Duodenal Mucosal Expression of COVID-19-Related Genes in Health, Diabetic Gastroenteropathy, and Functional Dyspepsia.

CONTEXT: SARS-CoV-2 infects the gastrointestinal tract and may be associated with symptoms that resemble diabetic gastroparesis. Why patients with diabetes who contract COVID-19 are more likely to have severe disease is unknown.
OBJECTIVE: We aimed to compare the duodenal mucosal expression of SARS-CoV-2 and inflammation-related genes in diabetes gastroenteropathy (DGE), functional dyspepsia (FD), and healthy controls.
METHODS: Gastrointestinal transit, and duodenal mucosal mRNA expression of selected genes were compared in 21 controls, 39 DGE patients, and 37 FD patients from a tertiary referral center. Pathway analyses were performed.
RESULTS: Patients had normal, delayed (5 FD [13%] and 13 DGE patients [33%]; P = 0.03 vs controls), or rapid (5 FD [12%] and 5 DGE [12%]) gastric emptying (GE). Compared with control participants, 100 SARS-CoV-2-related genes were increased in DGE (FDR < 0.05) vs 13 genes in FD; 71 of these 100 genes were differentially expressed in DGE vs FD but only 3 between DGE patients with normal vs delayed GE. Upregulated genes in DGE include the SARS-CoV2 viral entry genes CTSL (|Fold change [FC]|=1.16; FDR < 0.05) and CTSB (|FC|=1.24; FDR < 0.05) and selected genes involved in viral replication (eg, EIF2 pathways) and inflammation (CCR2, CXCL2, and LCN2, but not other inflammation-related pathways eg, IL-2 and IL-6 signaling).
CONCLUSION: Several SARS-CoV-2-related genes were differentially expressed between DGE vs healthy controls and vs FD but not between DGE patients with normal vs delayed GE, suggesting that the differential expression is related to diabetes per se. The upregulation of CTSL and CTSB and replication genes may predispose to SARS-CoV2 infection of the gastrointestinal tract in diabetes.

Diabetes and obesity are independent and synergistic risk factors for worse outcomes in patients with COVID-19 infection (1-6). Both conditions affect the gastrointestinal tract (7-9). Up to 50% of patients infected with SARS-CoV-2 have gastrointestinal symptoms (eg, nausea, vomiting, diarrhea, and abdominal pain) (10), which are similar to the symptoms of diabetic gastroenteropathy (7, 8, 11-13). The more severe gastrointestinal complications of COVID-19 (e.g., ileus, mesenteric ischemia) are also associated with diabetes (14, 15). It has been suggested that “the underlying chronic inflammatory state in diabetes may be ‘locked and loaded’ for virus-induced damage, promoting a vicious cycle of cytokine release and hyperglycemic surges, leading to more widespread multiorgan damage, including injury to tissues already weakened by preexisting diabetes complications (16).” However, the extent to which the increased severity of SARS-CoV-2 infection in diabetes is explained by underlying host factors (eg, dysimmunity) is unclear.

Gastrointestinal involvement may also affect the severity of COVID-19 independent of diabetes. In China, gastrointestinal involvement was associated with less favorable outcomes (17). By contrast, in the United States, patients with COVID-19 and gastrointestinal symptoms were more likely to have a favorable outcome, reduced small intestine mucosal expression of several inflammatory cytokines and chemokines (eg, interleukin [IL]-1β and interferon [IFN]-γ]), and reduced circulating levels of key inflammatory proteins than patients with COVID-19 who did not have gastrointestinal symptoms (9).

After SARS-CoV-2 binds to the angiotensin-converting–enzyme 2 (ACE2) receptors on epithelial cells, the proteases transmembrane-serine-protease-2 (TMPRSS2) and, to a lesser extent, CTSL and CTSB (Cathepsin L and B) are responsible for priming the spike (Increased) protein of SARS-CoV-2 (18, 19). It has been suggested that “old age, obesity, and diabetes produce a deadly symbiosis of dysregulated immunometabolism and chronic systemic inflammation that intensifies virally induced hyperinflammation associated with SARS-CoV-2 infection” (20). The bronchial and alveolar expression of the ACE2 protein but not mRNA is increased in diabetes patients who were not infected with SARS-CoV-2 (21). Since the gastrointestinal tract provides a route for SARS-CoV2, the aims of this study were to compare the duodenal mucosal expression of genes and pathways that are involved in viral replication and the pathogenesis of illness related to SARS-CoV-2 infection between patients with 1) diabetic gastroenteropathy and healthy controls and 2) functional dyspepsia, which served as a disease-control group, and healthy controls.

Methods

Study Design and Participants

All participants provided informed consent to participate in this study, which was approved by the Institutional Review Board at Mayo Clinic. From June 2014 to April 2017, 40 patients with diabetic gastroenteropathy and 40 patients with functional dyspepsia, who were identified from our clinical practice, and 24 healthy volunteers, who were identified by public advertisement, participated in this study that primarily sought to evaluate the effects of the GLP-1 antagonist exendin 9-39 on symptoms during enteral lipid infusion (22, 23). This report is based on 21 healthy controls, 39 diabetic gastroenteropathy patients, and 37 functional dyspepsia patients in whom the duodenal mucosal transcriptome was satisfactorily assessed (24, 25). The participants, who were aged between 18 and 70 years, did not have major systemic diseases that may interfere with study objectives or pose safety concerns, or gastrointestinal operations other than appendectomy, cholecystectomy, or hysterectomy.

