Islet endocrine cell composition among individuals with and without type 1 diabetes

A total of 26,376 individual islets were identified across the 38 individuals with or without type 1 diabetes included in this study. The number and proportion of islets comprising each endocrine cell composition (INS+GLU−, INS+GLU+ and INS−GLU+) are shown as a mean per group in Table 1 and for individual donors in ESM Table 5. The density and proportion of islets comprising each endocrine cell composition, and separated by islet bin size, are shown in ESM Fig. 4. These data highlight a marked reduction in both the density and proportion of small-to-medium insulin-containing (INS+GLU− and INS+GLU+) islets for individuals with type 1 diabetes. This is consistent with recent findings that smaller insulin-containing islets are largely absent at onset of type 1 diabetes, and those islets that retain insulin are larger in size [35]. Our 2D data from individuals without diabetes align with recent observations made in adult pancreas using 3D imaging technologies, which highlight that a significant proportion (approximately 50%) of insulin-containing islets are devoid of glucagon [36].

Table 1 Mean number and percentage of islets comprising different endocrine cell compositions among individuals with and without type 1 diabetes from the EADB and nPOD collections

In line with the classification of individuals with type 1 diabetes into two groups based on age at diagnosis and therefore aggressiveness in beta cell destruction, the proportion of INS+GLU+ islets tended to be higher in individuals ≥13 years at diagnosis of type 1 diabetes compared with individuals <13 years at diagnosis of type 1 diabetes [mean ± SEM; 22.93±4.59% vs 10.54±3.53%, respectively; unpaired t test: t(17)=2.002, p=0.06]. This aligns with previous observations demonstrating a higher proportion of insulin-containing islets in individuals with T1DE2 compared with T1DE1 [29, 30]. The identification of islets according to the three defined endocrine cell compositions allowed us to investigate whether pMSCs were preferentially associated with islets of a particular endocrine cell composition.

An extensive 6-plex panel is required to accurately identify pMSCs in situ

MSCs share phenotypic markers with endothelial cells, pericytes, fibroblasts and stellate cells. Thus, no single phenotypic marker can distinguish MSCs from other stromal or perivascular cell types. To ensure the most efficient use of rare and finite human pancreatic tissue, we initially evaluated whether a simplified 3-plex panel (CD90+, CD105+, CD31−) could accurately phenotype pMSCs in situ, or whether a more comprehensive 6-plex panel was required. CD90 and CD105 are widely used as positive human MSC markers; however, they are also expressed by endothelial cells. To address this, we included the endothelial cell marker CD31 as a negative MSC marker, in line with the International Society for Cell and Gene Therapy (ISCT) minimal criteria for MSC characterisation. CD90 and CD105 are two of the three (CD90, CD105 and CD73) positive human MSC markers outlined by the ISCT for defining human MSC identity [27, 28]. Accordingly, the 6-plex panel included CD73 as an additional positive MSC marker, as well as CD34 (endothelial cell and haematopoietic stem cell marker) and CD45 (pan-immune cell marker) as additional negative MSC markers (CD73+, CD90+, CD105+, CD31−, CD34−, CD45−).

Whole pancreas MSC density across EADB and nPOD donors was markedly overestimated when phenotyped using a core 3-plex (CD90+, CD105+, CD31−), compared with the comprehensive 6-plex panel (CD73+, CD90+, CD105+, CD31−, CD34−, CD45−) [mean ± SEM; 123±19 vs 14±3 MSCs/mm2, respectively; t(74)=5.647, p<0.001]. This validates the importance of characterising pMSCs using a 6-plex panel, to ensure cell types of distinct but similar phenotypes are excluded from the pMSC population. Using the comprehensive 6-plex panel, a total of 53,375 pMSCs (CD73+, CD90+, CD105+, CD31−, CD34−, CD45−) were identified across the 38 individuals included in this study. The pMSC population defined by the comprehensive 6-plex panel was further analysed to determine whether their number, density and/or islet-protective phenotype was altered in type 1 diabetes.

