Pathology study showing the association of sensory neuropathy and stem cell mobilopathy in patients with type 2 diabetes
We first sought evidence of sensory neuropathy in the BM of patients with type 2 diabetes by conducting an analysis of neuronal fibre density on specimens collected from iliac crest biopsies or bone left over from hip reconstructive surgery. Since nerve density was similar in the two source groups, data were pooled in the final analysis. By immunohistochemistry, we recognised a striking decrease in the total number of nerve fibres that express the general neuronal marker PGP9.5 in the BM of diabetic patients with neuropathic (T2DM-N) and neuropathic/ischaemic complications (T2DM-NI) (Fig. 1a, ANOVA p < 0.001). This defect extends to nociceptive fibres that co-express PGP9.5 and SP (Fig. 1b, ANOVA p < 0.01) and to tyrosine hydroxylase-positive sympathetic fibres (ESM Fig. 1a, p < 0.05 vs non-diabetic controls). Additionally, in the BM of T2DM-N and T2DM-NI patients, both PGP9.5-positive and SP-positive nerve fibres showed a marked increase in diameter (Fig. 1c, ANOVA p < 0.0001 and p < 0.01, respectively), as well as vacuolisation and infiltration with CD68-positive macrophages (Fig. 1d). These degenerative/inflammatory features were not observed in non-diabetic controls. This association of diabetes with nociceptive fibre rarefaction and remodelling was additionally confirmed in a multivariable analysis adjusted by age, BMI, fasting glucose and HbA1c. Furthermore, neuronal rarefaction was associated with reduction in BM capillary density (ESM Fig. 1b), which identifies a distinct form of microangiopathy we have described previously in animal models and patients with diabetes [23, 24]. Furthermore, by immunofluorescence microscopy, we detected a significant reduction in SP expression in the BM of diabetic patients with neuropathy (p < 0.01 vs non-diabetic controls) but not in those with superimposed CLI (Fig. 2a). In addition, SP levels were remarkably reduced in the PB of neuropathic diabetic patients with or without CLI (p < 0.001 vs non-diabetic controls; p < 0.001 vs uncomplicated diabetes) (Fig. 2b).
Having documented the presence of structural and functional nociceptive alterations in the BM of patients with type 2 diabetes, we next conducted immunofluorescence and flow cytometry analyses of CD34+ HSPCs, which co-express the SP receptor NK1R. Patients with complicated diabetes showed a depletion of CD34+NK1R+ HSPCs in their BM (Fig. 3a–d) and PB (Fig. 3e–h). Mean fluorescence intensity analysis indicated that NK1R was downregulated in cells from diabetic patients (Fig. 3g). Furthermore, superimposed CLI failed to induce an increment in circulating CD34+NK1R+ HSPCs (Fig. 3e–h). Altogether, these data indicate that sensory neuropathy is associated with BM depletion and altered release of a subpopulation of HSPCs that is responsive to SP chemoattraction. In particular, the presence of ischaemia, which reportedly acts as a potent stimulus for CD34+NK1R+ HSPC release , fails to promote cell mobilisation in patients with type 2 diabetes.
To confirm the presence of a specific form of mobilopathy affecting the subpopulation of CD34+NK1R+ HSPCs, we investigated an independent cohort of 24 non-diabetic individuals and 74 patients with type 2 diabetes (ESM Table 3 and ESM Fig. 2). A multivariable analysis showed reduced PB levels of CD34+NK1R+ HSPCs in the diabetic patients as compared with non-diabetic individuals, even after adjustment for BMI and medications that were significantly different at univariate analysis.
Experimental study showing the inhibitory effect of type 2 diabetes on nociceptive-mediated HSPC mobilisation and homing in mice with induced limb ischaemia
We next conducted studies to compare the abundance of nerve fibres in the BM of Lepr
db/db diabetic mice and non-diabetic controls and to investigate the effect of experimental type 2 diabetes on NK1R+ HSPC liberation as stimulated by peripheral ischaemia. The abundance of neuronal fibres expressing the pan-neuronal marker PGP9.5 and nociceptive SP-expressing nerves was measured in the marrow (Fig. 4a, b) and compact bone (Fig. 4c). Morphometric analyses indicated that diabetes induces a marked reduction in the density of both PGP9.5-positive fibres (Fig. 4d, p < 0.01 vs non-diabetic control) and SP-containing sensory terminals (Fig. 4e, f, p < 0.01 vs non-diabetic control).