Clinical Assessment

All participants were carefully evaluated by a physician. Patients had symptoms of dyspepsia documented by a gastroenterologist while controls did not. Symptoms over the preceding 2 weeks were characterized by the Patient Assessment of Upper Gastrointestinal Disorders-Symptom Severity (PAGI-SYM) questionnaire (26). The complications of diabetes mellitus (DM) (ie, retinopathy, peripheral and autonomic neuropathy, and nephropathy) were defined by established criteria (23).

Gastric Emptying and Small Intestinal Transit

Gastric emptying of solids was assessed with a meal (296 kcal; 32% proteins, 35% fats, and 33% carbohydrates) with 2 eggs labeled with technetium Tc100m sulfur colloid (1 mCi) and served on 1 slice of bread with milk (23). Rapid and delayed emptying were defined as ≥ 36% emptied at 1 hour, and < 76% emptied at 4 hours respectively. Small bowel transit time was calculated as the percentage of isotopically labeled meal that was in the colon at 6 hours (27).

Upper Gastrointestinal Endoscopy

An investigator (A.E.B.) performed an upper gastrointestinal endoscopy. In order to reduce the likelihood that gene expression was confounded by gastropathy related to no-steroidal anti-inflammatory drugs or other conditions, the mucosal biopsies were obtained from the second part of the duodenum and not from the stomach.

RNA Extraction and Sequencing

RNA was extracted from freshly frozen duodenal tissue using TRIzol (Ambion, Waltham, MA) and chloroform. Upon centrifugation, the aqueous phase was purified using an RNeasy Mini Kit (Qiagen, Valencia, CA) following the manufacturer’s instructions. Total RNA was analyzed using the Agilent Bioanalyzer 2100 and samples had an RNA integrity number of at least 7. The total RNA underwent a size selection process, optimized for mRNA products, and was sequenced on an Illumina HiSeq 4000. The RNA-Seq files from this experiment can be accessed in GEO under the accession number, GSE151497. After sequencing, the mRNA data was processed using Mayo Clinic’s RNA-Seq bioinformatics software, MAPR-Seq v2 in stranded mode (28). The transcriptome used for this analysis was Ensembl’s GRCh38.78 reference. After evaluating mapping percentages, total reads, gene body distribution, and other quality metrics, samples in 21 controls, 39 diabetic gastroenteropathy patients, and 37 functional dyspepsia patients were deemed satisfactory for further analysis. The Bioconductor edgeR package identified differentially expressed genes with a false discovery rate (FDR) of 0.05 (29). The results of differentially expressed genes in diabetic gastroenteropathy and functional dyspepsia have been published in papers that focused respectively on mitochondrial disturbances in diabetic gastroenteropathy and duodenal mucosal secretory disturbances in functional dyspepsia (24, 25). This paper is focused on the genes of putative relevance to COVID-19 infection.

Ingenuity Pathway Analysis, Gene-Set Enrichment Analysis, and nFerence Analysis

QIAGEN’s Ingenuity Pathway Analysis (IPA) algorithm (www.qiagen.com/ingenuity, QIAGEN, Redwood City, CA, USA) was used to identify the pathways associated with differentially expressed genes (ie, for genes with FC ± 1.5 and overlap P < 0.05). However, there are limitations of considering only the differentially expressed genes in IPA. Firstly, after adjusting for multiple comparisons, no individual gene may be statistically significant, perhaps because the biological differences are modest relative to the noise inherent to the microarray technology (30). Secondly, single-gene analysis may miss important effects on pathways which result from sets of genes that act in concert. For example, a tiny increase in all genes encoding members of a metabolic pathway may dramatically alter the flux through the pathway and may be more important than an overwhelming increase in a single gene (30). In order to address these gaps, the pathway analysis was also performed with gene-set enrichment analysis (GSEA) using the MsigDB database (30). An FDR cutoff of 0.25 was used to identify the significant gene-sets in GSEA. The IPA and GSEA analyses cataloged the differentially expressed genes into pathways. The analysis focused on those pathways that are known to be relevant to the pathogenesis and/or manifestations of SARS-CoV-2 infection.

The IPA and GSEA uncover pathways that are associated with differentially expressed genes in the literature. However, they do not consider the strength of the association between individual genes and disease processes or pathways. Hence, some of the differentially expressed genes are not strongly connected with the relevant disease mechanisms or pathways identified by the IPA and GSEA. Hence, we used the nferX software (nFerence.ai), which is based on a comprehensive corpus of the literature, to quantify the association between genes and disease processes (eg, diabetes mellitus, viral infections) (31). This software provides a local score that measures how frequently 2 tokens A and B, which, in this instance, respectively reflect a differentially expressed gene and a disease process, are found in close proximity to each other (ie, within 100 words or fewer) in the full set of considered documents (corpus), normalized by the individual occurrences of each token in that corpus. The corpus includes all abstracts in PubMed along with all full PubMed Central (PMC) articles. To calculate the score, the pointwise mutual information between A and B as PMIAB = log10([AdjacencyAB * NC]/[NA * NB]), where AdjacencyAB is the number of times token A occurs within 100 words of Token B (or vice versa), NA and NB are the number of times that Tokens A and B each occur individually in the corpus, and NC is the total number of occurrences of all tokens in the corpus. The local score between Tokens A and B is then calculated as LSAB = ln (AdjacencyAB + 1)/[1 + e-(PMIAB−1.5)]. A local score of 0 indicates that Tokens A and B have never occurred within 100 words of each other and a local score of 3 indicates a co-occurrence likelihood of approximately 1 in 20. A local score of 3 or greater suggests a significant association between the gene and the disease process (31).

