Since the beginning of the SARS-CoV-2 pandemic, there has been a concern about the possibility of infection precipitating the new onset diabetes (1-5). It has been postulated that there may be a bidirectional relationship between COVID19 and diabetes, but it is unclear if that relationship can be directly attributed to loss of beta cell function after SARS-CoV-2 infection of cells, or due to indirect beta cell stress from increased insulin resistance, steroid treatment, and global inflammation. Several high profile reports have provided conflicting results about the presence of ACE2, TMPRSS2, and SARS-CoV-2 in the beta cells of infected patients (6-10). While these reports have been very informative, they are limited by the difficulty in acquiring patient tissue and that the available post-mortem tissues are limited to patients who have expired due to severe illness. Because most illness from SARS-CoV-2 infection is not severe, it is critical to identify a reproducible model system to study the effects of SARS-CoV-2 infection on beta cells during and after the disease course.
During the SARS-CoV-2 pandemic, non-human primates have proven to be a consistent, robust, and reproducible model for studying COVID19 disease pathophysiology and for preclinical evaluation of vaccines and therapeutics (11-25). It has been reported that rhesus macaques infected intranasally/intratracheally with SARS-CoV-2 show a mild to moderate disease pathology consistent with the overwhelming majority of human COVID19 cases. We capitalized on our access to this model system to evaluate the effects of mild/moderate COVID19 disease pathogenesis on beta cells in vivo. We obtained the pancreas from adult Rhesus Macaques inoculated with SARS-CoV-2 and of uninfected adult macaques collected at necropsy. We evaluated cellular histology, subcellular ultrastructure, and metabolic signatures to assess if SARS-CoV-2 infected beta cells in vivo and whether SARS-CoV-2 infection resulted in aberrant cellular pathology characteristic of functional beta cell impairment.
SARS-COV-2 infects the beta cells of NHP in vivo
As previously reported, adult rhesus macaques (6 to 12 years of age) were inoculated with 1.1 × 106 plaque-forming units (PFU) of SARS-COV-2 administered as 1 ml intranasally and 1 ml intratracheally (13, 15, 16). In this model, viral RNA levels peak at 2 days post inoculation (dpi). Interstitial viral pneumonia is present and resolves by around 4 dpi. Clinical disease course is resolved by around 12dpi (16). To assess if there was an acute effect of SARS-COoV-2 infection on beta cells and if it resolved by the end of the disease course, we evaluated the pancreas collected at necropsy during the acute phase (7-10dpi, n=3) and in the post-acute phase (14dpi, n=4) of the disease course. The pancreas from 3 uninfected adult macaques and 2 pregnant macaques infected with the Zika virus served as controls (26).
Because there have been conflicting reports about the robustness of ACE2 and TMPRSS2 expression in human islets, we sought to identify if either transcript was present in the islets of rhesus macaques. We interrogated a publicly available single cell sequencing data set from multiple organs of the rhesus macaque (27). Cells were clustered into organ specific clusters, and the both ACE2 and TMPRSS2 were present in the pancreas cluster. We then clustered cells into pancreatic cell subtypes. ACE2 and TMPRSS2 expression was highest in beta cells (Supplemental Figure 1A). We confirmed ACE2 expression in beta cells by immunohistochemistry. ACE2 expression was low but present in most beta cells; a subset of cells exhibited robust expression (white arrows, Supplemental Figure 1B).
After establishing that ACE2 was present in the beta cells of this system, we used immunohistochemistry to detect the SARS-COV-2 nucleocapsid protein in the islets of acute or post-acute pancreas. Because SARS-CoV2 RNA is no longer detectable in the bronchioalveolar fluid of post-acute subjects by Day 14, it was unclear if the infection would be present only during the acute phase or would resolve by the post-acute phase (16). Islets from acute and post-acute pancreas were positive for the nucleocapsid protein (Figure 1A). Controls tissues were negative for nucleocapsid protein expression. To confirm the presence of the virus in acute and post-acute islet cells, we used transmission electron microscopy to assess if viral particles were present in beta cells from inoculated subjects. Beta cells were identified by the characteristic halos around secretory granules. Viral particles were present in 4 of the 4 samples assessed by electron microscopy. An active viral replication complex was also present and contained structures that were representative of multiple stages of viral particle assembly (Figure 1B, red box and red arrows) (28). We noted characteristic double membrane vesicles inside of the profusion replication complex are also hallmarks of SARS-CoV2 infection we also present (Figure 1B, blue box and blue arrows) (29).