Whole pancreas pMSC density is similar in individuals with and without type 1 diabetes

The density of pMSCs across the total pancreas area including acinar, islets, ducts and vessels was variable between individuals both with and without type 1 diabetes (range 0–63 and 1–94 pMSCs/mm2, respectively), as expected given the established migratory capacity of MSCs [19]. Together, the density of pMSCs across the whole pancreas area was similar for individuals with and without type 1 diabetes [14.70±3.98 vs 14.07±5.17 pMSCs/mm2, respectively; t(36)=0.0966, p=0.92]. When separated by age, total pancreas pMSC density was similar between individuals with or without type 1 diabetes [one-way ANOVA; F(3, 34)=0.9442, p=0.43]. Representative spatial plots showing the distribution of MSCs across the endocrine and exocrine pancreas are shown in ESM Fig. 5.

Detection of pMSCs in proximity to pancreatic islets

pMSCs, defined as immunopositive for CD90, CD73 and CD105 (Fig. 2a–d) and immunonegative for CD31, CD34 and CD45 (Fig. 2e–h), that were associated (either inside or within 10 μm of the outside of the islet) with islets (Fig. 2i, j) were identified (Fig. 2a–k, ESM Fig. 6). Across all 38 donors, 3796 pMSCs (7% of total pancreas MSCs) were associated with islets. A total of 1699 pMSCs at the islet periphery (within 10 µm of the outside of the islet) and 2097 intraislet pMSCs were identified. pMSCs appear to be wrapped around the islet periphery and display an elongated, spindle-shaped morphology, as expected (Fig. 2l).

Fig. 2

Phenotype and morphology of pMSCs at the islet periphery. pMSCs (shown by white arrows, a–l) at the islet periphery were identified in HALO as CD90+ (a), CD73+ (b), CD105+ (c); overlay of positive markers (d); CD31− (e), CD34− (f), CD45− (g); overlay of negative markers (h). Islets were identified by insulin (i) and glucagon (j) immunostaining (inner white annotation) and the islet annotation expanded by 10 µm or until another annotation was reached (outer white annotation). Next, pMSCs were quantified using the HALO HighPlex FL package (k). A magnified micrograph of (k) is shown in (l) (grey box shows magnified area), demonstrating that pMSCs appear to be wrapped around the islet and display an elongated spindle-shaped morphology. Micrographs used in the representative figure were adjusted to optimise contrast and visibility without altering the underlying image data or quantification. Adjustments were made to the ‘Black In’, ‘White In’ and ‘Gamma’ settings. Donor ID: individual ≥13 years at type 1 diabetes diagnosis, 6563, nPOD

Islets that had an associated pMSC were defined as islets with one or more pMSCs either within 10 µm of the islet periphery or within the islet itself (intraislet). The percentage of INS+GLU+ islets that had one or more associated pMSCs was higher in individuals with type 1 diabetes (≥39.5%) either <13 years or ≥13 years at diagnosis, compared with respective individuals of similar age without diabetes (≤20.3%; ESM Fig. 7).

pMSC number and density are increased at the islet periphery of insulin-containing islets in individuals with type 1 diabetes

The number and density of pMSCs within a 10 µm region immediately outside the islet (islet periphery) were quantified (Fig. 3a–d). The density of pMSCs within 10 µm of insulin-containing (INS+GLU− and INS+GLU+) but not insulin-deficient (INS−GLU+) islets was higher in individuals with type 1 diabetes compared with individuals without diabetes (p<0.001 for INS+GLU− and INS+GLU+ islets, Fig. 3e). Accordingly, in type 1 diabetes the density of pMSCs at the periphery of insulin-deficient islets was dramatically reduced compared with islets containing residual beta cells (p<0.001), consistent with an islet-protective role for pMSCs. The density of pMSCs at the periphery of INS+GLU+ islets was higher in individuals who were ≥13 years at diagnosis of type 1 diabetes compared with those <13 years at diagnosis (p<0.001, Fig. 3f). When separated by islet size (Fig. 3g, h), the more pronounced increase in pMSC number at the periphery of islets in individuals ≥13 years compared with <13 years at diagnosis of diabetes was true for larger islets containing ~64–128 or ~265–512 cells (islet bin size 6 or 8; Fig. 3h).