Statistical Analysis

The clinical features were compared with Wilcoxon rank sum test. A P value less than 0.05 was considered significant. A Chi-square test was used to compare categorial variables between groups. The statistical analysis was performed using JMP Pro 14 (SAS Institute) and R-3.6.3.

Results

Demographic Features and Gastrointestinal Manifestations

The age, sex distribution, and body mass index were not significantly different among healthy controls, functional dyspepsia patients, and diabetic gastroenteropathy patients (of whom 23 had type 1 DM and 17 had type 2 DM) (Table 1). Among DM patients, the complications were peripheral neuropathy (30 patients), nephropathy (16 patients), and retinopathy (16 patients). The medications for DM included insulin (32 patients), biguanides (ie, metformin, 11 patients), sulfonylureas (2 patients), dipeptidyl peptidase 4 inhibitors (2 patients), glucagon-like peptide 1 (GLP-1) receptor analog (1 patient), and sodium-glucose cotransporter-2 inhibitor (1 patient). Compared with controls, gastrointestinal symptoms were more severe in diabetic gastroenteropathy, and to a greater extent in functional dyspepsia (Table 1). For example, the PAGI satiety subscore was greater (P < 0.03) in functional dyspepsia (median [interquartile range (IQR)], 3.3 [2.4, 4]) than in diabetic gastroenteropathy (2.6 [1.6, 3.5]). Compared to controls, the gastric emptying half-time of solids was longer in diabetic gastroenteropathy (P = 0.008) and in functional dyspepsia (P = 0.04); differences between diabetic gastroenteropathy and functional dyspepsia were also significant (P = 0.02) (Table 1). The gastric emptying of solids was normal in 18 controls (75%), 30 functional dyspepsia patients (75%) and 22 diabetic gastroenteropathy patients (55%). Two controls (8%), 5 functional dyspepsia patients (13%) and 13 diabetic gastroenteropathy patients (33%, P = 0.03 vs controls) had delayed gastric emptying while 4 controls (17%), 5 functional dyspepsia patients (12%), and 5 diabetic gastroenteropathy patients (12%) had rapid emptying. The colonic filling at 6 hours (%), which is a surrogate marker of small intestinal transit, was lower in diabetic gastroenteropathy (P = 0.04) and in functional dyspepsia (P = 0.002) vs controls (Table 1).

Table 1.

Demographic and baseline clinical characteristics

Characteristic	Controls (n = 21)	Diabetes gastroenteropathy (n = 39)	Functional dyspepsia (n = 37)	P value, diabetes gastroenteropathy vs controls	P value, functional dyspepsia vs controls	P value, diabetes gastroenteropathy vs functional dyspepsia
Age, y,	40 (13)	45 (14)	42 (13)	0.33	0.80	0.36
BMI, kg/m2	26.1 (4.7)	27.8 (5.6)	26.6 (6.6)	0.39	0.60	0.23
Female sex, n (%)	11 (52)	30 (79)	28 (76)	0.08	0.09	1
PAGI-SYM subscores, median (IQR)
Heartburn	0 (0, 0.1)	1.1 (0.6, 1.9)	2.0 (1.1, 2.6)	< 0.0001	< 0.0001	0.02
Nausea and vomiting	0 (0, 0)	1.3 (0.3, 3.0)	1 (0.3, 2.2)	< 0.0001	< 0.0001	0.32
Satiety	0 (0, 0.3)	2.6 (1.6, 3.5) -+	3.3 (2.4, 4)	< 0.0001	< 0.0001	0.03
Bloating	0 (0, 0.3)	2.5 (1.0, 4)	3.5 (2.5, 4.8)	< 0.0001	< 0.0001	0.04
Upper abdominal pain	0 (0, 0)	2.0 (1.0, 3.5)	3.5 (2.0, 4.0)	< 0.0001	< 0.0001	0.009
Lower abdominal pain	0 (0, 0)	1.5 (0.5, 2.5)	2.0 (1.0, 3.25)	< 0.0001	< 0.0001	0.13
GCSIa	0 (0, 0.2)	2.3 (0.6, 4.1)	2.7 (1.7, 5.4)	< 0.0001	< 0.0001	0.14
GE t1/2 (minutes)	104 (35)	144 (48)	121 (35)	0.0008	0.04	0.02
Colonic filling at 6 hours, %	59 (20)	45 (28)	40 (18)	0.04	0.002	0.62
HbA1c, %	5.2 (0.3)	8.7 (2.1)	5.2 (0.4)	< 0.0001	0.51	< 0.0001
Fasting glucose, mg/dL	93 (7)	160 (59)	91 (6.0)	< 0.0001	0.48	< 0.0001

BMI, body mass index; GCSI, Gastroparesis Cardinal Symptom Index; GE, gastric emptying; IQR, interquartile range; PAGI-SYM, Patient Assessment of Upper Gastrointestinal Disorders-Symptom Severity.

