Novel mechanisms of bone cancer pain revealed with in vivo GCaMP6s imaging

Cancer pain remains a major area of unmet medical need, with few studies existing in proportion to that need. One common form, affecting 400,000 people each year in the US alone, is associated with skeletal metastases. These pains are typically mechanoceptive in nature and poorly managed by available analgesics. Here, we employed in vivo imaging using GCaMP6s to assess the properties of the nerve fibres that convey bone cancer pains. We find that a subclass of nociceptors, those that are normally mechanically insensitive, are recruited and activated in a rodent model of bone cancer, and this dramatically increases sensory input from the diseased tissue to the central nervous system. The recruitment of these so-called silent afferents was found to be Piezo2-dependent. The unique properties of these silent afferents offer several novel opportunities for targeting metastatic bone pain.


Introduction
Following a cancer diagnosis, patients must deal with a range of unpleasantries resulting not only from cancer growth but also from treatment. A significant side effect of certain cancer types is severe pain, which can persist even in remission [1][2][3] .
Cancer pain is particularly prominent in cases where malignant tumours have invaded the skeleton 2,4 . Neurologists broadly classify pain into two types: nociceptive (sharp, throbbing; stimulus driven) and neuropathic (abnormal sensations due to nervous system damage) 3 . Cancer-induced bone pain (CIBP) is a unique mix of both, with sufferers experiencing tonic, spontaneous and movement-evoked pain.
The latter is also known as "break-through" or suprathreshold pain. This type of pain, being mechanoceptive in nature, is difficult to manage in mobile subjects, as by definition, it 'breaks-through' the barriers of analgesia 3  Over the past decade, rodent studies of CIBP have revealed that neurons and cancer cells are engaged in bi-directional crosstalk. For instance, cancer causes a reorganization of normal anatomy, driving neurons to sprout and more densely innervate the tumour-bearing bone 2,5,6 . This sprouting process is mediated via tyrosine kinase A (TrkA) receptor activation by nerve growth factor (NGF) released from both cancer and stromal cells [7][8][9] . Conversely, neurons release factors which support tumour growth and vascularization [10][11][12] . This complex dialogue involves numerous mediators and different local cells, including fibroblasts, osteoclasts and newly recruited immune cells 3,10 .
Despite pioneering work, many questions remain regarding bone afferent function, both in health and disease. For instance, do they encode mechanical stimuli, and if so, how? And, most importantly, how are their functional responses altered in cancer conditions? This study was designed to address these outstanding questions using cell-type specific molecular approaches together with in vivo 4 functional imaging. We utilised an animal model of breast cancer metastasis to study CIBP. The model is recognised as having high face and construct validity compared to its human counterpart 3 .
The data gathered reveal important new information about bone afferent expression patterns, encoding of mechanical stimuli and potential functional mechanism explaining hyperexcitablity in the presence of a tumour. Taken together, these findings could help identify new therapeutic avenues for treatment of patients with CIBP.

A high proportion of deep body CGRP afferents express little or no Advillin
To study bone afferents in the absence of a selective marker, we injected the retrograde tracer fast blue (FB) into the tibial cavity. All lumbar DRG, ipsi-and contralaterally to the injection were dissected, cryosectioned and analysed. We found 18%, 51%, 24% and 7% of all FB+ cells in L2, L3, L4 and L5 DRG respectively ( Fig S1A, B), and no FB+ cells on the contralateral side (Not shown), indicating that our tracing was specific. Next, we showed that the traced tibial afferents stained positively for markers of peptidergic nociceptors (e.g. calcitonin gene-related peptide (CGRP) and TrkA, a receptor for NGF), but were rarely positive for isolectin B4 (IB4), a marker for non-peptidergic neurons (Fig S1A, B).
We examined co-expression of a virally delivered genetically-encoded calcium indicator (AAV9-GCaMP6s) and a commonly used marker for sensory neurons, Advillin (Avil). We performed this analysis using a knock out-validated antibody 13 .
Surprisingly, around 40% of FB/GCaMP++ cells were Avil negative (Fig. S1C). To ensure validity, we investigated FB-traced tibial afferents from rats that were not subjected to GCaMP6s delivery. The tissue was co-stained for CGRP and TubulinβIII (TubβIII, a known pan-neuronal marker) to further ensure the presence of well-defined neuronal populations ( Fig. 1A and S1D). Almost 41% of tibial afferents did not express Avil (Fig. 1B). These Avil negative fibres often (55%) co-localise with CGRP (Fig. 1B). Similarly, in the mouse, we employed a PACT-clearing technique 14 to visualise all neurons within the DRG of Avil-eGFP mice. Counterstaining with CGRP revealed a discrete population of CGRP+, Avil-neurons (around 10% of the total) (Fig.1C, S1E, See also Movie 1,2). A search through two separate, publicly 5 available single-cell RNA sequencing databases further supported our finding of differential expression between CGRP (and Tac1, a gene for substance P, another peptide found in peptidergic nociceptors) and Avil in sensory neurons (Fig. S2A, B) 15,16 .

