Non-human primate model of long-COVID identifies immune associates of hyperglycemia

Between 10 and 30% of people infected with SARS-CoV-2 develop long-term health complications (post-acute sequelae of SARS-CoV-2, PASC/Long-COVID). Metabolic diseases, including type 2 diabetes (T2D), as well as conditions with metabolic components such as ME/CFS, breathlessness, thrombosis, and neuropsychiatric sequelae, are part of the PASC spectrum. Compared to uninfected controls, individuals infected with SARS-CoV-2 have elevated risks and 12-month burdens of diabetes. A hyperinflammatory response to SARS-CoV-2 is implicated in acute COVID-19 severity and the development of metabolic PASC such as hyperglycemia, MAFLD, and cardiovascular diseases. Immune cell metabolic reprogramming is thought to balance viral survival and host responses, but how early immune/metabolic disruptions shape metabolic PASC remains unclear due to a lack of appropriate animal models. New-onset hyperglycemia and DKA also occur in people without pre-existing diabetes and are associated with poor outcomes. Glucose homeostasis involves hormonally regulated uptake and hepatic glucose production via gluconeogenesis/glycogenolysis, with potential contributions from pancreatic β-cell dysfunction. Circulating inflammatory mediators, including chemokines such as CCL25, have been linked to impaired glucose homeostasis. The study aims to establish a non-human primate model to interrogate how early immunometabolic disturbances during acute SARS-CoV-2 infection contribute to persistent hyperglycemia and to test whether early vaccination can modulate these outcomes.

The paper summarizes evidence that: (1) Long-COVID is common (10–30%) and includes metabolic sequelae such as T2D; (2) SARS-CoV-2 infection increases incident diabetes risk and 12-month burden compared to controls; (3) Hyperinflammation during acute COVID-19 drives severity and may promote metabolic PASC (hyperglycemia, MAFLD, cardiovascular disease); (4) Inflammatory cytokines/chemokines (e.g., CCL25) can impair β-cell insulin secretion and promote proinflammatory responses; (5) Pancreatic infection by SARS-CoV-2 and intrahepatic inflammation (e.g., HCV) have been linked to diabetes onset; (6) Elevated glucose favors SARS-CoV-2 infection and proinflammatory monocyte responses; (7) Hepatic infection by SARS-CoV-2 may stimulate gluconeogenesis and contribute to hyperglycemia; (8) Prior NHP work in AGMs showed pancreatic tropism and new-onset hyperglycemia, mirroring human observations. These studies underscore inflammation and organ-specific processes as contributors to dysglycemia after SARS-CoV-2 and motivate the development of an animal model to dissect mechanisms and test interventions.