Values are mean (SD) unless specified otherwise.

aAverage of nausea and vomiting, satiety, and bloating subscores.

mRNA Expression of all Genes in Diabetic Gastroenteropathy

There were 3175 differentially expressed genes (FDR < 0.05) in patients with diabetic gastroenteropathy vs healthy controls. Of these 3175 genes, 1558 and 1617 genes, respectively, were down- and upregulated in diabetic gastroenteropathy, as detailed previously (25). Of these genes, 201 genes were differentially expressed between diabetic gastroenteropathy patients who had normal vs delayed gastric emptying. The IPA and GSEA identified 48 and 50 significantly enriched pathways, of which 14 pathways that are known to be relevant to the pathogenesis and/or manifestations of SARS-CoV-2 infection are included in this report (Table 2). These pathways pertain to disease processes such as viral entry and replication (eg, for eukaryotic initiation factor-2 [EIF2] signaling), antiviral response, inflammation, coagulopathy, and dysmetabolism. Taken together, these pathways comprised 100 differentially expressed genes between diabetic gastroenteropathy and controls, of which only 3 genes (HCLS1-associated protein X-1 [HAX1], G0/G1 switch gene-2 [G0S2], and phosphomevalonate kinase [PMVK]) were differentially expressed between diabetic gastroenteropathy patients with normal and delayed gastric emptying.

Table 2.

Differentially expressed mucosal genes that are differentially expressed (FDR < 0.05) in diabetic gastroenteropathy vs healthy controls