Fig. 3

The density and number of pMSCs within a 10 µm region at the periphery of islets comprising different endocrine cell compositions and size. Representative quantifications of pMSCs (HighPlex FL package, HALO) at the islet periphery are shown in (a–d) (pMSCs highlighted by white arrows). For (e, f), data were derived from n=26,376 individual islets from 38 individuals (n=8 individuals <13 years at type 1 diabetes diagnosis and n=11 individuals ≥13 years at type 1 diabetes diagnosis [n=19 individuals with type 1 diabetes]; n=8 individuals <13 years without diabetes and n=11 individuals >13 years without diabetes [n=19 individuals without diabetes]). For (g–i), islet bin size containing <5 islets or bin sizes devoid of data for one or more groups were excluded; data were derived from n=23,290 individual islets. Bars represent mean ± SEM. Two-way ordinary ANOVA with post hoc tests adjusted for multiple comparisons with Bonferroni correction. Post hoc tests in (e): islets of the same endocrine cell composition; diabetes or no diabetes across islets of different endocrine cell compositions. Post hoc tests in (f): type 1 diabetes group and individuals of similar age without diabetes; type 1 diabetes and age of diagnosis; individuals without diabetes. Comparisons in (e, f) are shown with a horizontal black line between the two compared groups. Post hoc comparisons within each islet bin size in (g, h): a<13 years at type 1 diabetes diagnosis vs ≥13 years at type 1 diabetes diagnosis; b<13 years at type 1 diabetes diagnosis vs <13 years without diabetes; c≥13 years at type 1 diabetes diagnosis vs ≥13 years without diabetes; d<13 years without diabetes vs ≥13 years without diabetes. ***p≤0.001; **p≤0.01; *p≤0.05. Donor IDs: nPOD 6407 (a); nPOD 6578 (b); nPOD 6333 (c); nPOD 6362 (d). <13 y ND, <13 years without diabetes; <13 y T1D, <13 years at type 1 diabetes diagnosis; ≥13 y ND, ≥13 years without diabetes; ≥13 y T1D, ≥13 years at type 1 diabetes diagnosis; ND, no diabetes; T1D, type 1 diabetes

In individuals without type 1 diabetes, pMSC density was also higher at the periphery of insulin-containing islets in individuals ≥13 years compared with individuals <13 years (Fig. 3f). Nonetheless, the density of pMSCs at the periphery of insulin-containing islets was increased to a greater extent in individuals with type 1 diabetes irrespective of the age at diagnosis when compared with individuals of similar age without diabetes (Fig. 3f). Irrespective of age or type 1 diabetes diagnosis, the mean number of pMSCs at the periphery of islets increased with islet size for insulin-containing (two-way ANOVA main effect islet bin size, p<0.001 for both) but not insulin-deficient (two-way ANOVA main effect islet bin size, p=0.61) islets (Fig. 3g–i). Thus, the number of pMSCs at the islet periphery increased with increasing numbers of beta cells.

Intraislet pMSC density is increased within insulin-containing islets in individuals with type 1 diabetes

In accordance with our quantification of pMSCs at the islet periphery (Fig. 3a–f), the density of intraislet pMSCs within insulin-containing but not insulin-deficient islets was higher in individuals with type 1 diabetes compared with individuals without diabetes (p<0.001 for INS+GLU− and INS+GLU+ islets, Fig. 4a–f). However, there did not appear to be a more pronounced intraislet pMSC density in individuals ≥13 years at diagnosis of type 1 diabetes compared with <13 years at diagnosis when all islet sizes were grouped together. When separated by islet size (Fig. 4g–i), only the largest INS+GLU+ islets showed increased intraislet pMSC number in individuals ≥13 years compared with <13 years at diagnosis of type 1 diabetes. In agreement with our observations for pMSC number at the islet periphery, the mean number of intraislet pMSCs increased with islet size for insulin-containing islets (two-way ANOVA main effect islet bin size, p<0.001 for both).

