Although most patients with type 1 diabetes (T1D) continue to produce small amounts of insulin decades after disease onset, very few β-cells persist within their pancreata. Consequently, the source of persistent insulin secretion within T1D remains unclear. We hypothesized that low-level insulin content within non-β-cells could underlie persistent T1D insulin secretion. We tested for low levels of insulin (insulinlow) within a large cohort of JDRF Network for Pancreatic Organ Donors With Diabetes (nPOD) human pancreata across a wide range of ages and T1D disease durations. Long exposures, high-throughput imaging, and blinded parallel examiners allowed precise quantification of insulinlow cells. Of note, abundant islet endocrine cells with low quantities of insulin were present in most T1D pancreata. Insulinlow islet abundance and composition were not influenced by age, duration of diabetes, or age of onset. Insulinlow islets also contained β-cell markers at variable levels, including Pdx1, Nkx6.1, GLUT1, and PC1/3. Most insulinlow cells contained abundant glucagon and other α-cell markers, suggesting that α-cells drive much of the insulinlow phenotype in T1D. However, pancreatic polypeptide, somatostatin, and ghrelin cells also contributed to the insulinlow cell population. Insulinlow cells represent a potential source of persistent insulin secretion in long-standing T1D and a possible target for regenerative therapies to expand β-cell function in disease.

Endogenous insulin secretion persists in long-standing type 1 diabetes (T1D), but the cellular mechanism of this phenomenon is poorly understood (1). The majority of patients with long-standing T1D continue to secrete small amounts of insulin, as measured by ultrasensitive C-peptide assays (25). Of note, endogenous insulin secretion can be detected in 79% of patients with long-standing T1D disease duration (>28.5 years) (5). Despite these clinical observations, the anatomical source and implications of persistent insulin secretion in long-standing T1D remain enigmatic (1).

Perspectives regarding the mechanism of persistent insulin secretion in long-standing T1D traditionally have been β-cell centric. β-Cells can be readily found in pancreata of new-onset T1D when islets exhibit insulitis (6). Solitary β-cells and a few β-cell–containing islets can persist in T1D pancreata years after diagnosis (68) and have even been reported in long-standing disease (9,10). A variety of mechanisms have been proposed to explain β-cell persistence in T1D (11). We recently analyzed β-cell turnover, duct-associated β-cell neogenesis, and β-cell transdifferentiation in T1D samples but found no evidence for ongoing β-cell generation (12). We also were interested in quantifying residual β-cells in T1D samples as a potential explanation for persistent insulin secretion in T1D (12), but β-cells were extremely rare in most samples from long-standing T1D. Indeed, no residual β-cells were detected in one-half of the T1D samples (23 of 47), consistent with previous JDRF Network for Pancreatic Organ Donors With Diabetes (nPOD) studies (1315). The near total absence of β-cells in most pancreata from individuals with long-standing T1D is challenging to reconcile with recent reports of low-level insulin secretion in T1D (25,16). It remains possible that T1D donors without residual β-cells might contain β-cells somewhere else in the pancreas, akin to a needle buried in a haystack. Without sifting through an entire human pancreas cell by cell to test for residual β-cells, this hypothesis is difficult to formally exclude. Alternatively, residual insulin secretion could be due to insulin production from non-β-cell sources. Thus, the cellular basis for persistent insulin secretion in long-standing T1D remains the subject of speculation and intensely controversial.

Some have suggested that β-cells might become dedifferentiated in acute diabetes. For example, Accili and colleagues (17) reported the presence of plentiful dedifferentiated β-cells within pancreata from patients with type 2 diabetes. Similarly, Butler and colleagues (18) reported that T1D is associated with increased numbers of chromogranin-positive, hormone-negative cells. We found, however, a novel population of highly proliferative α-cell–related islet endocrine cells to be present in equal abundance within nondiabetic and T1D human pancreata of adolescent and young adult organ donors (19). As a result, the paradox of persistent endogenous insulin secretion in long-standing T1D remains unsolved.

Very-low-level insulin can be reliably detected in diabetic rodent β-cells using special techniques. We previously observed degranulated β-cell low-level insulin in newly diabetic NOD mice (20). These studies used methods developed in our laboratory designed to detect low-level insulin in β-cells. Our approach can even detect β-cells that are severely degranulated. By combining infrared fluorophores with a high-efficiency detector optimized for infrared wavelengths, this technique minimizes pancreatic autofluorescence, thereby increasing the signal-to-noise ratio of rare β-cell antigens. In islets from four NOD mice newly diagnosed with diabetes, we observed very weak insulin staining and GLUT2 staining (20). These findings revealed degranulated β-cells in acutely diabetic samples that would not have been identified by conventional imaging approaches.

In the current study, we considered the novel hypothesis that low-level insulin content from non-β-cell sources might be present in T1D pancreata. We imaged T1D pancreata with our method to detect small amounts of insulin and found abundant cells in T1D islets that exhibit very low levels of insulin (insulinlow). These insulinlow cells were undetectable by standard imaging practices but readily detected when imaged with very long exposures (to gather ∼20-fold more light). Insulinlow islet cells were observed in almost every T1D pancreas, even those from donors with long-standing disease (≥50 years). Insulinlow islet cells contained low levels of other β-cell markers as well. However, many insulinlow islet cells also had abundant α-cell markers: glucagon, ARX, GC, and PC2. Some other insulinlow islet cells contained other islet hormones indicative of pancreatic polypeptide (PP), somatostatin (SS), and ghrelin cell phenotypes. Thus, α-, PP, SS, and ghrelin cells each may contribute to the insulinlow islet cell phenotype in T1D pancreata.