Disease process	Pathway	Diabetic gastroenteropathy vs controls	Diabetes gastroenteropathy vs functional dyspepsia	Functional dyspepsia vs controls
Viral entry		CTSB** (1.24), CTSL* (1.16)	CTSB*** (1.28), CTSL***(1.25)	CTSB (0.96), CTSL (0.92)
Viral replication	Coronavirus specific	COPB1** (1.28), COPB2* (1.18), CTSL* (1.16), DDX1* (1.11), HNRNPA1** (1.14), TUBA1A* (1.20), TUBA1B* (1.28)	COPB1***(1.36), COPB2***(1.19), CTSL*** (1.25), DDX1*** (1.14), HNRNPA1***(0.90), TUBA1A (1.11), TUBA1B*** (1.22)	COPB1 (0.93), COPB2 (0.98), CTSL (0.92), DDX1 (0.97), HNRNPA1(1.27), TUBA1A (1.09), TUBA1B (1.04)
	EIF2 signaling	AGO2* (0.88), AGO4* (0.89), DHX9* (1.12), EIF4A1** (1.43), MT-RNR1*** (0.38), MT-RNR2*** (0.19), PIK3C2B* (0.84), RPL23A*** (1.34), RPL31*** (1.31), RPL34*** (1.32), RPL39*** (1.33), RPL9** (1.31), RPLI8A*** (1.48), RPS29*** (1.50), RPSA*** (1.43), TRIB3*** (3.53)	AGO2***(0.88), AGO4*** (0.91), DHX9*** (1.14), MT-RNR1*** (2.62), PIK3C2B*** (0.84), RPL23A*** (1.17), RPL31* (1.14), RPL34 (1.05), RPL39 (0.95), RPL9 (1.00), RPSA (0.0)	AGO2 (1.01), AGO4 (0.97), DHX9 (0.100), EIF4A1* (1.27), MT-RNR1 (0.21), MT-RNR2 (0.37), PIK3C2B (0.100), RPL23A* (1.14), RPL31* (1.16), RPL34*** (1.26), RPL39 (1.42), RPL9*** (1.31), RPLI8A*** (1.47), RPS29*** (1.39), RPSA (1.38), TRIB3 (1.08)
	Pyrimidine Ribonucleotides De Novo Biosynthesis	DDX3X* (1.16), DHX9* (1.12), NME1* (1.40), NME2* (1.22), NUDT15* (1.12), NUDT5*** (1.16), SLC25A42* (1.31), UMPS*** (1.15)	DDX3X***(1.20), DHX9***(1.13), NME1 (1.17), NUDT15*** (1.13), UMPS* (1.08)	DDX3X (0.97), DHX9 (0.100), NME1 (1.18), NME2 (1.10), NUDT15 (0.100), NUDT5 (1.03), SLC25A42 (1.04), UMPS (1.05)
	Pyrimidine Ribonucleotides Interconversion	DDX3X* (1.16), DHX9* (1.12), NME1* (1.40), NME2* (1.22), NUDT15* (1.12), NUDT5*** (1.16), SLC25A42* (1.31)	DDX3X*** (1.20), DHX9*** (1.13), NME1 (1.17), NUDT15*** (1.13)	DDX3X (0.97), DHX9 (0.100), NME1 (1.18), NME2 (1.10), NUDT15 (0.100), NUDT5 (1.03), SLC25A42 (1.04), UMPS (1.05)
	tRNA charging	DARS2* (1.11), FARSA* (1.17), MARS2* (1.18), YARS2* (1.11)	DARS2*** (1.14), FARSA*** (1.14), MARS2** (1.13), YARS2** (1.08)	DARS2 (0.97), FARSA (1.02), MARS2 (1.04), YARS2 (1.02)
Antiviral response		BCL3** (0.75), BNIP3* (0.80), BRD4***(0.83), CANX* (1.14), DAPK1* (0.85), FYCO1* (0.90), GZMB* (1.61), HAT1* (1.18), LCN2** (2.42), SAE1* (1.14), TGFB1** (0.85)	BCL3(0.92), BNIP3 (0.00), BRD4*** (0.84), CANX*** (1.26), DAPK1*** (0.85), FYCO1*(0.93), HAT1*** (1.26), LCN2*** (2.62), SAE1*** (1.19)	BCL3 (0.82), BNIP3 (0.80), BRD4 (0.100), CANX (0.90), DAPK1 (1.00), FYCO1 (0.96), GZMB (0.87), HAT1 (0.93), LCN2 (0.93), SAE1 (0.95), TGFB1(0.90)
Inflammation	Cytokine storm	CCR2*(1.53), CXCL2* (1.63), LCN2** (2.42)	CCR2(1.18), CXCL2** (1.48), LCN2*** (2.62)	CCR2* (1.31), CXCL2 (1.01), LCN2 (0.93)
	IL-2, IL-6 signaling	HAX1***a (1.24), INHBE* (2.37)	HAX1*** (1.39)	HAX1 (0.88), INHBE (1.34)
	IFN-gamma response	BPGM* (1.18), BST2* (1.33), GCH1* (1.34), IFI30*** (1.31), PLA2G4A* (1.20), PSMB2** (1.18), PSMB8* (1.28), PSMB9** (1.36), TXNIP (1.38)	BPGM*** (1.24), BST2 (1.14), GCH1*** (1.55), IFI30*** (1.44), PLA2G4A*** (1.33), PSMB2***(1.16), PSMB8*** (1.25), PSMB9*** (1.28), TXNIP (1.37)	BPGM (1.07), BST2 (1.17), GCH1 (0.86), IFI30 (0.90), PLA2G4A (0.90), PSMB2 (1.00), PSMB8 (0.97), PSMB9 (0.92), TXNIP (1.00)
	IFN-alpha response	BST2* (1.33), IFI30* (1.31), PSMB8* (1.28), PSMB9* (1.36), TXNIP* (1.38)	IFI30*** (1.44), PSMB8*** (1.25), TXNIP(1.37)	BST2 (1.17), IFI30 (0.90), PSMB8 (0.97), PSMB9 (0.92), TXNIP (1.00)
	ROS induced damage and nitric oxide production	CHUK* (1.13), CREBBP*** (0.82), DIRAS3* (1.40), IKBKB* (0.90), MAP3K12* (0.85), NOS2** (0.49), PIK3C2B* (0.84), PIK3R2** (1.14), PIK3R4** (1.12), PPP1R14D** (1.31), PPP1R7** (1.17), RBP4* (1.52), RHOBTB1* (1.35), RHOBTB2* (0.77)	CHUK*** (1.25), CREBBP*** (0.84), DIRAS3*** (1.49), IKBKB*** (0.82), MAP3K12* (1.16), NOS2(1.17), PIK3C2B*** (0.84), PIK3R2*** (1.25), PIK3R4*** (1.10), PPP1R14D*** (1.43), PPP1R7*** (1.15), RBP4*** (1.53), RHOBTB1***(1.38), RHOBTB2** (0.80)	CHUK (0.89), CREBBP (0.97), DIRAS3 (0.90), IKBKB* (0.90), MAP3K12 (0.97), NOS2 (0.41), PIK3C2B (0.100), PIK3R2 (0.93), PIK3R4 (1.01), PPP1R14D (0.91), PPP1R7 (1.00), RBP4 (0.98), RHOBTB1 (0.98), RHOBTB2 (0.98)
Coagulopathy		CTSB** (1.24), CTSH* (1.23), PEF1** (1.21)	CTSB*** (1.28), CTSH*** (1.32), PEF1*** (1.44)	CTSB (0.96), CTSH (0.92), PEF1* (0.84)
Dysmetabolism	Oxidative Phosphorylation	ACAT2*** (1.39), ACSL1** (1.64), CEL** (3.05), COX7A2*** (1.37), FABP2* (1.44), G0S2*a (1.77), HMGCS1* (1.42), IDIL1* (1.47), MT-CO1** (0.79), MT-CO2** (0.79), MT-CYB*** (0.59), MT-ND1*** (0.43), MT-ND2*** (0.51), NSDHL*** (1.41), UQCR11*** (1.37)	ACAT2*** (1.25), ACSL1***(1.69), CEL*** (2.84), COX7A2*** (1.41), FABP2*** (1.51), G0S2*** (1.58), HMGCS1*** (1.50), MT-CO1*** (1.32), MT-CO2*** (1.51), MT-ND1 (0.82), MT-ND2 (0.84), NSDHL*** (1.35), UQCR11*** (1.31)	ACAT2 (1.10), ACSL1 (0.94), CEL (1.07), COX7A2 (0.96), FABP2 (0.93), G0S2 (1.11), HMGCS1 (0.93), IDIL1 (0.84), MT-CO1 (0.59), MT-CO2 (0.52), MT-CYB* (0.71), MT-ND1** (0.53), MT-ND2** (0.61), NSDHL (1.02), UQCR11 (0.100)
	TCA cycle	DLD* (1.14), MDH1*** (1.30), SDHC* (1.11), SDHD** (1.23)	DLD*** (1.14), MDH1*** (1.34), SDHC (1.04), SDHD*** (1.41)	DLD (0.87), MDH1 (0.95), SDHC (1.06), SDHD(0.87)
	Cholesterol synthesis	ACAA2* (1.22), ACAT2** (1.39), DHCR7* (1.46), FDFT1** (1.31), FDPS* (1.37), FH** (1.19), IDI1* (1.47), PMVK*a (1.32)	ACAA2 *(1.14), ACAT2*** (1.25), DHCR7* (1.34), FDFT1*** (1.38), FDPS*** (1.33), FH*** (1.19), PMVK*** (1.23)	ACAA2 (1.07), ACAT2 (1.10), DHCR7 (1.09), FDFT1 (0.95), FDPS (1.02), FH (0.98), IDI1 (0.84), PMVK (1.06)

