also observed a range of soma sizes of bone afferents [22,37], with most neurons classified as large-and medium-sized. Using three different reporter lines, we identified the nociceptive neuronal marker Na v 1.8 to be expressed in over three-quarters of bone marrow afferent neurons. This population largely overlapped with TrkA-expressing neurons, a known marker for these cells [13,14,23]. Interestingly, ablation of Na v 1.8-expressig neurons using Cre-mediated diphtheria toxin expression neither prevents nor attenuates the pain phenotype and bone degradation induced by intra-femoral injection of cancer cells [38]. This suggests that non-Na v 1.8-expressing bone afferents, although a smaller population of DRG cells, are sufficient to induce physiological and nociceptive changes leading to the development of CIBP. Despite the reported absence of nonpeptidergic C-fibres in mouse bone tissue [14,39], depletion of nonpeptidergic IB4+ fibres was sufficient to relief breakthrough pain in a rat model of CIBP [40]. Recent work employing anterograde tracing of sensory afferents revealed the presence of unmyelinated nonpeptidergic fibres in the mouse marrow and periosteum [41]. Based on singe cell RNA-seq data Tmem233 is expressed in nonpeptidergic neurons [42]. We found that Tmem233 is expressed in 1/5th of retrogradely labelled bone afferents, where it also largely overlapped with TrkA immunoreactivity. Our findings are in line with retrograde labelling of bone afferents in the rat [37] and support the involvement of these cells in CIBP. In addition, we observed an almost complete absence of Fast Blue retrogradely labelled cells in the Pvalb-cre+ neuronal subpopulation, indicating bone afferents do not play a role in proprioception. Single cell RNA sequencing data also demonstrates that the Pvalb-expressing population of mouse lumbar DRG neurons is distinct to the Na v 1.8-expressing population of nociceptors [42,43].

Cancer Bearing Mice with a Moderate Reduction of Limb Use Develop Secondary Mechanical Hypersensitivity
Cancer bearing mice presented clear signs of limping, a reduction in weight bearing on the affected limb, and increased nocifensive responses to non-noxious palpation of the distal femur head, as has been previously reported [44,45]. Withdrawal thresholds to von Frey filaments applied to the ipsilateral paw were substantially reduced, while withdrawal to the application of the Randall-Selitto pressure clamp to the ipsilateral paw did not differ between groups. These seemingly contradicting results relate to the difference in stimulus quality; while Randall-Selitto measures static hyperalgesia generally confined to the area of primary hypersensitivity, fine von Frey filaments can detect punctate mechanical hypersensitivity expanding beyond the primary site of hypersensitivity [46]. Overall, the results of this study suggest that CIBP is associated with the development of mechanical hypersensitivity at distant sites, as has been previously described in mice [47] and humans [48]. On the other hand, we did not observe any differences in heat pain thresholds between sham and cancer animals, suggesting thermal sensitivity is unaffected in mice with a moderate cancer pain phenotype, even though others have previously reported a reduction in paw withdrawal thresholds to radiant heat (Hargreaves) in murine models of CIBP [49,50].