Human Pancreatic Samples

Paraffin-embedded pancreas sections were obtained from the JDRF nPOD with a waiver from institutional review board. Pancreata studied were based on availability. Tissues were processed through standardized operating procedures (www.jdrfnpod.org/for-investigators/standard-operating-procedures), and serum C-peptide was quantified by nPOD using an AIA-2000 analyzer (Tosoh). The assay was sensitive to <0.05 ng/mL.

Sample Population

Ten control samples (from 6 males and 4 females) and 55 T1D samples (from 29 males and 26 females) were studied. These were selected to include various ages from infants to older adults (0–89 years). The cohort included some individuals with long-standing diabetes (59–89 years of age with >50 years T1D). Recent onset was defined as ≤10 years’ T1D duration. Sample population details are listed in Supplementary Tables 1 and 2 and in our previous work (12,19).

Immunohistochemistry

Sections were incubated with primary antisera (Supplementary Table 3), and secondary antisera were conjugated to aminomethylcourmarin, Cy2, Cy3, or Cy5 (Jackson ImmunoResearch) and DAPI (Molecular Probes) (12). Primary antisera were (1:100) ARX (AF7068; R&D systems), C-peptide (Alpco), GAD65 (sc-5601; Santa Cruz Biotechnology), vitamin D–binding protein GC (ab65636; Abcam), ghrelin (H-031-77; Phoenix Pharmaceuticals), amidated GLP17–36 (ab26278; Abcam), GLUT1 (07-1401; Millipore), insulin A and B chains (gifts from Alpco confirmed by the company to specifically detect the insulin A chain and insulin B chain), Nkx6.1 (F55A12; Developmental Studies Hybridoma Bank), pancreatic polypeptide (18-0043; Invitrogen), prohormone convertase (PC) 1/3 (AB10553; Millipore), PC2 (sc-374140; Santa Cruz Biotechnology), Pdx1 (NBP2-38865; Novus), somatostatin (ab30788; Abcam), and synaptophysin (18-0130, Thermo Fisher Scientific, and AB6245, Abcam); (1:250) glucagon (ab8055 and ab10988; Abcam); and (1:1,500) insulin (A0564; Dako).

Islet Morphometry

Islet morphometry was analyzed with Volocity version 6.1.1 (PerkinElmer) (12,19). Axio Imager (Zeiss) with X-Y stage and Orca-ER digital camera (Hamamatsu) was used to acquire thousands of islet images, with tens of thousands of nuclei analyzed per sample (Supplementary Table 4). All visible islets within one pancreatic section per individual were imaged for insulin with long exposure imaging (≥20× shutter time of standard exposure). Acinar and ductal tissue images were captured as negative controls for insulin staining.

Blinded Study 1: Classifying Insulinlow Islets

Virtually all islets in a pancreatic section were identified by DAPI and imaged. Negative control images (nonislet containing) were inserted into image stacks for subsequent blinded classification. For classification of insulinlow cells, T1D islets were separated into those exhibiting strong, moderate, or no insulinlow (described in detail below). Two examiners blinded to disease duration or phenotype independently classified and quantified insulinlow phenotypes in T1D islets. Islets were defined as containing five or more islet endocrine cells.

Blinded Study 2: Calculating Insulinlow Islet Abundance

Most, if not all, T1D islets were imaged for synaptophysin, insulin, and Nkx6.1 for 10 control and 15 T1D pancreatic sections. Control islets were imaged with standard exposure; T1D islets were imaged with long exposure for insulin only. Blinded investigators quantified insulinlow islets as percentage of total synaptophysin-positive islets.

Insulinlow Characterization

Insulinlow cells were characterized using a panel of α-cell (glucagon, ARX, glucagon-like peptide 1 [GLP-1], PC2, and GC) and β-cell (insulin, Nkx6.1, Pdx1, GLUT1, GAD65, and PC1/3) markers.

Many T1D Islets Exhibit Low Levels of Insulin That Are Detectable Only With Long-Exposure Imaging

To characterize the developmental basis of persistent insulin secretion, we recently surveyed β-cell persistence and regeneration in a large cohort of T1D pancreata and controls (12). However, we found no detectable β-cells within 23 of 47 T1D pancreata and very few β-cells within the rest. On the basis of the paucity of phenotypically detectable normal β-cells, we considered the alternative hypothesis that persistent and low insulin secretion could derive from β-cells that have adopted an alternative phenotype or that insulin could be present within non-β-cells of T1D pancreata. Consequently, we adapted our existing high-throughput imaging method (20) to search for very-low-level insulin content (and other β-cell markers) within T1D pancreata.

We used very long exposure times without image offset or gain to detect rare insulin granules in T1D pancreata. Islets were identified as distinct clusters of DAPI-positive nuclei and captured insulin staining with standard and long exposures (Fig. 1A–L). With standard exposures, we were able to detect readily insulin in nondiabetic control pancreata (Fig. 1E), but we were unable to detect any insulin with equivalent exposures in the majority of T1D islets (Fig. 1F–H). We then imaged the same T1D islets with much longer exposures (20-fold) and, surprisingly, readily observed small amounts of insulin within most islet endocrine cells of T1D pancreata, even in samples obtained from individuals with ≥50 years of disease duration (Fig. 1J–L and Supplementary Fig. 1). Insulinlow cells were highly abundant in T1D pancreata (Fig. 1M).

Figure 1

T1D islets with low levels of insulin. T1D islets (from donors of various ages and diabetes duration) stained with DAPI (white) (AD) and insulin (white) (EL) showed low levels of insulin detectable only with long-exposure imaging. Islets (green dashed line) identified by cellular clusters of DAPI-stained nuclei (AD). Control (A, E, and I) and T1D (BD, FH, and JL) islets imaged at standard (AH) and long exposure (IL). Scale bar = 50 μm. M: Abundant insulinlow islets present in T1D pancreas from a patient with long-standing disease. Scale bar = 2.5 mm.