SARS-CoV2 infection drives a massive loss of beta cell mass
Previous reports suggested that SARS-CoV infection may cause beta cell injury, and other reports have suggested that certain viral infections can cause beta cell loss (30-34). We quantified fractional beta cell area in the control and infected NHP pancreas to determine if SARS-CoV-2 infection resulted in beta cell loss. Tissue was collected from the head, body, and tail of the pancreas and beta cell area was quantified as insulin+ pixels divided by total tissue area pixels. Representative images are shown in Figure 2A. Pregnant Zika-infected macaques were excluded from this analysis because pregnancy drives temporary increases in beta cell mass (35, 36). Total beta cell area from acute and post-acute pancreas beta cell area averaged approximately 1.8%, while total beta cell area in control pancreas was approximately 3.8% (n=3-4, p<0.05, Figure 2B). Because it was not clear if the loss in beta cell area was a result of decreased number or decreased size of beta cells (37), we measured the proportion of beta and alpha cells per islet. We found that the percentage of beta cell area per islet did not change (Figure 2C), which suggested that either beta cell atrophy or pan-islet apoptosis could have been driving this phenotype. Neither control, acute, or post-acute beta cells expressed cleaved caspase-3, a marker of apoptosis (data not shown). To measure cellular atrophy, we measured individual beta cell area in each subject (n=300 cells per group). Cellular boundaries were marked with ß-actin staining. We used Image J to trace and calculate individual cell size. Mean beta cell size decreased by 18% in the acute phase when compared to control and by 29% when comparing the post-acute beta cells to those from controls. Within inoculated subjects, individual beta cell size continued to decreased between the acute and the post acute phase, suggesting that SARS-CoV-2 infection continued to drive beta cell atrophy after disease resolution. We measured fasting serum insulin and glucose levels prior to necropsy in a small number of subjects (Supplemental Figure 2). Control subjects (n=2) had very low fasting glucose levels. 4 of 8 inoculated animals had glucose levels 60mg/dL, which has been characterized as dysmetabolic (metabolically normal <60mg/dL, dysmetabolic 60-100mg/dL, diabetic >100mg/dL) (38). Of those 4 animals, 3 also had elevated serum insulin levels (>45mU/ml, dark gray bars).
SARS-CoV2 infection induces subcellular ultrastructure indicative of diminished beta cell function
Because primates infected with SARS-CoV2 are restricted to BSL3 restricted facilities and these subjects were all participants in other studies, we were not able to pursue in vivo beta cell function studies, such as glucose tolerance tests or hyperglycemic/euglycemic clamp studies, to measure how SARS-CoV2 may affect beta cell function. To maximize information we can discern from the tissues we have available, we examined beta cell ultrastructure to assess if the viral infection induced any ultrastructural markers of beta cell stress or dysfunction.
Close examination of beta cells during individual beta cell size measurement revealed that atrophied beta cells had a degranulated appearance. Degranulation has been proposed to be a driving cause of beta cell deficits in the context of metabolic stress (39, 40). To discern if atrophied beta cells from SARS-CoV2 inoculated pancreas were degranulated, we visualized granular density using super-resolution fluorescence microscopy. Insulin granules were evenly distributed and filled most of the beta cell cytoplasm in control tissues (Figure 3A-A’’). In both acute and post-acute beta cells, insulin granules were concentrated in speckles and large areas of the cytoplasm were devoid of insulin granules (Figure 3A-A’, purple boxed insets and white arrows). We used ImageJ to determine the density of granules per square um of insulin area in an islet, then used our previous measurements of beta cell size (Figure 2) to estimate how many granules were present per beta cell. We measured a 66% decrease in insulin granularity between control and inoculated islets (n=10-20 islets per condition, p<0.001). Within the inoculated samples, there was no difference in granularity between the acute and post-acute time points.