Study design: Fifteen African green monkeys (Chlorocebus aethiops sabaeus; 13 females, 2 males; 7.92–19.32 years) were infected with SARS-CoV-2 strain 2019-nCoV/USA-WA1/2020 via intranasal (~1e6 TCID50; 0.5 mL/nares) and intratracheal (1 mL) routes. Animals were followed for 18 weeks with weekly clinical, virologic, blood chemistry, and immunometabolic assessments (some biweekly after week 4). Two groups: unvaccinated (n=10) and vaccinated (n=5) that received a single dose of BNT162b2 (Pfizer/BioNTech) on day 4 post-infection. One vaccinated female (PB24) underwent necropsy at week 8 due to anorexia. Morning blood (fasted overnight) was collected at baseline (11 days pre-infection) and at day 3, and weeks 1, 4, 8, 12, 16, and 18 post-infection. Historical pre-baseline glucose data every few months over ~3 years were also analyzed. Tissues (liver, pancreas, duodenum, lung) were collected at necropsy (18 weeks p.i.) for virologic, metabolic, and histopathologic analyses. Virology: Quantified nasal and pharyngeal swab genomic (gRNA) and subgenomic (sgRNA) SARS-CoV-2 RNA by qRT-PCR (TaqPath master mix, QuantStudio 6). Standard curves from in vitro transcribed RNA standards (10^10 to 10 copies). Primers/probes targeted genomic N and subgenomic N and E. Serology: Anti-SARS-CoV-2 IgA, IgG, IgM responses against spike, S1-RBD, and nucleocapsid measured using MSD S-PLEX assays per manufacturer’s instructions (plasma diluted 1:500). Data log10-transformed for longitudinal visualization with LOESS smoothing. Immune phenotyping and ex vivo stimulation: Longitudinal PBMCs were stimulated with PMA/ionomycin/brefeldin A/monensin for 6 h; stained for viability, surface markers (CD45, CD3, CD4, CD8, CD28, CD95, CD69, CD107a) and intracellular cytokines (IL-2, IL-4, TNF, IFNγ, IL-17A). Flow cytometry acquired on BD FACSymphony. Naïve/memory CD4+ T cells delineated by CD95/CD28. Targeted analyses of memory T-cell cytokine expression; additionally, data exported for unsupervised Spectre analysis (arcsinh transform; FlowSOM clustering; UMAP reduction; identification of 18 CD3+ T-cell populations; populations correlated with glucose and plasma analytes). Plasma proteomics (OLINK PEA): Olink Target 96 Inflammation panel measured plasma proteins (NPX units). QC applied; 65 analytes passed stringent QC. Differential expression assessed between baseline and week 1; BH FDR correction applied. STRING and PANTHER used for protein-protein interaction networks and functional enrichment; BioGRID used for validation of interactions. A confirmatory OLINK dataset including vaccinated and unvaccinated animals at BL and weeks 1–3 was used for GDNF. Insulin: Plasma insulin quantified by monkey insulin ELISA (AssayGenie), samples 1:8, measured in triplicates. Tissue viral persistence: RNAscope 2.5 HD RED ISH with anti-sense SARS-CoV-2 S probe on FFPE duodenum, liver, pancreas at 18 weeks p.i.; historical 4-week p.i. sections as comparators; lung sections used as positive controls. Quantification with HALO ISH algorithm. qPCR on tissues (n=14 at 18 weeks p.i.) assessed sgN, sgE (none detected) and genomic N (low-level in duodenum of 2 animals; negligible in lungs). SARS-CoV-2 immunohistochemistry performed on duodenum. Histopathology: Lungs (multiple lobes) and pancreas assessed by H&E for inflammation (graded 1–4), fibrosis, and pneumocyte II hyperplasia. Comparators included 3 uninfected and 4 infected animals with 3–4 week endpoints. Liver glycogen: Periodic acid-Schiff (PAS) staining with and without diastase to quantify hepatic glycogen in uninfected (n=3), short-term (4 weeks p.i.; n=4), and long-term (18 weeks p.i.; unvaccinated n=10, vaccinated n=5) groups. ImageJ (Fiji) used for intensity quantification (absolute and Δ mean intensity diastase-subtracted). Correlated with blood glucose at weeks 8 and 12. Statistics: Wilcoxon matched-pairs signed-rank tests for paired comparisons; Mann–Whitney U for unpaired; Spearman’s rank correlations; PERMANOVA for longitudinal group comparisons; BH FDR corrections applied where indicated. PCA and heatmaps in R (ggplot2).