Figure 1

T1D islets with low levels of insulin. T1D islets (from donors of various ages and diabetes duration) stained with DAPI (white) (AD) and insulin (white) (EL) showed low levels of insulin detectable only with long-exposure imaging. Islets (green dashed line) identified by cellular clusters of DAPI-stained nuclei (AD). Control (A, E, and I) and T1D (BD, FH, and JL) islets imaged at standard (AH) and long exposure (IL). Scale bar = 50 μm. M: Abundant insulinlow islets present in T1D pancreas from a patient with long-standing disease. Scale bar = 2.5 mm.

We considered the alternative hypothesis that nonspecific staining could underlie insulin immunoreactivity in T1D islets. Consequently, we stained control and T1D pancreata with either primary insulin antibody or secondary antibody alone followed by imaging with extra-long exposure times (∼2.5-fold longer than the very long exposures used to detect insulinlow cells throughout the study). However, we found no evidence of nonspecific staining with primary insulin or secondary antibody alone in islets, even with the very long exposure times (Supplementary Fig. 2A–H). As previously, combined insulin primary and appropriate secondary antisera resulted in intense insulin staining in control and low-level insulin staining in T1D islets, although the insulin staining intensity was exaggerated by the very long exposures (Supplementary Fig. 2I–L).

Insulin A Chain, Insulin B Chain, and C-Peptide in T1D Islets

To further interrogate the specificity of low-level insulin staining in T1D islets, we stained control and T1D pancreata with monoclonal antisera specific to various proinsulin cleavage products (insulin A chain, insulin B chain, or C-peptide). In control pancreata, insulin A and B chain and C-peptide antibodies detected cytoplasmic islet endocrine cell antigens at standard exposures (Fig. 2). We also detected low-level immunoreactivity with insulin A and B chain and C-peptide antisera in T1D islets. Islet T1D staining was present with long exposures times (20-fold) but not by standard exposures with insulin fragment antisera. Moreover, proinsulin cleavage product antisera exhibited cytoplasmic islet staining within pancreata of varying disease duration. These observations confirm our initial studies with polyclonal insulin antisera (Fig. 1 and Supplementary Figs. 1 and 2) and collectively support the notion that T1D islets contain low-level insulin immunoreactivity. We considered the possibility that proinsulin also might be detected in T1D islets in accordance with a recent publication by Wasserfall et al. (15). However, proinsulin was detected only in control or T1D samples with residual β-cells with intense insulin staining but never in insulinlow cells (data not shown).

Figure 2

Insulinlow staining confirmed by differential insulin-specific antibodies. Control and T1D islets stained for DAPI (white) and an insulin-specific monoclonal antibody: insulin A chain (white), insulin B chain (white), and C-peptide (white). Islets imaged at standard (A, B, D, E, G, H, J, K, M, N, P, and Q) and long (C, F, I, L, O, and R) exposures. Insulin A chain, insulin B chain, and C-peptide were detected at standard exposures in control islets (B, H, and N) unlike T1D islets (E, K, and Q). T1D insulinlow islets with low levels of insulin A chain, insulin B chain, and C-peptide captured only at long exposures (F, L, and R). Scale bar = 50 μm.

Figure 2

Insulinlow staining confirmed by differential insulin-specific antibodies. Control and T1D islets stained for DAPI (white) and an insulin-specific monoclonal antibody: insulin A chain (white), insulin B chain (white), and C-peptide (white). Islets imaged at standard (A, B, D, E, G, H, J, K, M, N, P, and Q) and long (C, F, I, L, O, and R) exposures. Insulin A chain, insulin B chain, and C-peptide were detected at standard exposures in control islets (B, H, and N) unlike T1D islets (E, K, and Q). T1D insulinlow islets with low levels of insulin A chain, insulin B chain, and C-peptide captured only at long exposures (F, L, and R). Scale bar = 50 μm.

Insulinlow Islet Composition Is Not Influenced by Age, Diabetes Duration, or Age of Onset

We quantified insulinlow islet composition in T1D pancreata by developing a system to classify insulinlow islets because the phenotype varied between islets and samples (Research Design and Methods). Two independent investigators classified islet images as strong, moderate, or no insulinlow (Fig. 3A–C). Insulinlow islets with strong low-level insulin showed intense, clearly identifiable cytoplasmic insulin with long exposures. Insulinlow islets with moderate low-level insulin were identified by medium intensity with diffuse cytoplasmic insulin staining with long exposures (Fig. 3B). Finally, no insulinlow islets did not contain detectable insulin but were clearly identified as collections of DAPI-positive cells in islets (Fig. 3C).

Figure 3

Insulinlow islet composition is not influenced by age, duration of diabetes, or age of onset. AC: Representative strong, moderate, and no insulinlow islets stained for insulin and imaged using long exposure. Scale bar = 50 μm. DF: Strong, moderate, and no insulinlow islets expressed as percentage of total islets analyzed vs. age (D), duration of diabetes (years) (E), and age of onset (F). Results demonstrate persistent insulinlow cells in individuals with T1D of all ages, duration of diabetes, and ages of onset. G: Correlation of two independent investigators for classification of strong, moderate, and no insulinlow islets in a blinded study demonstrates concordant evaluations of insulinlow islets.

Figure 3

Insulinlow islet composition is not influenced by age, duration of diabetes, or age of onset. AC: Representative strong, moderate, and no insulinlow islets stained for insulin and imaged using long exposure. Scale bar = 50 μm. DF: Strong, moderate, and no insulinlow islets expressed as percentage of total islets analyzed vs. age (D), duration of diabetes (years) (E), and age of onset (F). Results demonstrate persistent insulinlow cells in individuals with T1D of all ages, duration of diabetes, and ages of onset. G: Correlation of two independent investigators for classification of strong, moderate, and no insulinlow islets in a blinded study demonstrates concordant evaluations of insulinlow islets.