Increased Sensitivity of Cutaneous Afferents Innervating the Paw in Animals with CIBP
Several lines of evidence in the current study point towards changes in excitability at the level of the DRG, which contributes to the peripheral drive of nociceptive signalling in CIBP. We found that a higher proportion of L3 cutaneous plantar afferents respond to pinch, as opposed to heat, in cancer bearing animals. This is contrary to what we observed in sham animals, and reflects the behavioural phenotype of mechanical hypersensitivity associated with CIBP in our model. Similar changes in excitability of cutaneous afferents have been reported in a model of osteoarthritis associated with secondary hypersensitivity, with afferents innervating the hind paw showing increased responses to pinch as revealed by in vivo calcium imaging [51]. Our data also shows a significantly higher response intensity of mechanoresponsive plantar afferents in cancer bearing animals, which further supports the data demonstrating that mechanosensitive cutaneous neurons are sensitized in CIBP. Similarly, in vivo electrophysiological recordings from L4 DRG soma have also found that C-, Aβ-, and Aδ-fibres become more excitable, with decreased mechanical thresholds in rats with CIBP [52]. A more recent study of CIBP in rats reported a recruitment of silent small-to-medium sized muscle afferents surrounding the cancer bearing bone in response to knee compression [53].
We observed a substantial increase in polymodal responses of cutaneous afferents in cancer bearing animals, which may reflect a phenotypic shift in thermo-specific DRG neurons that acquire sensitivity to mechanical stimuli. This may underlie the observed increase in response intensity of heat-responding nociceptive neurons in cancer bearing animals, although this was not sufficient to induce behavioural changes in noxious heat thresholds. A previous study reported that subcutaneous administration of inflammatory soup produces a response to mechanical stimulation in previously silent mechanically insensitive heat nociceptors [54]. Moreover, we have also previously reported that intraplantar Prostaglandin E 2 leads to a substantial increase in polymodal responses in mice and encourages a phenotypic switch in response modalities of DRG neurons [34].
We observed a trend of increased DRG coupling in response to mechanical, but not thermal stimuli, although size distribution of these recruited cells did not differ between sham and cancer animals. A substantial increase in neuronal coupling events, which are mediated by glial gap junctions, has previously been reported for both inflammatory and neuropathic pain, with both imaging and behavioural data indicating this phenomenon is more prominent for mechanical, rather than thermal hypersensitivity [35]. The discrepancy in effect size between these findings and the current study may depend on our approach to image at the level of L3, whereas L4 DRG contain somas of the majority of neurons with receptive fields in the hind paw [55]. While CIBP shares some features of inflammatory and neuropathic pain, it is seen as a distinct pain state [9,56] for which cross-activation of sensory neurons may play a minor role in peripheral sensitization.
The secretome of the cancer cells constitutes a potential mechanism for the sensitization of cutaneous afferents innervating distant sites to the tumour bearing bone. Neuronal sensitization has been observed using no-contact co-cultures of fibrosarcoma cells and DRG neurons [57]. Moreover, conditioned medium from painful, but not non-painful schwannomatosis tumours sensitizes sensory neurons. Higher levels of pro-inflammatory cytokines were detected in the secretome of these cells and may mediate the observed hyperexcitability [58]. In addition to these findings, a recent study reported that intraplantar injection of conditioned medium from an oral cancer cell line leads to mechanical and thermal hyperalgesia, but not if depleted from exosomes [59]. We cannot exclude the possibility that cancer cells release inflammatory mediators in the local microenvironment, and with the increased vascularization of the metastatic femur [60] these mediators may act systemically to produce hypersensitivity of cutaneous afferents.