Cancer progression impacts bone innervation
We investigated how the presence of bone cancer affects these patterns of innervation by generating a validated rat CIBP model using syngeneic mammary gland carcinoma cells (MRMT1) 17 (Fig. 2B). Damage to afferents innervating cancerous tissue was quantifiable using activating transcription factor 3 (Atf3), a protein induced by cellular stress 18 . Representative micrographs of the L3 DRG ipsilateral to the injury site clearly demonstrated the characteristic nuclear expression pattern of Atf3 in bone and other afferents (Fig. 2C). This is especially evident in early stages of our CIBP model (Fig. 2F, S2A, B). Interestingly, by the late stage, Atf3 positivity normalises in both groups, with almost no occurrence in late-stage sham animals, suggesting full postsurgical recovery (One Way ANOVA [group]: F 3, 17 = 14.37, P < 0.0001, Tukey post-hoc: CIBP early vs. Sham early P < 0.0001, CIBP early vs. CIBP late P < 0.01, Sham late vs. Sham early P < 0.01) (Fig. 2F, S2A, B). Bone afferents are more likely than other afferents to express Atf3 at early disease stages, suggesting higher levels of stress in this population (One Way ANOVA [group]: F 3, 17 = 8.844, P < 0.001, Tukey post-hoc: CIBP early vs. all the others P < 0.01) (Fig. 2G). Moreover, there is 6 a visible shift in the expression pattern of Atf3+/FB+ from L3 to L4 DRG between early and late CIBP (Fig. S2B).
To examine whether tumour progression translates to animal behaviour, we monitored rats for up to 16 days after cancer cell implantation. Body weight gain remained stable in this period when compared to the sham animals (Fig. S2C).
Behavioural data clearly demonstrate that animals with CIBP manifest mechanical hypersensitivity, with significant changes in static weight bearing between rear legs starting at day 7 post-surgery and progressing with time (CIBP: Kruskal-Wallis for independent samples [days]: P < 0.0001: day 2 and day 7 vs. baseline P < 0.05, day 14 vs. baseline P < 0.001) (Fig. 2H). Sham animals did not demonstrate any significant alterations in the weight-bearing test after day 7, confirming valid model establishment. Sham surgery did affect pain thresholds in the first 2 days suggesting the presence of transient postsurgical pain (Sham: Kruskal-Wallis for independent samples [days]: P < 0.05: day 2 vs. baseline P < 0.01) (Fig. 2H).