We hypothesized that increasing age, diabetes duration, or age of onset might influence the insulinlow phenotype such that pancreata from older individuals, long-standing disease, or early onset might contain fewer insulinlow islets and less intense insulinlow immunoreactivity. However, examination of insulinlow islets in T1D pancreata revealed no association between insulinlow phenotype and age (Fig. 3A and D and Supplementary Table 5). Both young and older individuals exhibited considerable percentages of strong and moderate islets (up to 100% of islets analyzed). Similarly, diabetes duration (3 months–84 years) and age of onset (1–32 years) also did not affect the percentage of strong, moderate, or no insulinlow islets in T1D pancreata. Of note, a few samples from individuals with ≥50 years’ T1D duration were found to contain a substantial amount of strong insulinlow islets (Fig. 3E and F, Supplementary Fig. 3B and C, and Supplementary Table 5). Every T1D pancreatic section contained both strong and moderate insulinlow islets, with fewer to zero islets in the no insulinlow islet group (Fig. 3D–F, Supplementary Fig. 3, and Supplementary Table 5). Concordant scoring of strong, moderate, or no insulinlow islets by two investigators (each blinded to T1D disease duration or each other’s impression) further supported the imaging tools, and classification criteria quantified insulinlow islets as robust and reproducible (Fig. 3G and Supplementary Table 5). Thus, insulinlow phenotypes were not influenced by age, disease duration, or age of onset.

Insulinlow Islets Are Abundant in T1D, Irrespective of Donor Age, Disease Duration, or Age of Onset

We aimed to determine the abundance of insulinlow islets as a fraction of total islets in T1D pancreata. We imaged all islets in T1D pancreatic sections simultaneously with synaptophysin and insulin at long exposure in a representative subset of T1D samples (n = 25) (Fig. 4A–H and Supplementary Table 6). Insulinlow cells uniformly and strongly costained synaptophysin, a marker of secretory islet endocrine cells (19). Insulinlow islets were highly abundant in most T1D pancreata and accounted for ∼65% of total synaptophysin-positive islets, an assessment consistent between the blinded investigators (Fig. 4I and Supplementary Table 6). Although the abundance of insulinlow islets varied across T1D pancreata (33–99% of total islets), analysis of abundance by the two independent investigators was strongly correlated (Fig. 4I and J and Supplementary Table 6). We performed correlation analyses to determine whether insulinlow islet abundance was influenced by age, diabetes duration, or age of onset (Supplementary Fig. 4). However, insulinlow islet abundance was not affected by age: young and older individuals with T1D had comparable insulinlow islets (Supplementary Fig. 4A). Diabetes duration and age of onset were not associated with insulinlow islet abundance (Supplementary Fig. 4B and C).

Figure 4

Abundant insulinlow islets persist in long-standing T1D. AH: Insulinlow islets stained for insulin (long exposure) (AD) and synaptophysin (standard exposure) (EH). Scale bar = 50 μm. I: Insulinlow islet abundance expressed as percentage of total synaptophysin-positive (% total Syn+) islets. T1D pancreata across individuals vary in insulinlow islet abundance; on average, though, approximately two-thirds of T1D islets per individual contained low-level insulin. J: Correlation of insulinlow islets (% total Syn+ islets) as measured by two independent investigators demonstrates concordant evaluations of insulinlow islet proportion in T1D. K: Correlation of insulinlow islets (% total Syn+ islets) vs. islet cell area (% total). Insulinlow islet abundance is positively associated with islet cell area. L: Insulin-negative (Insulin) islets from T1D pancreata are smaller than control insulin-positive (Insulin+) islets and T1D insulinlow-positive (Insulinlow+) islets. Data points represent the mean cross-sectional area of control insulin+ or T1D insulin and insulinlow+ islets for each donor sample. Data are mean ± SEM. One-way ANOVA was performed. ***P ≤ 0.001.

Figure 4

Abundant insulinlow islets persist in long-standing T1D. AH: Insulinlow islets stained for insulin (long exposure) (AD) and synaptophysin (standard exposure) (EH). Scale bar = 50 μm. I: Insulinlow islet abundance expressed as percentage of total synaptophysin-positive (% total Syn+) islets. T1D pancreata across individuals vary in insulinlow islet abundance; on average, though, approximately two-thirds of T1D islets per individual contained low-level insulin. J: Correlation of insulinlow islets (% total Syn+ islets) as measured by two independent investigators demonstrates concordant evaluations of insulinlow islet proportion in T1D. K: Correlation of insulinlow islets (% total Syn+ islets) vs. islet cell area (% total). Insulinlow islet abundance is positively associated with islet cell area. L: Insulin-negative (Insulin) islets from T1D pancreata are smaller than control insulin-positive (Insulin+) islets and T1D insulinlow-positive (Insulinlow+) islets. Data points represent the mean cross-sectional area of control insulin+ or T1D insulin and insulinlow+ islets for each donor sample. Data are mean ± SEM. One-way ANOVA was performed. ***P ≤ 0.001.