Response Properties of Bone Afferents Are Unaffected in Mice with Moderate Bone Cancer Pain
To investigate the response properties of bone afferents to acute stimulation, a 10 µL solution was injected in the mouse femur and activity of DRG neurons was recorded by in vivo calcium imaging. The time required to inject a 10 µL solution in the intramedullary cavity of the mouse femur was significantly increased in animals with CIBP and the pressure of the injected solution tended to be higher in cancer bearing animals. Overall, these findings indicate that an enhanced intraosseous pressure is created by the presence of tumour and are in line with previous work showing a two-fold increase in intramedullary pressure by bone cancer [18]. To stimulate bone afferents, a solution of saline was injected to produce increased intraosseous pressure [22]. Citric acid injection was used to simulate the acidic microenvironment, which is thought to contribute to sensitization of primary afferents innervating cancerous tissue [26,28]. In both cases, we were surprised to find that the response properties of bone afferents, unlike primary afferents innervating the glabrous hindpaw skin, were unaffected by the presence of cancer. We acknowledge some limitations of our study; Although we always injected 10 µL of solution, which is the maximum capacity of the mouse femur [61], we cannot assure pressure thresholds were met in each animal. Moreover, as intraosseous pressure between cancer and sham animals varies, the injection stimulus may produce different changes in pressure in the two groups. Finally, a limitation of our in vivo imaging setup is that neurons are recorded only within a particular field of view of the entire DRG. Given the small proportion of bone afferents within all DRG neurons, the probability of imaging from one of these cells that is responsive to a particular stimulus type is low and could explain why we only detected few responding cells in each group. Nevertheless, our findings are in line with a recent study indicating that muscle afferents innervating the tissue surrounding the cancer bearing bone contribute to peripheral sensitization in rats, rather than bone innervating neurons themselves [53]. It should be noted that this study applied pressure outside the bone using a cuff, which resulted in activation of very few bone marrow afferents, while we instead produced increased intraosseous pressure. Similarly, electrophysiological studies point towards ectopic activity in cutaneous C-fibres surrounding the cancer bearing bone [62]. Overall, these results suggest bone afferents are not mediating nociceptive processing in intermediate to late stage cancer. However, we cannot exclude the potential role of bone marrow afferents in early establishment of CIBP, as during disease progression the fibres innervating the bone undergo a continuous cycle of sprouting, degeneration and re-sprouting [63,64].

Cell Culture
LL/2 Lewis Lung carcinoma cells (ATCC) were cultured in DMEM supplemented with 10% FBS and 1% Penicillin/Streptomycin for at least 2 weeks prior to surgery. Cells were split at 70-80% confluence two or one day prior to surgery (cell culture reagents supplied by Thermo Fisher). On the day of surgery, cells were harvested with 0.05% Trypsin-EDTA, resuspended in DMEM at a final concentration of 4 × 10 7 cells/mL and kept on ice until use. Cells were counted before and after intrafemoral injection to confirm viability.

Animals
For genetic labelling of neuronal subsets, homozygous Rosa-flox-stop tdTomato mice [65] were crossed with homozygous Na v 1.8-cre (Scn10a tm2(cre)Jnw ) [66], Tmem233-cre (generated in our lab), or Pvalb-cre (B6;129P2-Pvalb tm1(cre)Arbr /J) [67] homozygous mice. In vivo calcium imaging experiments were performed using heterozygous Pirt-GCaMP3-expressing mice (at least 12 weeks old; male and female) on a C57BL/6J background, generated by X.D. (Johns Hopkins University, Baltimore, MD) [68]. Where indicated, homozygous Pirt-GCaMP3 and Rosa-flox-stop tdTomato mice were crossed with homozygous Calb1-cre [69] or Tmem233-cre mice. For genotyping, genomic DNA was isolated from ear tissue. Primers used for PCR are summarized in Table S1. Mice were housed in groups of 2-5 animals with a 12-h light/dark cycle and allowed free access to water and standard diet. All animals were acclimatized for 2 weeks before the start of the experiment. All experiments were performed with approval of personal and project licenses (licence number P413329A2) from the United Kingdom Home Office according to guidelines set by the Animals (Scientific Procedures) Act 1986 Amendment Regulations 2012, as well as guidelines of the Committee for Research and Ethical Issues of IASP. Any mice that developed limping within 4 days after surgery were excluded (n = 2). Cancer animals that failed to develop a reduction in limb use score over the course of 20 days (n = 8) were also excluded.

Surgery
Cancer cells or DMEM were administered intrafemorally as previously described [33,38], with a few adaptations to allow for injection of a retrograde tracer. A Hamilton syringe with an attached canulae was used to inject 5 µL of 2 × 10 5 LL/2 cells or DMEM medium (sham). This was kept in place for 2 min to allow cells to set. The canulae was quickly replaced with a new canulae to deliver 1 µL of 2% Fast Blue solution in water. Another 5 min were left to allow the tracer to set, before closing the hole. For colocalization experiments, only Fast Blue was injected.