Values in parentheses are fold changes.

The genes with a local score > 3.0 are shown in bold font.

Abbreviation: FDR, false discovery rate.

*FDR: 0.01-0.05 **FDR: 0.005-0.01,***FDR: <0.005.

aAlso differentially expressed between diabetic gastroenteropathy with delayed gastric emptying and diabetic gastroenteropathy with normal gastric emptying with fold change changes being HAX1*** (1.23), G0S2*** (2.14), and PMVK*** (1.36).

Of note, there are 100 differentially expressed genes between diabetic gastroenteropathy vs controls. Genes that pertain to 2 or more pathways are listed more then once in the table.

Genes Involved in Cellular Entry of COVID-19 in Diabetic Gastroenteropathy

The mRNA expression of CTSL (|FC|=1.16) and CTSB (|FC|=1.24) were upregulated (FDR < 0.05) in diabetic gastroenteropathy compared with healthy controls; however, differences were modest. By contrast, the expression of ACE2 (|FC|=1.16, FDR = 0.31), TMPRSS2 (|FC|=1.07, FDR = 0.55), and FURIN (|FC|=1.05, FDR = 0.54) were not different in diabetic gastroenteropathy vs controls (Table 2). None of the viral entry or SARS-CoV-2 viral replication genes shown in Table 2 were differentially expressed between patients who were taking metformin vs those not taking metformin (data not shown).

Genes Regulating SARS-CoV2 Replication, Inflammation, Coagulopathy, and Dysmetabolism in Diabetic Gastroenteropathy

The IPA identified upregulation of several genes in pathways that are responsible for viral replication, inflammation, coagulation, and metabolism in diabetic gastroenteropathy compared with controls (Table 2). This differential expression suggests upregulation of several steps in viral replication, including EIF2 signaling (tribbles pseudokinase 1 [TRIB3] [FC 1.8], ribosomal protein S29 [RPS29] [FC 0.6], ribosomal protein L18A [RPLI8A] [0.6], eukaryotic initiation factor 4A1 [EIF4A1] [0.5]); biosynthesis and interconversion of pyrimidine ribonucleotides (nucleoside diphosphate kinase [NME1] [0.5], NME2 [0.3], solute carrier 25A4A2 [SLC25A4A2] [0.4]); tRNA charging (methionyl-TRNA synthetase-2 [MARS2] [0.2], phenylalanyl-tRNA synthetase [FARSA] [0.2]); and protein transport (coat complex subunit beta 1 [COPB1] [0.4], coat complex subunit beta 2 [COPB2] [0.2]) in diabetic gastroenteropathy.

Diabetic gastroenteropathy was also associated with increased expression of selected inflammation-related genes, including genes that mediate the cytokine storm (lipocalin-2 [LCN-2] [1.3], chemokine ligand-2 [CXCL2] [0.7], and chemokine receptor-2 [CCR2] [0.6]); interleukin-2 and -6 (IL-2 and IL-6) signaling (HCLS1 Associated Protein X-1 [HAX1] [0.3], inhibin subunit beta-E [INHBE] [1.3]); IFN-gamma and increased responses (thioredoxin-interacting protein [TXNIP] [0.5], proteasome subunit-B9 [PSMB9] [0.5], inducible thiol reductase [IFI30] [0.4]); the antiviral response ([LCN2] [1.3], granzyme-B (GZMB) [0.7]), and damage induced by reactive oxygen species (rho related BTB Domain [RHOBTB1] [0.4], retinol binding protein-4 [RBP4] [0.6], and DIRAS Family GTPase-3 [DIRAS3] [0.5]) (Table 2).

Among genes that regulate metabolic processes, the expression of several mitochondrial genes encoded by mitochondrial DNA (eg, mitochondrial cytochrome oxidase-1 [MT-CO1], mitochondrial NADH dehydrogenase-2 [MT-ND2]) was reduced in diabetic gastroenteropathy. By contrast, in diabetic gastroenteropathy there was increased expression of several mitochondrial genes encoded by nuclear DNA that are critical to oxidative phosphorylation (carboxyl ester lipase [CEL] [1.6], G0/G1 switch 2 [GOS2] [0.8] and acyl-CoA synthetase long chain [ACSL1] [0.7]); TCA cycle (malate dehydrogenase-1 [MDH1] [0.4], succinate dehydrogenase-D [SDHD] [0.3], and fumarate hydratase [FH] [0.3]); and cholesterol synthesis (isopentenyl-diphosphate delta isomerase 1 [IDI1] [0.6], 7-dehydrocholesterol reductase [DHCR7] [0.6], and acetyl-CoA acetyltransferase 2 [ACAT2] [0.5]).