Insulinlow Cells in Large Islets

To study the morphology of insulinlow islets, we quantified insulinlow islets, islet endocrine mass, and islet size in a representative subset of age-matched control and T1D pancreata. T1D islet cell area was correlated with insulinlow islet abundance: Individuals with T1D with the most islet cell area had the highest percentage of insulinlow islets (Fig. 4K and Supplementary Table 7). Islets from individuals without diabetes contained fully granulated β-cells. In contrast, only a few samples had occasional T1D islets with fully granulated β-cells. T1D islets without fully granulated β-cells were subdivided into insulinlow and insulin-negative subtypes (Supplementary Table 7). Consistent with previous studies (19), mean islet endocrine area of control islets was 220% of T1D insulin islets (Fig. 4L, Supplementary Figs. 57, and Supplementary Table 7). Of note, mean islet area for insulinlow islets was comparable to control islets. In contrast, mean islet area for insulinlow islets was 250% of insulin negative islets (Fig. 4L, Supplementary Figs. 57, and Supplementary Table 7). Taken together, these studies indicate that the insulinlow phenotype is more likely to be found in large islets and to correlate with total islet area, hinting that an islet size threshold exists for the insulinlow phenotype within T1D pancreata.

Insulinlow Cells Contain Other β-Cell Markers: Nkx6.1, Pdx1, PC1/3, and GLUT1

We considered the possibility that insulinlow cells might contain other β-cell identity factors, albeit at reduced levels equivalent to the low insulin in T1D islets (17). We imaged T1D islets for various β-cell factors and found that insulinlow cells variably contained the β-cell transcription factors Nkx6.1 and Pdx1 (Fig. 5A and B). Insulinlow cells also costained GLUT1 and PC1/3, which determine glucose sensing and insulin processing, respectively (Fig. 5C and D). In contrast, insulinlow islets did not stain for GAD65, which is present in virtually all human β-cells (Supplementary Fig. 8). Thus, insulinlow cells maintain some, but not all, signatures of β-cell identity.

Figure 5

Insulinlow islets contain β-cell markers. T1D islets were stained with Nkx6.1, Pdx1, PC1/3, GLUT1, and insulin. Insulinlow islets contain variable amounts of β-cell markers as follows: Nkx6.1 (A, B, and D), Pdx1 (B and C), PC1/3 (C), GLUT1 (D), and insulinlow (A, C, and D). Insets highlight T1D islets that contain Nkx6.1 and insulinlow (A), show variable costaining of Nkx6.1 and Pdx1 in a T1D islet (B), or contain PC1/3, Pdx1, and insulinlow (C). Not all cells in T1D insulinlow islets uniformly contain Nkx6.1, Pdx1, or PC1/3. However, most cells in T1D insulinlow islets contained GLUT1 (D). Scale bar = 50 μm.

Figure 5

Insulinlow islets contain β-cell markers. T1D islets were stained with Nkx6.1, Pdx1, PC1/3, GLUT1, and insulin. Insulinlow islets contain variable amounts of β-cell markers as follows: Nkx6.1 (A, B, and D), Pdx1 (B and C), PC1/3 (C), GLUT1 (D), and insulinlow (A, C, and D). Insets highlight T1D islets that contain Nkx6.1 and insulinlow (A), show variable costaining of Nkx6.1 and Pdx1 in a T1D islet (B), or contain PC1/3, Pdx1, and insulinlow (C). Not all cells in T1D insulinlow islets uniformly contain Nkx6.1, Pdx1, or PC1/3. However, most cells in T1D insulinlow islets contained GLUT1 (D). Scale bar = 50 μm.

We then tested for prohormone convertase in insulinlow cells. As above, β-cells in both control and T1D islets with residual β-cells had PC1/3 (Fig. 5C). Consistent with previous literature (21), α-cells in both control and T1D islets with residual β-cells preferentially contained PC2 (Supplementary Fig. 9A and B), but insulinlow islets costained PC2 and PC1/3, which was in sharp contrast to control islets (Supplementary Fig. 9C). Simultaneous detection of PC2 and PC1/3 supports the biphenotypic nature of insulinlow cells, in contrast to previous studies of human pancreata (21).

Insulinlow Cells Also Contain α-Cell Markers: Glucagon, ARX, GLP-1, and GC

We hypothesized that insulinlow cells could be derived from other islet endocrine cells. Because α-cells are the most frequent cell type in T1D islets, we wondered whether they could be insulinlow cells. We first tested whether α-cells in control pancreata exhibit low-level insulin secretion, imaging control islets with standard and long exposures, but the intense insulin staining of nondiabetic β-cells made it extremely difficult to definitively exclude any low-level insulin within typical β-cell–containing control islets (Fig. 1I). To counter this potential issue, we performed an extensive survey in control pancreata to search for the rare islets that entirely lacked β-cells and, therefore, mainly comprised α-cells (Fig. 6). However, we did not find any evidence of bihormonal cells (with insulin and glucagon) in the rare nondiabetic islets without β-cells (Fig. 6B). Control α-cells only stained the α-cell markers glucagon and ARX (Fig. 6C). Similarly, control β-cells only stained β-cell markers: insulin and Nkx6.1 (Fig. 6A). However, insulinlow cells in T1D consistently costained both α-cell markers (glucagon and ARX) and β-cell markers (Nkx6.1 and insulin) (Fig. 6E–H). Thus, some insulinlow cells seemed to be biphenotypic α-cells in T1D that contained low-level insulin.

Figure 6

Insulinlow islets contain α-cell markers. Islets were stained simultaneously with insulin, Nkx6.1, glucagon, and ARX. AD: Control α-cells do not contain insulin and Nkx6.1, and control β-cells do not contain glucagon and ARX. Insets emphasize distinct and specific staining of glucagon and ARX in α-cells (yellow boxes) and insulin and Nkx6.1 in β-cells (green boxes) within their respective cells. EH: However, insulinlow (InsLow) cells in T1D islets costain insulin, Nkx6.1, glucagon, and ARX. Insets (white) show an insulinlow cell costaining insulin, glucagon, ARX, and Nkx6.1, suggesting that insulinlow cells exhibit an α-β biphenotypic identity.