Behavioural Tests
For behavioural experiments, animals were acclimatized to the equipment for at least 2 days prior to testing. The experimenter was blind to the groups. The progression of CIBP was assessed by limb use score and a cut-off of 2 or 1 (i.e., significant limping) was used as the endpoint for cancer animals, as previously described [33]. Deficits in static weight bearing were assessed using an Incapacitance Meter as previously described [70]. Mechanical sensitivity of the ipsilateral hindpaw was measured using the von Frey up-down method [71] and Randall-Selitto apparatus [72]. Mechanical sensitivity of the tumour-surrounding tissue was determined by non-noxious palpation of the distal femur head [44]. Thermal nociception was assessed on the hot-plate test at 50 • C as previously described [70]. Limb use score and weight bearing was performed on all animals. For von Frey data, some animals were excluded as withdrawal thresholds were measured before the animals reached a limb use score of 2. All other tests were performed on a different subset of animals to accommodate for multiple testing and avoid stress evoked by multiple procedures carried out on a single day (detailed in Table S2).

In Vivo Calcium Imaging
In vivo calcium imaging was performed as previously described [34], with a few adaptations. The laminectomy was performed at spinal level L2-L4 to expose the L3 DRG for imaging. Additionally, the lateromedial aspect of the left femur was exposed, by separating the biceps femoris posterior from the biceps femoris anterior. A hole was drilled at about 1 cm from the distal femur head; a canulae was inserted and fixed in place with dental cement. In vivo imaging was performed using a Leica SP8 confocal microscope (Dry ×10, 0.4-N.A. objective with 2.2-mm working distance, Leica). Scans were taken at a bidirectional scan speed of 400-600 Hz at a resolution of 512 × 512 pixels in either one or two z-planes. Pinhole A.U. was kept between 1.22 and 3.85 to visualize single cells accurately. A laser line of 405 nm, 488 nm was used to excite Fast Blue and GCaMP3, respectively. The collection of the resulting emission was system optimized to maximize yield and minimize crosstalk (Leica Dye Finder, LASX software; Leica, Wetzlar, Germany).

Peripheral Stimulation
Study design of peripheral stimulation is shown in Figure 4. In vivo calcium imaging of L3 DRG was performed once cancer bearing animals reached limb score 2 or 1, representing a time point for significant sensory and motor dysfunction. (1) First, the glabrous skin of the ipsilateral hind paw was stimulated in order to activate calcium transients in cutaneous plantar afferents. Tweezers were used to apply pressure (pinch stimulation) for 3 s across the dermatome covering L3-L4 for mechanical stimulation, followed by transient immersion of the paw in hot water (55 • C) and ice-cold water (0 • C) for 10 s for thermal stimulation. Each stimulus application was separated by 30 s. (2) Second, for activation of bone marrow afferents a 10 µL solution of either saline (0.9% NaCl) or 0.1 M citric acid (pH = 4), containing 2.5 mg/mL Blue Evans was injected in the mouse femur. Bone marrow afferents were counted as responders if there was a change in unevoked baseline fluorescence either during the injection or within 30 s after the injection. In a subset of animals, the solution was delivered by a pump-controlled system (Harvard Apparatus Plus), where a 1 mm syringe was connected to both pressure gauge and a canulae containing 40 µL volume. The gauge recorded changes in pressure with a maximum limit reading of 250 mmHg during the constant flow rate set at 10 uL/s until at total of 10 µL solution was delivered. All but two animals reached the maximum limit reading before 10 µL solution was injected, and we therefore assumed that pressure was still rising above this value. In another subset of animals, pressure was applied manually using a syringe in order to produce consistent changes in intraosseous pressure. The changes in pressure were not recorded during manual delivery.
(3) Finally, cutaneous plantar afferents were stimulated again as described in point 1. After in vivo calcium imaging, a z-stack was recorded for counting the overlap between Fast Blue+ and tdTomato labelled cells. A necropsy was performed to confirm that the injected solution stayed within the femur and confirm that only neurons in the ipsilateral DRGs were labelled with retrograde tracer Fast Blue. We did not observe any leakage, and cancer growth was confined to the bone. Animals showing Fast Blue labelling in contralateral DRG were excluded from retrograde tracing analysis.