Table 3 shows the genes for which the nFerence local score was greater than 3, which suggests a significant association between these genes and disease processes that are broadly relevant to the pathogenesis of COVID-19. For example, the local score for DEAD-Box Helicase 3 X-Linked (DDX3X) and viral infection is 3.6. In some pathways (eg, EIF2 signaling, cytokine storm, and interferon-gamma response), the local score for several genes, which are indicated in bold font in Table 2, was greater than 3. In other pathways (eg, IL-2 and IL-6 signaling), most differentially expressed genes had a local score less than 3, which suggests that the connection between these genes and the disease process (eg, inflammation) is weak.

Table 3.

Local scores and relevant disease processes of differentially expressed genes in diabetic gastroenteropathy

Infection

DDX3X (3.6), DHX9 (3.6), HNRNPA1 (Rhinovirus: 3.4), RPS29 (Hepatitis C: 4), EIF4A1 (Hepatitis C: 4.5), RPSA (Gastroenteritis: 4.4, Colitis: 4.3, Infection: 4.0), RPL31 (Hepatitis C: 4.5), LCN2 (3.8), CXCL2 (4.6)

CCR2 (Encephalomyelitis/neuroinflammation: 5, HIV infection: 4.2, Influenza infection: 3.5), CANX (Hepatitis C: 5.1; Lyme disease: 3), BRD4 (RSV infection: 3.1), TXNIP (Neuroinflammation: 3), PSMB9 (EBV infection: 4, Vaccinia virus Infection: 3.4), PIK3R4 (Rubella infection: 3.2), PSMB8 (Coxsackie myocarditis: 3.1), GZMB (3.8), TGFB1 (Influenza: 4.2), NOS2 (SIRS: 4.8), MAP3K12 (Zika virus infection: 3.1, HIV related neurodegeneration: 4.2), MDH1(4.2)

Inflammation

CCR2 (Encephalomyelitis/neuroinflammation: 5), SAE1(Myositis: 3.1), BCL3 (Colitis: 3.2), TXNIP (Neuroinflammation: 4.6), PSMB8 (Coxsackie myocarditis: 3.1, Ulcerative colitis: 3.2), IKBKB (Rheumatoid arthritis: 3.8, Skin inflammation: 3.9), CREBBP (Prostatitis: 5.2)

Dysmetabolism

RBP4 (Type 2 DM: 5.6, cardiovascular disease: 4.2), TGFB1 (Diabetic neuropathy: 3.8), G0S2 (Hepatic steatosis: 3.2), ACAT2 (Atherosclerosis and cardiovascular disease: 4.7), ACSL1 (Atherosclerosis and cardiovascular disease: 3.2), DHCR7 (Lathosterolosis: 5.8), FABP2 (Type 2 DM: 4.3), ACAA2 (Lathosterolosis: 3.5, Neurometabolic diseases: 4.1), FDFT (Sudden cardiac death: 3.4), DLD (Neurometabolic diseases: 4.7), FH (Hypercholesterolemia: 8.7),

Differentially expressed genes in diabetic gastroenteropathy with a local score < 3.0

COPB1, COPB2, DDX1, NME1, NME2, TUBA1A, NUDT15, TRIB3, NUDT5, UMPS, DARS2, FARSA, RPL23A, RPL39, RPL34, MARS2, YARS2, HAX1 IFI30, HAT1, BNIP3, PPP1R7 , PLA2G4A, BPGM, DIRAS3, BPGM, GCH1, PIK3R2, DAPK1, FYCO1, PMVK, FDPS, COX7A2, HMGCS1, IDI1

Genes for which a local score is unavailable

SLC254A2, RPLI8A, RHOBTB1, PSMB2, PPP1R14D, CHUK, UQCR11

Relationship Between COVID-Related Proteins and Other Features of Diabetes

Among diabetic gastroenteropathy patients, the HbA1c was inversely correlated with the mRNA expression of ACE2 (r = −0.33, P < 0.04) and CTSB (r = −0.33, P < 0.04) but not with CTSL, TMPRSS2, or FURIN (data not shown). The mRNA expression of CTSL (type 1 DM—4.8 [1], type 2 DM—5.5 [1], P = 0.03) and FYVE and coiled-coil domain autophagy Adaptor 1 (FYCO1) (type 1 DM—2.9 [0.4], type 2 DM—3.2 [0.4], P = 0.04) was greater in type 2 than type 1 diabetes. The mRNA expression of CTSB was lower (21.7 vs 14.5; P < 0.04) in patients with vs without nephropathy. Other correlations between the mRNA expression of other COVID-19 related genes and complications of DM were not significant (data not shown).