Figure 6

Insulinlow islets contain α-cell markers. Islets were stained simultaneously with insulin, Nkx6.1, glucagon, and ARX. AD: Control α-cells do not contain insulin and Nkx6.1, and control β-cells do not contain glucagon and ARX. Insets emphasize distinct and specific staining of glucagon and ARX in α-cells (yellow boxes) and insulin and Nkx6.1 in β-cells (green boxes) within their respective cells. EH: However, insulinlow (InsLow) cells in T1D islets costain insulin, Nkx6.1, glucagon, and ARX. Insets (white) show an insulinlow cell costaining insulin, glucagon, ARX, and Nkx6.1, suggesting that insulinlow cells exhibit an α-β biphenotypic identity.

To confirm that glucagon-staining α-cells exhibited the insulinlow phenotype, we tested with additional markers. GLP-1 can be detected in control human α-cells with antisera to amidated GLP17–36 (22). Similarly, the vitamin D–binding protein GC is an α-cell marker (23). As expected, nondiabetic α-cells stained with ARX, GLP-1, and GC (Supplementary Fig. 10), but these α-cell markers were distinctly absent in nondiabetic β-cells (Supplementary Fig. 10A and C). In contrast, ARX, GLP-1, and GC were readily detected in insulinlow cells of T1D islets (Supplementary Fig. 10B and D). These results further support the notion that α-cells contribute to the insulinlow phenotype. Moreover, the absence of low-level insulin in nondiabetic cells with ARX, GLP-1, and GC supports the T1D-specific nature of the insulinlow phenotype.

PP, SS, and Ghrelin Cells With Low Levels of Insulin

Although the presence of glucagon, ARX, and other α-cell markers in some insulinlow islets indicates an α-cell phenotype, we considered the possibility that PP, SS, and/or ghrelin cells also might contribute to insulinlow cells in T1D. In control pancreata, PP, SS, and ghrelin cells did not contain low-level insulin (Fig. 7A–C). However, some (but not all) PP and SS cells contained low-level insulin in T1D islets (Fig. 7D and E). Ghrelin cells were rare in both nondiabetic and T1D islets; still, a few ghrelin cells also contained low levels of insulin (Fig. 7F). These results suggest that any of the non-β islet endocrine cell types might exhibit the insulinlow phenotype in T1D.

Figure 7

Insulinlow phenotype within PP, SS, and ghrelin cells of T1D pancreata. Islets were stained with PP, SS, and ghrelin (magenta) and insulin (Ins) (white). Insets from top to bottom show islet endocrine hormone at standard exposure, insulin (standard, then long exposure), and combined. AC: Control islets do not exhibit low levels of insulin. PP, SS, and ghrelin cells do not exhibit low levels of insulin, as highlighted by insets for AC. DF: However, some PP, SS, and ghrelin cells in T1D islets show low levels of insulin. Insets show PP (A), SS (B), and ghrelin (C) cells with small amounts of insulin that were detected using long-exposure imaging. Scale bar = 100 μm.

Figure 7

Insulinlow phenotype within PP, SS, and ghrelin cells of T1D pancreata. Islets were stained with PP, SS, and ghrelin (magenta) and insulin (Ins) (white). Insets from top to bottom show islet endocrine hormone at standard exposure, insulin (standard, then long exposure), and combined. AC: Control islets do not exhibit low levels of insulin. PP, SS, and ghrelin cells do not exhibit low levels of insulin, as highlighted by insets for AC. DF: However, some PP, SS, and ghrelin cells in T1D islets show low levels of insulin. Insets show PP (A), SS (B), and ghrelin (C) cells with small amounts of insulin that were detected using long-exposure imaging. Scale bar = 100 μm.

Defining the cellular source of persistent low-level insulin secretion in individuals with T1D is critical to understanding the potential compensatory mechanisms to insulin deficiency. Here, we report the existence of low-level insulin (insulinlow) content in T1D islets devoid of fully granulated insulin-positive β-cells from 55 individuals with T1D of a wide range of ages and disease duration. Of note, insulinlow cells in T1D islets exhibited a mixed islet endocrine cell phenotype with both β- and α-cell markers. Non-β-cell islet endocrine cell types (α-, PP, SS, and ghrelin cells) seemed capable of exhibiting the insulinlow phenotype. In contrast, insulinlow islets were not present in nondiabetic pancreata. Thus, insulinlow cells often are present in T1D but not control islets. Insulinlow cells seem to represent a phenotypic adaption to a T1D environment. In the future, we hope to perform extensive surveys to determine whether the insulinlow phenotype is specific to T1D or more commonly found in pancreata from other forms of diabetes.

Our findings of insulinlow islets are supported by a number of unique strengths. We combined unusual imaging techniques (long exposures and no contrast) and overlapping blinded examination with a comprehensive panel of islet endocrine markers to conclusively reveal the cellular phenotype of insulinlow cells. We used redundant markers (insulin, Nkx6.1, etc.) to quantify robustly the relative abundance of insulinlow cells in T1D pancreata, confirming the insulinlow phenotype. Our studies also revealed that T1D insulinlow islet cells exhibit a mixed islet endocrine phenotype with classical α- and β-cell markers, albeit at differential hormone content. Of note, we did not see any evidence of dysfunction of α-, PP, SS, and ghrelin cells in T1D, except that they make minor amounts of insulin.