• Virology and antibodies: All animals had detectable nasal and pharyngeal sgRNA at day 3 (>3.64×10^6 VC/ml); by week 5, sgRNA largely declined (detected in nasal swabs of only 2 animals). gRNA peaked at day 3 and declined by week 5; at week 5, 20% of vaccinated and 60% of unvaccinated animals had detectable pharyngeal gRNA. Robust IgG and IgA responses to spike and S1-RBD peaked between weeks 3–6 and were maintained to 18 weeks; nucleocapsid responses were also induced. No significant differences in antibody magnitudes between groups over time.
• Hyperglycemia: Serum glucose rose significantly by day 3 (mean 102.2 mg/dL; range 76–154) vs latest baseline (mean 71.1 mg/dL; range 51–86; p<0.0001), exceeding reported normal AGM ranges. Unvaccinated animals had significantly higher glucose longitudinally than vaccinated (PERMANOVA p=0.001), with a higher proportion of readings >100 mg/dL and persistence of hyperglycemia to at least week 12. Animals with higher preceding glucose (week 2) had lower peak IgA to nucleocapsid and S1-RBD at week 3. Historical pre-baseline glucose values over ~3 years were stable; post-infection, 9/13 timepoints differed from pre-baseline in unvaccinated vs 2/13 in vaccinated.
• Lipids: Triglycerides increased early (significant by week 1 in unvaccinated; overall higher in unvaccinated over time, PERMANOVA p=0.001) but were not consistently elevated beyond baseline across all timepoints. Total cholesterol showed minimal changes (significant only BL to week 1).
• Early inflammatory signature: OLINK PEA (n=9 unvaccinated) identified 15 analytes dysregulated at week 1 vs baseline (14 up, IL-8 down). After BH correction (FDR<0.05): CCL25, CDCP1, Flt3L, CCL8 (MCP-2), and SCF remained significant. Elevated chemokines (CCL25, CCL8, CCL19) and inflammatory mediators (IL-18, TNF) peaked at week 1 and normalized by week 12. STRING/PANTHER analyses highlighted networks for leukocyte migration, chemokine receptor binding, macrophage proliferation, and cytokine-cytokine receptor interactions.
• Correlations with glucose: Across baseline, weeks 1, 4, and 12 (unvaccinated), plasma glucose correlated positively with 8 analytes (including CCL25, GDNF, ADA, ST1A1, CXCL9, IL-10RB, FGF-19, CDCP1) and negatively with IL-8. After BH correction: CCL25 (r=0.57; p=0.0003; FDR=0.004), GDNF (r=0.55; p=0.0004; FDR=0.006), ADA (r=0.44; p=0.007; FDR=0.049) remained significant. Week 1 CDCP1 correlated with glucose at weeks 4 and 16; week 1 GDNF strongly correlated with week 4 glucose. GDNF remained elevated to week 12 (p=0.019) and showed significant correlation with CCL25 (text reports r=0.64; p<0.0001). In a confirmatory dataset (BL and weeks 1–3), GDNF increased significantly after infection in unvaccinated but not in vaccinated animals. Insulin did not significantly correlate with glucose.
• T-cell activation: Memory CD4+ T cells expressing IL-2 increased vs baseline across weeks (FDR<0.01). Vaccinated animals had lower memory CD4+ TNF responses to PMA/I vs unvaccinated (PERMANOVA p=0.033). Memory CD8+ TNF responses were elevated vs baseline (FDR<0.01). Spectre analysis identified: Population 3 (CD8+CD69+ producing IFNγ and TNF) correlated with glucose across time (r=0.41; p=0.0003). Population 4 (activated CD8+ producing TNF and IL-2) correlated with plasma GDNF (r=0.47; p=0.016).
• Tissue viral persistence and pathology: RNAscope at 18 weeks p.i. showed negligible SARS-CoV-2 spike RNA-positive cells in duodenum (mean 0.018%), liver (0.005%), and pancreas (0.012%); similar minimal signals at 4 weeks p.i.; lung controls at 4 weeks p.i. were positive. Tissue qPCR at 18 weeks p.i. detected no sgN/sgE in liver, duodenum, pancreas; low genomic N in duodenum of 2/14 animals (Ct ~30–32); lung signals near negative control Ct. Duodenal IHC was negative. Lung histopathology at 18 weeks showed minimal inflammation (comparable to uninfected controls); severe inflammation seen only in some 3–4 week endpoint comparators. Pancreas showed none to minimal inflammation and no fibrosis in 13/15 (mild in 2).
• Liver glycogen: PAS staining showed increased hepatic glycogen, reaching significance in long-term unvaccinated animals vs controls. Hepatic glycogen correlated with glucose at week 8 (r=0.54; p=0.04) and week 12 (Δ intensity r=0.74; p=0.003; absolute intensity r=0.64; p=0.014). No hepatic steatosis or fibrosis at necropsy.
• Vaccination effect: A single BNT162b2 dose at day 4 p.i. was associated with better glycemic control over time (PERMANOVA p=0.001 for between-group glucose trajectories), despite similar antibody magnitudes to unvaccinated animals.