Fig. 4

The density and number of intraislet pMSCs of islets comprising different endocrine cell compositions and size. Intraislet pMSCs (shown by white arrows, a–d) within INS+GLU+ islets are shown. Immunostaining for insulin (blue) and glucagon (green) outside of the analysed islets in (a–d) is shown. For (e, f), data were derived from n=26,376 individual islets from 38 individuals (n=8 individuals <13 years at type 1 diabetes diagnosis and n=11 individuals ≥13 years at type 1 diabetes diagnosis [n=19 individuals with type 1 diabetes]; n=8 individuals <13 years without diabetes and n=11 individuals >13 years without diabetes [n=19 individuals without diabetes]). For (g–i), islet bin size containing <5 islets or bin sizes devoid of data for one or more groups were excluded; data were derived from n=23,290 individual islets. Bars represent mean ± SEM. Two-way ordinary ANOVA with post hoc tests adjusted for multiple comparisons with Bonferroni correction. Post hoc tests in (e): islets of the same endocrine cell composition; diabetes or no diabetes across islets of different endocrine cell compositions. Post hoc tests in (f): type 1 diabetes group and individuals of similar age without diabetes; type 1 diabetes and age of onset; individuals without diabetes. Comparisons in (e, f) are shown with a horizontal black line between the two compared groups. Post hoc comparisons within each islet bin size in (g–i): a<13 years at type 1 diabetes diagnosis vs ≥13 years at type 1 diabetes diagnosis; b<13 years at type 1 diabetes diagnosis vs <13 years without diabetes; c≥13 years at type 1 diabetes diagnosis vs ≥13 years without diabetes; d<13 years without diabetes vs ≥13 years without. ***p≤0.001; **p≤0.01; *p≤0.05. Donor IDs: nPOD 6407 (a); nPOD 6578 (b); nPOD 6333 (c); nPOD 6362 (d). <13 y ND, <13 years without diabetes; <13 y T1D, <13 years at type 1 diabetes diagnosis; ≥13 y ND, ≥13 years without diabetes; ≥13 y T1D, ≥13 years at type 1 diabetes diagnosis; ND, no diabetes; T1D, type 1 diabetes

pMSCs express islet-protective factors including ANXA1

Isolated ‘exogenous’ MSCs secrete an array of cytoprotective, immunomodulatory and regenerative molecules in response to specific cues within their microenvironment to influence both islet cells and a wide range of immune cell subsets. In the current report we have determined whether the previously defined islet-protective and immunomodulatory factors, ANXA1 and IDO1, are expressed in human MSCs in vitro (ESM Fig. 8) and in pMSCs in the pancreas of individuals with and without type 1 diabetes (Fig. 5). The expression of IDO1 was induced in human MSCs in vitro when exposed to cytokine combinations reflective of type 1 diabetes (IFN-γ + IL-1β [Ct<18] and IFN-γ + IL-1β + TNF-α [Ct<17]) compared with the no cytokine control (Ct>30). ANXA1 expression remained constitutively high across all groups (Ct<21; ESM Fig. 8).

Fig. 5

IDO1 and ANXA1 immunostaining and quantification of ANXA1+ pMSCs. Representative INS+GLU+ islets are shown. Islets were identified by insulin and glucagon immunostaining (a–d). IDO1 (e–h) and ANXA1 (i–l) immunostaining was conducted and ANXA1+ MSCs were quantified using the HALO HighPlex FL package (pMSCs shown by white arrows, m–p). Each micrograph (a–p) shows the islet annotation (inner white annotation) and expanded annotation 10 µm from the islet periphery (outer white annotation). Donor IDs: nPOD 6407 (a); nPOD 6578 (b); nPOD 6333 (c); nPOD 6362 (d). <13 y ND, <13 years without diabetes; <13 y T1D, <13 years at type 1 diabetes diagnosis; ≥13 y ND, ≥13 years without diabetes; ≥13 y T1D, ≥13 years at type 1 diabetes diagnosis