Although we cannot formally exclude the possibility that insulinlow cells are not fully dedifferentiated β-cells, many T1D insulinlow islets contain markers of mature α-cells, including glucagon, ARX, GLP-1, GC, and PC2. Our findings therefore suggest that α-cells might malfunction and/or compensate for extreme insulin deficiency in T1D. Many investigators have proposed α- to β-cell transdifferentiation as a primary response to T1D. Herrera and colleagues (24) have described how α- to β-cell transdifferentiation might occur in response to near total α- to β-cell ablation in mice. Similarly, de Koning and colleagues (25) reported an increased frequency of insulin-positive cells that also contain glucagon in type 2 diabetes. Combined with β-cell dedifferentiation studies of Accili and colleagues (17), these reports revealed previously unrecognized phenotypic plasticity in islet endocrine cells under diabetic conditions. Alternatively, insulinlow cells simply could be dedifferentiated β-cells that now express mature markers of α-, PP, SS, or ghrelin cells. This possibility cannot be formally discounted given the impossibility of lineage tracing in human tissues. Moreover, a recent report described a “bottom” (Btm) population of residual β-cells with lower granularity in NOD mice (26). The Btm phenotype appears to be distinct from our insulinlow cells, as the authors did not find any insulin-glucagon or insulin-SS dual-positive cells by flow cytometry or electron microscopy. We look forward to additional studies of these Btm cells with other islet endocrine cell markers to definitively test the role of Btm cells in the insulinlow phenotype. Similarly, we look forward to tests of the insulinlow hypothesis in mice or with murine xenograft studies with cadaveric human islets or stem cell–derived islet cells.

Although all islet endocrine cell types appear to contribute to the insulinlow phenotype in T1D, why some islets contain insulinlow cells and others do not is unclear. We find no relationship between T1D duration and the insulinlow phenotype. However, it remains possible that patients with recently diagnosed T1D might harbor a distinct pattern. However, the rarity of samples precludes formal testing of this hypothesis until more rare new-onset T1D samples can be obtained. Although differences in fixation theoretically could drive intrapancreatic variation, we find no association between insulinlow cells and various geographical locations within pancreata. Similarly, we find no association between insulinlow cells within ducts, lobes, or other pancreatic structures. Finally, we note the lack of obvious insulinlow islet clustering. This is in sharp contrast to lobular concentration of insulin-positive islets often observed in recent-onset T1D (12).

The mechanisms that regulate hormones in insulinlow cells remain unclear. Insulin in healthy β-cells is regulated by elevated glucose, incretins (GLP-1, glucose-dependent insulinotropic polypeptide), and various other factors. In contrast, glucagon is regulated by low glucose and amino acids, such as glycine. Whether β- or α-cell signaling pathways govern hormones (insulin or glucagon) in insulinlow cells is unknown. In the future, it might be possible to determine the mechanisms that regulate insulin production and secretion in insulinlow cells in order to augment insulin secretion within patients with T1D.

Although residual β-cells are believed to contribute to persistent insulin secretion in T1D, we previously found that T1D pancreata contain few, if any, residual β-cells. As such, insulin microsecretion cannot be explained by anatomical preservation of β-cells. Insulinlow islets may represent another source of potential insulin secretion. Although we tested the relationship between insulinlow cells and serum C-peptide, we found no clear correlation (data not shown), possibly because of the standard assays used by nPOD that could only detect C-peptide down to 0.05 ng/mL. In this regard, ultrasensitive C-peptide assays will help to advance knowledge regarding the relationship between insulin microsecretors and the insulinlow phenotype of T1D. We look forward to prospective autopsy studies with ultrasensitive C-peptide assays and extensive pancreatic assays. Finally, we speculate that studies to identify the molecular drivers of the insulinlow phenotype might ultimately contribute to novel pharmacological approaches that enhance endogenous insulin secretion in patients with T1D.

J.A.K. is currently affiliated with the McNair Medical Institute and McNair Interests, Houston, TX.

Acknowledgments. The authors thank the pancreas donors and their families. This research was performed with the support of nPOD, a collaborative T1D research project sponsored by JDRF International. Organ procurement organizations partnering with nPOD to provide research resources are listed at www.jdrfnpod.org/our-partners.php. The authors thank M. Yang, I. Kusmartseva, A. Pugliese, C. Wasserfall, M. Campbell-Thompson, and M. Atkinson from JDRF nPOD for generous and essential contributions to this work. The authors also thank A. Myers, M. Beery, and E. Verney from JDRF nPOD for helping with the acquisition of samples for this study.

Funding. This study was supported by funding from the Robert and Janice McNair Foundation.

Duality of Interest. J.A.K. advises for Lexicon and Sanofi. J.A.K. is a full-time employee of McNair Interests, a private equity group with investments in life science–related companies. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. C.J.L., A.C., and E.S. performed the experiments. C.J.L., A.C., E.S., A.R.C., and J.A.K. conceived and designed the experiments and analyzed the data. C.J.L., A.R.C., and J.A.K. wrote the manuscript. J.A.K. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 2018 JDRF nPOD 10th Annual Scientific Meeting, Hollywood, FL, 20–23 February 2018.