DRG afferents encode mechanical pressure stimuli in a graded fashion
We implemented in vivo GCaMP6s imaging to thoroughly analyse bone afferent physiology in the late cancer stage since mechanical hypersensitivity was evident most clearly here. Considering our anatomical findings of low Avil expression in bone afferents, we opted for intrathecal delivery of an AAV9 viral vector containing GCaMP6s. This method of delivery was previously shown by our lab to ensure uniformly distributed expression between all subtypes of the DRG neurons 19 and Fig.   S1). We analysed 757 DRG neuronal cell bodies from 9 sham-operated rats, and 1750 neurons from 10 CIBP animals transduced with AAV9/Syn.GCaMP6s. Lumbar DRG L3 and L4 were imaged (based on our FB tracing studies (Fig. S1, Fig. 2D, E, and 18,20 ), and 6-7 L3 DRG and 3 L4 DRG from each group were imaged. No response difference was detected between L3 and L4 lumbar levels (not shown), hence results from all DRG were pooled.
To apply pressure stimuli, we implemented a novel method using a neonatal cuff connected to a manometer and air pump. The cuff was applied sequentially in all animals to the following rear limb regions: knee and tibial head, calf, and calf-ankle.
Incremental pressure (50 mmHg every 10 s, in the range of 0-400 mmHg) was applied (Fig. 3A, S3A). Utilising this stimulation regimen, individual DRG neurones showed phasic and tonic increases in Ca 2+ when pressures were increased (Fig.   7 3B). In addition to cuff stimulation, we also examined proprioceptive responses, by gently moving the limb along the body axis in 5 consecutive push-pull stretching cycles. Our results were analysed using in-house R scripts. Very stringent criteria were applied in order to select responses: fluorescence intensity was counted as a positive response when an average signal reached 70% above baseline fluorescence plus 4 standard deviations 19 .
A striking difference in the number of neurons recruited between sham and CIBP groups, especially after the knee compression (which mostly covered the tumour-growth area) was clear. The effect was detected with the naked eye on video-rate recording (Movie 4). Selected frames from before and after stimulation Interestingly, increased compression forces (range 100-400 mmHg) were reflected in the linear recruitment of responders in the CIBP group (Fig. 3D). In contrast, the number of responders in sham group did not appear to increase linearly. There appeared to be a threshold between 50-100 mmHg, after which all potential mechanoceptors within the imaged field of view (FOV) were responding to the chosen receptive field stimulation (Fig. 3D).
Fluorescent intensity coded in line with the pressure surge applied to the knee in both groups (Fig. 4A). Unlike cuff pressure, proprioception appears to be encoded by the same cells responding to different leg positions (push-pull stretches) without the change in fluorescence (Fig. 4B, S4B).

CIBP recruits previously silent nociceptive C fibres
We analysed the cell size distribution of responders in healthy and cancer states. We chose 700 µm 2 and 1200 µm 2 to crudely separate sensory neuron type (small and medium-size cells). Our results suggest that pressure is encoded mainly 8 by small to medium-size neurons. As expected, the average cell size decreased with the increase of the force applied (Fig. 4C). Intriguingly, there were significant differences in the average cell sizes of responders to knee compression (RM-ANOVA [group]: F 1, 202 = 16.24, P < 0.0001) (Fig. 4C), but not leg movement (RM-ANOVA [group]: F 1, 303 = 0.239, P = 0.625) (Fig. 4D, S4C), between cancer and sham animals. Specifically, in sham animals, the number of medium-size responders increased with increasing stimulus pressure (Fig. 4E), while dynamic brushing of the calf only recruited a few large-sized neurons. Meanwhile, in our cancer group, an additional population of small diameter neurons (likely C nociceptors) was activated proportionally with increasing stimulus strength, reaching almost 3 times the number of cells that responded to the 400 mmHg than the initial 50 mmHg (Fig. 4E).