A total of 327 IDO1+ MSCs were identified across all pancreas sections (total pancreas including acinar, islets, ducts and vessels). The percentage of pMSCs that expressed IDO1 across all 38 individuals with and without type 1 diabetes was 1.02±0.34%. The percentage of pMSCs that were IDO1+ was similar for individuals with or without type 1 diabetes, irrespective of age [<13 years at type 1 diabetes diagnosis: 2.06±1.22%; <13 years without diabetes: 0.84±0.54%; ≥13 years at type 1 diabetes diagnosis: 1.15±0.64%; ≥13 years without diabetes: 0.25±0.13%; one-way ANOVA, F(3, 34)=1.156, p=0.32]. The small number of pMSCs expressing IDO1 indicates that IDO1 is not the primary islet-protective mechanism employed by pMSCs. Further analysis was not conducted on IDO1+ pMSCs due to the low number of cells detected.

In contrast, a total of 16,825 ANXA1+ pMSCs were identified across all pancreas sections. The percentage of total pancreas pMSCs that expressed ANXA1 across all 38 individuals with and without type 1 diabetes was 33.2% and was similar across groups, irrespective of age or type 1 diabetes diagnosis [<13 years at type 1 diabetes diagnosis: 35.78±6.20%; <13 years without diabetes: 27.38±5.08%; ≥13 years at type 1 diabetes diagnosis: 35.67±5.80%; ≥13 years without diabetes: 33.02±5.32%; one-way ANOVA, F(3, 34)=0.1867, p=0.73]. Of the total pancreas ANXA1+ pMSCs identified, 7.45% were associated with islets. Specifically, we detected 663 (3.94%) ANXA1+ intraislet pMSCs and 590 (3.51%) ANXA1+ pMSCs at the islet periphery (within 10 μm).

The number and density of pMSCs expressing islet-protective ANXA1 at the islet periphery are higher in individuals with type 1 diabetes

For individuals with type 1 diabetes, the density of ANXA1+ pMSCs was increased at the periphery of insulin-containing but not insulin-deficient islets (Fig. 6a). When segregated according to age at type 1 diabetes diagnosis, the density of ANXA1+ pMSCs was enhanced at the periphery (within 10 µm) of INS+GLU+ islets for individuals ≥13 years compared with individuals <13 years at diagnosis (Fig. 6b). When further separated according to islet size (Fig. 6c–e), enhanced ANXA1+ pMSC number at the islet periphery of INS+GLU+ islets was apparent for the largest islets (islet bin size 8) containing over ~256 cells (Fig. 6d). The number of ANXA1+ pMSCs at the islet periphery increased with islet size for insulin-containing islets (two-way ANOVA main effect islet bin size, p≤0.002 for both). In agreement with total pMSCs (both ANXA1+ and ANXA1− pMSCs), this may suggest increased migration of ANXA1+ pMSCs to islets with an increasing number of beta cells.

Fig. 6

Number and density of pMSCs expressing islet-protective ANXA1 that are associated with islets. (a, b) and (f, g): data were derived from 26,376 individual islets from 38 individuals (n=8 individuals <13 years at type 1 diabetes diagnosis and n=11 individuals ≥13 years at type 1 diabetes diagnosis [n=19 individuals with type 1 diabetes]; n=8 individuals <13 years without diabetes and n=11 individuals >13 years without diabetes [n=19 individuals without diabetes]). (c–e) and (h–j): islet bin size containing <5 islets or bin sizes devoid of data for one or more groups were excluded; data were derived from n=23,290 individual islets. Bars represent mean ± SEM. Two-way ordinary ANOVA with post hoc tests adjusted for multiple comparisons with Bonferroni correction. Post hoc tests in (a) and (f): islets of the same endocrine cell composition; diabetes or no diabetes across islets of different endocrine cell compositions. Post hoc tests in (b) and (g): type 1 diabetes and individuals of similar age without diabetes; type 1 diabetes age of diagnosis; individuals without diabetes. Comparisons in (a, b) and (f, g) are shown with a horizontal black line between the two compared groups. Post hoc comparisons within each islet bin size in (c–e) and (h–j): a<13 years at type 1 diabetes diagnosis vs ≥13 years at type 1 diabetes diagnosis; c≥13 years at type 1 diabetes diagnosis vs ≥13 years without diabetes. ***p≤0.001; **p≤0.01. <13 y ND, <13 years without diabetes; <13 y T1D, <13 years at type 1 diabetes diagnosis; ≥13 y ND, ≥13 years without diabetes; ≥13 y T1D, ≥13 years at type 1 diabetes diagnosis; ND, no diabetes; T1D, type 1 diabetes