1.
Pietropaolo
M
.
Persistent C-peptide: what does it mean
?
Curr Opin Endocrinol Diabetes Obes
2013
;
20
:
279
284
[PubMed]
2.
Wang
L
,
Lovejoy
NF
,
Faustman
DL
.
Persistence of prolonged C-peptide production in type 1 diabetes as measured with an ultrasensitive C-peptide assay
.
Diabetes Care
2012
;
35
:
465
470
[PubMed]
3.
Oram
RA
,
Jones
AG
,
Besser
RE
, et al
.
The majority of patients with long-duration type 1 diabetes are insulin microsecretors and have functioning beta cells
.
Diabetologia
2014
;
57
:
187
191
[PubMed]
4.
Davis
AK
,
DuBose
SN
,
Haller
MJ
, et al.;
T1D Exchange Clinic Network
.
Prevalence of detectable C-Peptide according to age at diagnosis and duration of type 1 diabetes
.
Diabetes Care
2015
;
38
:
476
481
[PubMed]
5.
Oram
RA
,
McDonald
TJ
,
Shields
BM
, et al.;
UNITED Team
.
Most people with long-duration type 1 diabetes in a large population-based study are insulin microsecretors
.
Diabetes Care
2015
;
38
:
323
328
[PubMed]
6.
MacLean
N
,
Ogilvie
RF
.
Observations on the pancreatic islet tissue of young diabetic subjects
.
Diabetes
1959
;
8
:
83
91
[PubMed]
7.
Gepts
W
,
De Mey
J
.
Islet cell survival determined by morphology. An immunocytochemical study of the islets of Langerhans in juvenile diabetes mellitus
.
Diabetes
1978
;
27
(
Suppl. 1
):
251
261
[PubMed]
8.
Foulis
AK
,
Stewart
JA
.
The pancreas in recent-onset type 1 (insulin-dependent) diabetes mellitus: insulin content of islets, insulitis and associated changes in the exocrine acinar tissue
.
Diabetologia
1984
;
26
:
456
461
[PubMed]
9.
Meier
JJ
,
Bhushan
A
,
Butler
AE
,
Rizza
RA
,
Butler
PC
.
Sustained beta cell apoptosis in patients with long-standing type 1 diabetes: indirect evidence for islet regeneration
?
Diabetologia
2005
;
48
:
2221
2228
[PubMed]
10.
Keenan
HA
,
Sun
JK
,
Levine
J
, et al
.
Residual insulin production and pancreatic ß-cell turnover after 50 years of diabetes: Joslin Medalist Study
.
Diabetes
2010
;
59
:
2846
2853
[PubMed]
11.
Wu
J
,
Yang
X
,
Chen
B
,
Xu
X
.
Pancreas β cell regeneration and type 1 diabetes (Review)
.
Exp Ther Med
2015
;
9
:
653
657
[PubMed]
12.
Lam
CJ
,
Jacobson
DR
,
Rankin
MM
,
Cox
AR
,
Kushner
JA
.
β cells persist in T1D pancreata without evidence of ongoing β-cell turnover or neogenesis
.
J Clin Endocrinol Metab
2017
;
102
:
2647
2659
[PubMed]
13.
Gianani
R
,
Campbell-Thompson
M
,
Sarkar
SA
, et al
.
Dimorphic histopathology of long-standing childhood-onset diabetes
.
Diabetologia
2010
;
53
:
690
698
[PubMed]
14.
Campbell-Thompson
M
,
Fu
A
,
Kaddis
JS
, et al
.
Insulitis and β-cell mass in the natural history of type 1 diabetes
.
Diabetes
2016
;
65
:
719
731
[PubMed]
15.
Wasserfall
C
,
Nick
HS
,
Campbell-Thompson
M
, et al
.
Persistence of pancreatic insulin mRNA expression and proinsulin protein in type 1 diabetes pancreata
.
Cell Metab
2017
;
26
:
568
575.e3
16.
Williams
GM
,
Long
AE
,
Wilson
IV
, et al
.
Beta cell function and ongoing autoimmunity in long-standing, childhood onset type 1 diabetes
.
Diabetologia
2016
;
59
:
2722
2726
[PubMed]
17.
Cinti
F
,
Bouchi
R
,
Kim-Muller
JY
, et al
.
Evidence of β-cell dedifferentiation in human type 2 diabetes
.
J Clin Endocrinol Metab
2016
;
101
:
1044
1054
[PubMed]
18.
Md Moin
AS
,
Dhawan
S
,
Shieh
C
,
Butler
PC
,
Cory
M
,
Butler
AE
.
Increased hormone-negative endocrine cells in the pancreas in type 1 diabetes
.
J Clin Endocrinol Metab
2016
;
101
:
3487
3496
[PubMed]
19.
Lam
CJ
,
Cox
AR
,
Jacobson
DR
,
Rankin
MM
,
Kushner
JA
.
Highly proliferative α-cell-related islet endocrine cells in human pancreata
.
Diabetes
2018
;
67
:
674
686
[PubMed]
20.
Sherry
NA
,
Kushner
JA
,
Glandt
M
,
Kitamura
T
,
Brillantes
AM
,
Herold
KC
.
Effects of autoimmunity and immune therapy on beta-cell turnover in type 1 diabetes
.
Diabetes
2006
;
55
:
3238
3245
[PubMed]
21.
Marchetti
P
,
Lupi
R
,
Bugliani
M
, et al
.
A local glucagon-like peptide 1 (GLP-1) system in human pancreatic islets
.
Diabetologia
2012
;
55
:
3262
3272
[PubMed]
22.
Zhang
Y
,
Wu
M
,
Htun
W
, et al
.
Differential effects of linagliptin on the function of human islets isolated from non-diabetic and diabetic donors
.
Sci Rep
2017
;
7
:
7964
[PubMed]
23.
Ackermann
AM
,
Wang
Z
,
Schug
J
,
Naji
A
,
Kaestner
KH
.
Integration of ATAC-seq and RNA-seq identifies human alpha cell and beta cell signature genes
.
Mol Metab
2016
;
5
:
233
244
[PubMed]
24.
Thorel
F
,
Népote
V
,
Avril
I
, et al
.
Conversion of adult pancreatic alpha-cells to beta-cells after extreme beta-cell loss
.
Nature
2010
;
464
:
1149
1154
[PubMed]
25.
Spijker
HS
,
Song
H
,
Ellenbroek
JH
, et al
.
Loss of β-cell identity occurs in type 2 diabetes and is associated with islet amyloid deposits
.
Diabetes
2015
;
64
:
2928
2938
[PubMed]
26.
Rui
J
,
Deng
S
,
Arazi
A
,
Perdigoto
AL
,
Liu
Z
,
Herold
KC
.
β cells that resist immunological attack develop during progression of autoimmune diabetes in NOD mice
.
Cell Metab
2017
;
25
:
727
738
[PubMed]
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at http://www.diabetesjournals.org/content/license.

Supplementary data