This work establishes African green monkeys as a non-human primate model that recapitulates key virologic, immunologic, and metabolic features of human COVID-19 and metabolic PASC: early-onset hyperglycemia that persists months after acute infection, early chemokine-dominant inflammatory perturbations, and heightened T-cell cytokine responses to polyclonal stimuli. The persistence of hyperglycemia despite minimal evidence of long-term replicating virus in liver or pancreas, limited pancreatic pathology, and lack of correlation with insulin supports contributions from extra-pancreatic mechanisms such as hepatic glucose production and altered liver glycogen handling. The acute-phase plasma signature—especially chemokines (CCL25, CCL8, CCL19), inflammatory mediators (IL-18, TNF), and GDNF—correlated with subsequent glucose levels, implicating leukocyte migration/chemokine receptor pathways and neurotrophic signaling in dysglycemia. Activated CD8+ T-cell responses to PMA/I correlated with glycemia, linking heightened inflammatory potential to metabolic disturbance. Importantly, early mRNA vaccination (day 4 p.i.) was associated with improved glycemic control without large differences in antiviral antibody magnitudes, suggesting that early immunomodulation may mitigate downstream metabolic sequelae independent of humoral responses. Collectively, the findings support a model where early systemic inflammatory and chemokine signaling perturb glucose regulation and contribute to long-term hyperglycemia after SARS-CoV-2 infection, with the AGM model suited for testing targeted interventions.

SARS-CoV-2-infected African green monkeys develop early and persistent hyperglycemia resembling metabolic PASC. An acute inflammatory plasma signature dominated by chemokines (e.g., CCL25) and the neurotrophic factor GDNF significantly associates with dysglycemia over months. Activated CD8+ T-cell responses to polyclonal stimulation correlate with glucose levels, further linking immune activation to metabolic impairment. Minimal long-term viral nucleic acid/protein in liver, pancreas, and duodenum and limited organ pathology suggest mechanisms beyond persistent organ infection, potentially involving hepatic glucose metabolism and systemic inflammation. Early mRNA vaccination (day 4 post-infection) is associated with improved glycemic trajectories, offering a potential strategy to mitigate metabolic PASC. Future research should: (1) mechanistically dissect roles of chemokines (e.g., CCL25/CCR9 axis) and GDNF/RET pathways in glucose regulation post-infection; (2) test therapeutic targeting of chemokines or their autoantibodies; (3) evaluate hepatic gluconeogenic regulators (e.g., GP73, PEPCK) and glucagon; (4) examine sex-specific effects; (5) expand inflammatory analyte coverage (e.g., interferons); and (6) define optimal vaccination timing to prevent metabolic sequelae.

• Skewed sex distribution (predominantly female AGMs) limits assessment of sex-specific effects.
• OLINK panel QC excluded some proteins; interferons were under-represented, potentially missing relevant pathways.
• Study focus and design did not allow detailed mechanistic dissection of glucometabolic regulation (e.g., direct measures of hepatic gluconeogenesis, glucagon, or β-cell function beyond insulin correlations).
• Small cohort size typical of NHP studies may limit generalizability; one vaccinated animal necropsied at week 8 due to anorexia.
• Certain glucose timepoints (weeks 10 and 18) were excluded due to housing transitions (BSL-3 to BSL-2 and back) that could transiently affect measurements.
• Tissue viral assessments were at endpoint (18 weeks); intermittent earlier persistence in some tissues cannot be fully excluded.