Remarkable differences were observed in the tissue redistribution of HSPC populations following LI in non-diabetic mice compared with diabetic mice. In both groups, LI caused an initial expansion of LSK-HSPCs in the BM, and this was followed by a return to basal levels at 3–7 days post-LI and then a second peak at 14 days (Fig. 5a, b). The subpopulation of LSK-NK1R-HSPCs was reduced in the BM of non-diabetic mice during the acute phase of LI, returning to basal levels by day 7. In contrast, no change was observed in diabetic mice before and after LI (Fig. 5c, d). When looking at cellular changes in PB, we observed a striking increase in LSK-HSPCs and LSK-NK1R-HSPCs from day 1 to day 14 post-LI in non-diabetic mice, with this response being remarkably attenuated and delayed in diabetic mice (Fig. 5e–h, p < 0.01 vs non-diabetic mice). Additionally, the diabetic mice manifested a reduction in the homing of LSK-HSPCs and LSK-NK1R-HSPCs at the level of the ischaemic limb muscle, whereas no difference between groups was seen in the contralateral muscle (Fig. 5i–l, p < 0.01 vs non-diabetic mice).
We next investigated whether these deficits associate with alterations of nociceptor-related mechanisms. To this aim, we measured the levels of SP in BM supernatant fractions, PB and ischaemic muscles of diabetic and non-diabetic mice before and after LI. Two-way ANOVA detected an effect of time (p < 0.01) and group factor (p < 0.01) on the levels of SP in these different compartments; specifically, non-diabetic mice manifested a positive gradient of SP between muscle, PB and BM following LI, whereas this phenomenon was much attenuated in diabetic mice (Fig. 5m–o and ESM Fig. 3).
Since recruitment of LSK-NK1R-HSPCs has been shown to be essential for post-ischaemic vasculogenesis , we anticipated that the observed cellular defect may participate in depressing the spontaneous recovery from ischaemia. Consistently, the diabetic mice showed remarkable deficits in limb blood-flow recovery (ESM Fig. 4a, b, p < 0.01) and reparative capillary angiogenesis (ESM Fig. 4c, d, p < 0.001) when compared with non-diabetic mice.
Clinical studies showing that dysfunctional nociception contributes to defective G-CSF-induced mobilisation
Having shown that diabetes-induced nociceptive dysfunction contributes to depressed HSPC release in ischaemia, we next investigated the relation between pain, circulating levels of SP and CD34+ HSPC mobilisation following direct stimulation of BM with G-CSF.
A cohort of healthy volunteers received placebo or recombinant human G-CSF. Those receiving G-CSF were grouped according to the level of pain (graded as pain score) induced by the growth factor. Interestingly, individuals who experienced moderate to severe bone or back pain showed a significantly greater increase in PB CD34+ HSPC counts than those who reported no pain or mild pain (Fig. 6a, b). There was a significant direct correlation between the pain score and the increase in the number of CD34+ HSPCs induced by G-CSF (r = 0.36, p < 0.02; Fig. 6c).
We also re-analysed data from a trial of BM stimulation with human recombinant G-CSF in diabetic and non-diabetic individuals. Specifically, here we investigated the association of HSPC mobilisation and pain perception. In the whole cohort, the individuals who reported back or bone pain after G-CSF administration showed a significantly higher increase in PB CD34+ HSPCs than those reporting no pain (p < 0.01, Fig. 7a). Two-way ANOVA detected an inhibitory effect of diabetes (p < 0.0001) and an enhancing effect of pain (p < 0.01) on G-CSF-induced mobilisation, with no interaction between factors. Furthermore, in diabetic patients, mobilisation was completely abrogated in the absence of pain (Fig. 7a) and there was no incremental effect of vascular complications on mobilisation in comparison with diabetic patients without vascular disease (Fig. 7b). To determine whether depressed nociceptive signals may account for the reduced mobilisation of HSPCs in diabetic patients, we measured PB SP concentrations before and 24 h after G-CSF injection. Interestingly, G-CSF administration caused an increase in SP levels in non-diabetic individuals (p < 0.05), with this response being abrogated in diabetic patients (Fig. 7c). Additionally, in diabetic patients with neuropathy and vascular complications, G-CSF stimulation resulted in a decrease in PB SP levels (Fig. 7d, p < 0.01 vs non-diabetic individuals, p < 0.05 vs diabetes without complications). We also found an association between changes in SP concentrations and the degree of CD34+ HSPC mobilisation in response to G-CSF. In fact, CD34+ HSPC mobilisation was significantly higher in individuals showing an increase in SP concentrations compared with those showing unchanged or decreased SP levels (Fig. 7e, p < 0.05).