The density and number of intraislet pMSCs that express ANXA1 (Fig. 6f–j) mostly aligned with the data for total (both ANXA1+ and ANXA1− pMSCs) intraislet pMSCs within INS+GLU+ islets. An exception to this was that intraislet ANXA1+ pMSC density between individuals <13 years with and without type 1 diabetes was similar.

Cell types expressing ANXA1 within the pancreas

Cell types including MSCs, immune cells, endothelial cells, ductal cells, pericytes and fibroblasts can express ANXA1. Our panel of MSC markers enabled us to investigate ANXA1+ MSCs, immune (CD45+) cells and endothelial (CD31+) cells. Overall, the density of ANXA1+ cells across the whole pancreas was increased in individuals with type 1 diabetes, compared with those without [1583±190 vs 899±81 ANXA1+ cells/mm2, respectively; t(36)=3.314, p=0.002].

The percentage of ANXA1+ cells that were MSCs (ANXA1+, CD73+, CD90+, CD105+, CD31−, CD34−, CD45−) was similar between individuals with and without type 1 diabetes [1.62±0.64% vs 1.76±0.77%, respectively; t(36)=0.1358, p=0.89].

The percentage of ANXA1+ cells that were immune cells (ANXA1+, CD45+) was increased in individuals with type 1 diabetes compared with those without [5.92±0.65% vs 3.73±0.71%, respectively; t(36)=2.282, p=0.03]. When separated by age, the percentage of ANXA1+ cells that were immune cells was similar between groups [<13 years at type 1 diabetes diagnosis: 5.49±0.45%; <13 years without diabetes: 3.40±1.09%; ≥13 years at type 1 diabetes diagnosis: 6.23±1.09%; ≥13 years without diabetes: 3.97±0.97%; one-way ANOVA, F(3, 34)=1.811, p=0.16].

The percentage of ANXA1+ cells that were endothelial cells (ANXA1+, CD31+) was also similar between individuals with and without type 1 diabetes [11.94±1.37% vs 8.82±1.01%, respectively; t(36)=1.832, p=0.08].

Whole pancreas and individual-islet immune cell infiltration

Total pancreas CD45+ cell number was quantified as an index of inflammation. As expected, when categorised by age and diabetes status, whole pancreas inflammation was increased in individuals with type 1 diabetes compared with individuals without (two-way ANOVA main effect diabetes, p<0.001). When separated by age, irrespective of diabetes diagnosis, total pancreas inflammation was similar (p=0.76).

An islet was characterised as inflamed if there were more than five associated immune cells. This was determined by quantifying the total number of CD45+ cells located either inside the islet or within 10 µm of the islet periphery (Fig. 7a–d) [37, 38]. Very few (0.15%) INS+GLU+ islets from individuals without diabetes were categorised as inflamed, as expected [38]. Therefore, at the individual-islet level, inflammation was characterised further for all INS+GLU+ islets from individuals with type 1 diabetes only. The proportion of INS+GLU+ islets that were inflamed was 53.6% (n=126 islets) and 13.6% (n=130 islets) for individuals <13 years or ≥13 years at diagnosis of type 1 diabetes, respectively. The mean number of immune cells associated with each inflamed islet was higher for individuals <13 years at diagnosis of type 1 diabetes, compared with ≥13 years at diagnosis of type 1 diabetes [mean ± SEM; 27.48±3.35 vs 15.84±1.82 CD45+ cells; unpaired t test; t(254)=3.078, p=0.002]. This is consistent with the increased insulitis and more aggressive destruction of beta cells reported to occur in individuals with T1DE1 compared with T1DE2 [29, 30]. Whilst the number of beta cells per INS+GLU+ islet was not different between age groups, inflamed islets contained more beta cells than islets not under obvious immune attack (Fig. 7e), suggesting that larger islets may be preferentially infiltrated by immune cells.