To more closely examine the subcellular ultrastructure of the cytoplasm, we imaged control, acute, and post-acute beta cells by transmission electron microscopy. Beta cells from control pancreas exhibited normal ultrastructure, including dense insulin granulation, dense mitochondria, compact endoplasmic reticulum, and minimal vacuolization (Figure 4A). During the acute phase, we detected beta cells that were less electron dense than surrounding cells. These cells had increased vacuole-like spaces, dilated endoplasmic reticulum, and distended cristae within the mitochondria (Figure 4). In less electron dense cells, convoluted membranes predominated the cytoplasms and mitochondria membranes were disrupted. These hallmark attributes mirror observations in beta cells that are undergoing metabolic stress (41-43).
SARS-CoV2 infection shifts markers of cellular metabolism toward a glycolytic profile
The ultrastructural evidence of beta cell stress, and specifically of mitochondria disruption, raised the possibility these SARS-CoV2 could induce changes in beta cell metabolism. Recent reports concluded that SARS-CoV2 infection in shifts cells towards a more glycolytic metabolism to provide building blocks for viral replication (44). It has been argued that beta cells from Type 2 diabetic patients show a shift in cellular metabolism from oxidative phosphorylation towards glycolysis and that shifts in towards glycolysis can decrease insulin secretion (45). We sought to measure cellular metabolism of beta cells in fixed pancreas from control and SARS-CoV2 inoculated animals.
We developed a novel method to use fluorescence lifetime imaging (FLIM) to measure the levels of NADH in formalin-fixed paraffin embedded beta cells as a proxy measurement of cellular metabolism. FLIM measures the lifetime of excited NADH: unbound NADH exhibits short lifetimes (τ = 0.4 ns) and is a byproduct of glycolysis; enzyme-bound NADH exhibits a far longer lifetime (τ =1.2 – 3.7 ns (46) dependent on the bound enzyme and is a substrate of oxidative phosphorylation (47, 48). This ~10x difference allows FLIM to offer a measure of the glycolytic vs oxidative status of a cell that persists even after fixation and histological processing. We used immunohistochemistry to identify insulin producing cells on slides, then used a 2-photon laser to collect the lifetimes of the secondary antibody for insulin and the autofluorescence of NADH. The distribution of NADH lifetimes within the beta cells of an image is represented on the phasor plot. The closer the phasor plot’s centroid is to the relative position of free NADH on the circle, the higher the glycolytic metabolism of the cell. Figure 5A presents insulin masks, NADH intensity masks, and phasor plots from three representative islets. To capture the glycolytic vs oxidative status of each islet, we averaged the modes of the centroids for each islet per individual pancreas, and plotted the coordinates on a 2D plot (Figure 5B; n=10 islets from 3-5 pancreas per experimental group). We observed that the centroid plots for the acute and post-acute samples clustered separately from the control samples.
We next sought to understand if the separation of the sample populations on the phasor plots represented a change in beta cell metabolism. We calculated each islet’s glycolytic coefficient to report the proportion of NADH from glycolysis and identify cells as primarily glycolytic or primarily oxidative (49). Using this estimation, a higher glycolytic coefficient would represent more free NADH in the islet, suggestive of more glycolytic metabolism. We found that both uninfected control samples and zika-infected samples had a similar glycolytic coefficients (Figure 5C, pregnant- blue triangles, non-pregnant- blue circles). There was a 23% increase in the glycolytic coefficient in islets from the acute pancreas, suggesting that these cells employed a more glycolytic metabolism. Islets from the post-acute pancreas had a slightly lower glycolytic coefficient that was still significantly different from control samples. This indicates that beta cell metabolism may begin to recover in the post-acute phase of COVID19 pathogenesis. Because our study ended 14 days after infection, we were unable to measure if and when beta cell metabolism could return to baseline.