Intratibial afferent function in health and CIBP
We next focused on bone afferents using FB tracing. Firstly, we investigated whether the neurons innervating the tibial cavity express Piezo-type mechanosensitive ion channel component 2 (Piezo2) and TrkA; an abundance of both proteins on bone afferents was observed (Fig. 5A). Secondly, we examined the functional responses of bone afferents (Fig. 5B, C, D). Sensory cells innervating tibial cavity were responsive to whole limb mechanical stimulation (Fig. 5E, S5A).
Around 13% of traced tibial cavity afferents responded to knee compression (Fig.   5E), and considerably more responders (~25%) were observed following calf compression (Fig. S5A). Leg movement activated up to 25% of all traced cells (Fig.   5F). Compared to non-traced cells (Fig. 3D), the engagement of FB+ responders to their receptive field compression required higher pressures (around 200 mmHg) ( Fig.   5E), suggesting that bone has a higher pressure activation threshold than more cutaneous layers.
Interestingly, there was no difference between sham and cancer-bearing groups regarding the number of responders to knee compression and movement   Muscle and periosteum afferents are recruited and sensitized by bone cancer Bone afferents were not sensitised in CIBP but what about local periosteum and muscle afferents? Analogous to the bone afferents we traced muscle and periosteum (MP) afferents by injecting AAV-retrograde virus expressing tdTomato outside the tibia (Fig. 6A, B, C). Piezo2 and TrkA were both present in the traced cells (Fig. 6D). Further, we showed that virtually none of the MP traced cells from the sham group responded to mechanical stimulation (Fig. 6E, S6A In this study, we identified a discrete population of peripheral neurons that innervate bone and lack Avil expression. Using in vivo calcium imaging, we also revealed novel mechanisms that drive cancer-induced bone pain.
A high proportion of deep body CGRP afferents express little to no Advillin We demonstrate that many bone afferents express markers of peptidergic (e.g. CGRP and TrkA), but rarely non-peptidergic (IB4) nociceptors, in agreement with previous reports 9, [24][25][26] . Further, we discovered that a large population of bone afferents rich in CGRP lack Avil expression. Over the last decade there have been repeated efforts to more accurately define sub-populations of sensory neurones 15 .
Avil, an actin-binding protein, was anticipated to be selectively expressed by up to 97% of all sensory DRG cells 27 . Following this announcement, several murine models were created to drive transgene expression utilising Avil. Recent research suggests that Avil is expressed not only in sensory neurons, but in most adult neural crest-derived neurons including sympathetic fibres 13 . Others reported that Avil-driven transgene expression does not cover around 10-15% of sensory neurons. A restriction to any particular population of the DRG cells, however, was not described 28,29 . Publicly available single-cell RNA sequencing databases supported our finding of differential expression between CGRP and Avil on the whole population level 15,16 . Overall the presence of a discrete (around 10% of all DRG neurons) but significant population of peptidergic afferents that are low or almost not expressing Avil, although enriched in CGRP, is reported. This type of afferent appears to be particularly abundant in bone innervation. The physiological meaning of the lack of Avil in bone afferents remains an open question, not least since the exact function of Avil is unknown.

Cancer progression impacts bone innervation
We employed a highly-validated rat CIBP model 17

CIBP recruits previously silent nociceptive C fibres
We showed that pressure is encoded mainly by small to medium-size neurons. As expected, the average cell size decreased as the force applied increased (increasingly painful levels of pressure recruit small nociceptive afferents preferentially over large myelinated Aβ mechanosensors). Previously, dynamic brushing on the calf surface proved that light touch is encoded by Aβ fibres 19,37 .
We found a three-fold increase in the number of neurons responding to knee compression in CIBP, as compared to sham animals. Interestingly, these cells were small to medium-size, in contrast to results obtained from sham animals where pressure activated mainly medium-size neurons. The latter is in keeping with previous literature which suggests that in healthy animals, noxious compression is encoded preferentially by myelinated Aδ nociceptors 30 . The robust recruitment of cells, in particular a large cluster of small-size cells, in CIBP rats suggests activation of previously silent nociceptors. The existence of these kinds of fibres has been reported before 38 , but never in the context of cancer.
Next, we went on to investigate where these silent nociceptors might originate from. We traced afferents from within the bone with FB and found that they respond to mechanical forces, as demonstrated previously by electrophysiology 30,31 .
However, no additional recruitment was detected in this neuronal population suggesting that silent nociceptors in late cancer stage originate outside of the bone.
Again, we observed linear coding of pressure, in agreement with previous electrophysiological reports of tibial axon coding to increasing intraosseous pressure 30,31 . The reduced fluorescence of bone afferents to the proprioceptive stimulation in CIBP animals suggests impaired bone proprioceptor functioning and/or their loss in tumour conditions.
These functional results support our anatomical findings (decreased Atf3 staining in the late CIBP group) (Fig. 2G, S2) and suggesting a degree of bone afferent loss in advanced cancer states. This is also supported by the shift in FB positivity that occurs from L3 to L4 DRG in the late stage CIBP (Fig. S2). The anaerobic and toxic conditions of the tumour are likely to evoke degeneration of locally entrapped afferents. 1 3 Silent nociceptors in CIBP originate from muscle or periosteum via a Piezo2dependent mechanism Since bone afferents themselves do not appear to be the neurons responsible for sensitization, we next used a viral approach to label MP afferents. We found them to be silent in sham operated animals, but responsive in cancer conditions. This suggests that in the late stage of the disease, cancer induces employment of silent nociceptors from bone surroundings, rather than the bone cavity itself. Our anatomical results must be taken with some care. Virally-delivered tdTomato expression levels around the bone remained low and we had to limit the amount of virus used to ensure specificity and to avoid off-target labelling in contralateral DRG.
This meant that our cell numbers were low compared to the rest of our analyses ( Fig.   6B, C).