Image Analysis
All in vivo imaging data were acquired with the LAS-X analysis software (Leica) and analysed with ImageJ. All images were stabilized for XY movement using the TurboReg plug-in [73], with all images being registered to a stable image of the series. Raw traces of calcium signals were generated through the free hand selection tool of regions of interest (ROIs) surrounding cell bodies, which responded to stimulus application. Area was used to determine average cell size of responding cells and average pixel intensity as a measure of change in calcium transients. Data were analysed by a combination of MATLAB R2017a and Microsoft Office Excel 2013. Raw traces were first smoothed by averaging the preceding four frames of any test frame to reduce noise. To determine if a neuron was responsive to a given stimulus, the derivative of each frame was taken as ∆F/∆t. Neurons were counted as responders to a given stimulus if ∆F stim ∆t > ∆F basal ∆t + 4σ basal , where F stim is the maximum derivative value within a given window of stimulus application, F basal is the average of derivative values in a 10 s time window preceding stimulus application, and σ basal is the SD of the baseline derivative values. All neurons identified as responders were double-checked visually to avoid signal contamination by cells with partially overlapping ROIs. To generate normalised data for each trace, the following equation was applied F−F min F max −F min . Cutaneous afferents were classified as polymodal if they responded to at least two of the three test modalities. Coupled responses were determined manually and defined as two or more cell bodies within 5 µm of each other responding to the same stimulus application. Relative presence of coupling events was determined by dividing the number of coupled responses by the total number of responses to that particular stimulus. To identify if previously silent low-threshold mechanoreceptors (LTMRs) are recruited in cancer animals, we determined size distribution of responding cells, in which case we would expect a shift of the distribution to the right.

Statistical Analysis
Statistical analysis was performed using GraphPad Prism 8. Log-rank test was used to compare survival distributions between groups. Welch's t-test (two-tailed) and one-way ANOVA were used to compare means between two or more groups, respectively. Kolmogorov-Smirnov test was used to compare difference between two distributions. Comparison of multiple time points between groups was performed by two-way ANOVA with Bonferroni post-hoc test. Data is presented as mean ± standard error of the mean (S.E.M.), and significance as: * p < 0.05; ** p < 0.01; *** p < 0.001, **** p < 0.0001.

Conclusions
We observed that the majority of afferents innervating the mouse femoral bone marrow express Na v 1.8, and the response properties of these DRG neurons to both pressure and acidic stimuli are unaffected in a mouse model of CIBP. However, cancer-bearing animals did exhibit increased intraosseous pressure, as evidenced by prolonged duration to inject a solution into tumour bearing femurs. Surprisingly, we found that cutaneous afferents innervating the glabrous skin of the hindpaw in animals with CIBP show increased excitability at the level of the soma, as well as a phenotypic shift from thermosensitivity to increased mechanosensitivity. These findings are consistent with the behavioural phenotype of secondary mechanical hypersensitivity observed in cancer bearing mice compared to sham controls. Our data supports previous findings that pain behaviour and disease progression in a mouse model of CIBP occurs independently of bone marrow-innervating afferents, and is likely driven by an increase in polymodal neurons innervating secondary sites.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6694/12/12/3491/s1, Figure S1: Size distribution of retrogradely labelled femoral bone marrow afferents within each genetically labelled neuronal subset; Supplementary methods including Table S1: Primers used for genotyping, and Table S2: Overview of animals used for behavioural testing; Video S1: Representative recording of cutaneous afferents responding to stimulation of the paw in a sham animal; Video S2: Representative recording of cutaneous afferents responding to stimulation of the paw in a cancer animal, Video S3: Representative recording of bone marrow afferents responding to intraosseous injection.