Fig. 7

pMSCs are preferentially associated with inflamed islets in individuals older than 13 years at type 1 diabetes diagnosis. The number of CD45+ cells associated with islets (that were either within 10 µm of the islet periphery or inside the islet; a, b) were quantified, and islets were classified as not inflamed (≤5 CD45+ cells; c) or inflamed (>5 CD45+ cells; d). Islets are shown in (a–d) as inner white annotation and islet annotations expanded by 10 µm are shown by the outer annotation. (e–i) Data were derived from 1191 individual islets from n=8 individuals <13 years at type 1 diabetes diagnosis and n=11 individuals ≥13 years at type 1 diabetes diagnosis. Bars represent mean ± SEM. Two-way ordinary ANOVA with post hoc tests adjusted for multiple comparisons with Bonferroni correction. Post hoc tests in (e–i): between type 1 diabetes age of diagnosis at the same level of inflammation; between inflammation status for the same age of diagnosis. Comparisons are shown with a horizontal black line between the two compared groups. Donor ID: 6578, nPOD (a–d). ***p≤0.001; **p≤0.01; *p≤0.05. <13 y T1D, <13 years at type 1 diabetes diagnosis; ≥13 y T1D, ≥13 years at type 1 diabetes diagnosis

pMSC density (ANXA1+ and ANXA1−) at the periphery of, but not within (intraislet), inflamed INS+GLU+ islets was increased in individuals ≥13 years at diagnosis of type 1 diabetes compared with islets that were not inflamed from the same group and compared with inflamed islets in individuals <13 years at diagnosis (Fig. 7f, g). The density of pMSCs that express ANXA1 at the periphery of inflamed INS+GLU+ islets displayed similar results to those of total pMSCs (ANXA1+ and ANXA1−; Fig. 7h). Intraislet ANXA1+ pMSC density was increased in inflamed INS+GLU+ islets of individuals ≥13 years compared with <13 years at type 1 diabetes diagnosis (Fig. 7i).

Viability and proliferation of cytokine-exposed human MSCs in vitro

We investigated the ability of MSCs to survive and proliferate under cytokine exposures of increasing ‘aggression’ and duration (Fig. 8a–d). MSC viability was reduced in response to the more aggressive cytokine cocktail (IFN-γ + IL-1β + TNF-α), compared with exposure to less aggressive conditions (IFN-α alone or IFN-γ + IL-1β) or to the no cytokine control following 3 or 7 day exposure (Fig. 8a, b). The percentage of MSCs that had divided over the course of the 3 day experiment was also reduced in response to the more aggressive cytokine cocktail compared with the no cytokine control (Fig. 8c).

Fig. 8

Viability and proliferation of cytokine-exposed human MSCs in vitro. Human MSCs were exposed to a cytokine condition of increasing aggression and viability was assessed after 3 (a) or 7 (b) days. The percentage of divided cells, as calculated by the FlowJo proliferation algorithm, was used to determine proliferation of CellTrace-stained cells after 3 (c) or 7 (d) days. Each symbol represents an independent experiment (n=5 for viability and n=3 for proliferation). Bars represent mean ± SEM. Human MSCs between passages 5 and 8 were used for all experiments. One-way ANOVA with post hoc tests adjusted for multiple comparisons with Bonferroni corrections. Comparisons in (a–d) are shown with a horizontal black line between the two compared groups. ***p≤0.001; **p≤0.01; *p≤0.05

Overall, exposure of MSCs to an aggressive cytokine combination led to increased cell death and reduced proliferation. This is particularly reflective of the cytokine exposure expected around islets in individuals <13 years at diagnosis of type 1 diabetes, where immune cell infiltration is more intense [29, 30]. This may indicate that pMSCs are less able to survive in situ at the islet periphery in individuals <13 years at diagnosis of type 1 diabetes.