Conclusions
In agreement with Prato and colleagues 21 , we provide evidence that silent nociceptors are present in deep tissues. In our model of CIBP they are activated to potently increase nociceptive input to the CNS. We hypothesise that locally released NGF unlocks these dormant cells leading to additional nociceptive signalling in the spinal cord, ultimately amplifying pain perception (Fig. 7E). This mechanism may be one of the reasons why bone cancer patients experience a unique type of mechanoceptive pain even after remission: silent nociceptors may remain active.
Anti-NGF therapy could therefore benefit this patient population in two ways. First, by limiting sprouting of afferents into the affected bone and second, by preventing recruitment of silent nociceptors from the surrounding muscle. 1 4 Finally, we hypothesise that Piezo2 plays a critical role in unsilencing nociceptors in cancer. The channel, with its variety of splice variants, has been established as a definitive mechanosensitive protein in mammalian tissue 33 .
We are yet to confirm which isoform is the most important in the mechanism described here, but our data suggests that Piezo2 blockers could be a valuable additional target in the fight against CIBP.

Materials and Methods
Cell lines In vivo calcium imaging of sensory neurons Rats were anaesthetised using urethane (12.5% w/v in saline, Sigma, UK).
Starting with an initial dose of 0.5 ml given i.p., subsequent (0.5ml) given at approximately 10-15 minute intervals, depending on hind limb reflex activity, until surgical depth was achieved. The core body temperature was maintained close to 37°C using a homeothermic heating mat with a rectal probe (Harvard Apparatus).
Tracheotomy was performed to secure steady breathing. An incision was made to the skin on the back and the muscle overlying the L3, L4 and L5 vertebral segment was removed. The bone around either the L3 or L4 DRG was carefully removed and the underlying epineurium and dura mater over the DRG were washed and moistened with normal saline. The position of the animal's body was varied between prone and lateral recumbent to orient the DRG in a more horizontal plane. The exposure was then stabilised at the neighbouring vertebrae using spinal clamps (Precision Systems and Instrumentation) attached to a custom-made imaging stage.
The exposed cord and DRG were covered with silicone elastomer (World Precision Instruments, Ltd) to avoid drying and to maintain a physiological environment. The

Static Weight Bearing
Behaviour was assessed 2-4 hours before surgery (day 0) and at 2, 7 and 14 days following cancer cell injection. Testing was preceded by a 30 min acclimatisation period. Rooms conditions used for behavioural testing were strictly controlled, maintaining stable levels of humidity (40-50%) and temperature (22±2°C).
Weight bearing was assessed using a tester (Linton Instrumentation, Norfolk, UK) in which rats were placed in a plexiglass enclosure where each hindpaw rested on a separate weighing plate. After a few minutes of habituation, the force exerted by each hind paw was measured 5 times with a 10-20 s gap between measurements.
Measurements from each paw separately were averaged, and results were transformed to give the percentage of weight borne on each side to the total rear legs bearing (taken as 100%).

Micro-computed tomography of cancer-bearing legs
Rats were sacrificed by overdose of isoflurane (5% vol/vol) and transcardially perfused with 250 ml of cold phosphate buffer saline solution (PBS, pH=7.5, 1 9 Invitrogen, Paisley, UK) followed by 4% paraformaldehyde solution in 0.  See also Figure S1 and Movie S1, Movie S2.   See also Figure S2 and Movie S3.  See also Figure S3 and Movie S4. See also Figure S4 and Movie S4.