mRNA Expression of Genes in Functional Dyspepsia vs Controls

There were 4557 differentially expressed genes with a FDR less than 0.05 in patients with functional dyspepsia vs healthy controls. Of the 100 SARS-CoV-2-related genes that were differentially expressed in diabetic gastroenteropathy vs controls, 13 were also differentially expressed in functional dyspepsia vs healthy controls. This includes a few genes that are involved in EIF2 signaling pathways (eg, RPS29 [0.5], RPLI8A [0.6], EIF4A1 [0.4]), which may facilitate viral replication. CCR2 [0.4] was the only differentially expressed inflammation-related gene. The expression of selected mitochondrial genes that are encoded by mitochondrial DNA (eg, MT-CYB, MT-ND2, MT-ND1) was also reduced in functional dyspepsia. However, by comparison to diabetic gastroenteropathy, there were fewer such genes (Table 2).

mRNA Expression of Genes in Diabetic Gastroenteropathy vs Functional Dyspepsia

There were 14 049 differentially expressed genes between diabetic gastroenteropathy vs functional dyspepsia. Of the 100 SARS-CoV-2-related genes that were differentially expressed in diabetic gastroenteropathy vs controls, 71 were also differentially expressed in diabetic gastroenteropathy vs functional dyspepsia. This includes a few genes that are involved in EIF2 signaling pathways (eg, RPS29 [0.5], RPLI8A [0.6], EIF4A1 [0.4]), which may facilitate viral replication. CCR2 [0.4] was the only differentially expressed inflammation-related gene. The expression of selected mitochondrial genes that are encoded by mitochondrial DNA (eg, MT-CYB, MT-ND2, MT-ND1) was also reduced in functional dyspepsia. However, by comparison to diabetic gastroenteropathy, there were fewer such genes (Table 2).

Discussion

Compared with healthy controls, the duodenal mRNA expression of 100 genes that mediate the pathogenesis or manifestations of SARS-CoV-2, were increased in patients with diabetic gastroenteropathy. Of these 100 genes, 71 were also differentially expressed between diabetic gastroenteropathy and functional dyspepsia. By contrast, only 13 of these 100 genes were differentially expressed in functional dyspepsia patients vs controls and only 3 were differentially expressed between diabetic gastroenteropathy patients with normal vs delayed gastric emptying. Taken together, these findings suggest but do not establish that the observed differences between diabetic gastroenteropathy and controls are more likely related to diabetes per se rather than the gastrointestinal involvement secondary to DM. These differentially expressed genes broadly pertained to viral entry and replication, inflammation, coagulopathy, and dysmetabolism.

Viral Entry

After viral binding to epithelial ACE2 receptors, the proteases TMPRSS2, CTSL, and CTSB prime the spike (Increased) protein of the SARS-CoV2 virus (18, 19). Thereafter, the virus invades cells. Supporting the role of these proteases, TMPRSS2 inhibitors and nonselective cysteinyl cathepsin inhibitors used alone and in combination respectively attenuated and blocked the infection of human epithelial cells by SARS-CoV-2 (18). CTSL and CTSB are respectively expressed in approximately 16% and 53% of dendritic cells in the human duodenum (32). CTSL is also expressed in approximately 5% of goblet cells, Paneth cells, and fibroblasts. The observed differences between expression of CTSL and CTSB between healthy controls and diabetic gastroenteropathy are small, perhaps partly because dendritic cells comprise less than 1% of the population in these tissues (nFerence database). While plausible, there is no evidence that the expression of these proteases affects the likelihood of developing or the severity of SARS-CoV-2 infection. Nonetheless, the magnitude of these differences is comparable to corresponding differences measured with scRNA between smokers and nonsmokers in the alveolar submucosa, which suggests that they may be biologically relevant (33). These proteins were also upregulated in lung autopsy samples obtained from COVID-19 infected patients (34).

In this study, the duodenal mucosal mRNA expression of ACE2 was not different between healthy controls and diabetic gastroenteropathy. A proteomic analysis of 144 autopsy samples from 7 organs showed no differences in the pulmonary ACE2 expression in patients with vs without SARS-CoV-2 infection (34). However, our findings do not exclude the possibility that duodenal mucosal expression of ACE2 proteins differs between groups. For example, pulmonary bronchial and alveolar ACE2 protein but not mRNA expression were increased in DM (21).

Viral Replication

Following viral entry, the replication of SARS-Cov-2 entails the following steps: uncoating, primary translation and polyprotein processing, viral RNA synthesis, assembly, and finally exocytosis (35). At every step, there is considerable interaction between viral and host proteins. Some of the differentially expressed genes in this dataset have been studied in the replication of other viruses. These pathways and genes include (1) the eukaryotic translation initiation factor 2 (EIF2) signaling pathway (eg, RPS29, EIF4A, and DExH-Box Helicase 9 [DHX9]) that is responsible for translating the viral RNA into proteins (36); (2) genes that synthesize and interconvert pyrimidines (eg, Human Dead-box protein 3 [DDX3X] and DHX9) and charge tRNA (34); and (3) genes that encode RNA binding proteins (RBPs), such as Heterogeneous Nuclear Ribonucleoprotein A1 (HNRNPA1), which interacts with SARS-CoV-2, and acts as a template for viral gene transcription and posttranscriptional regulation, thereby facilitating the expression of viral genes (37). The IPA analysis labels selected genes (eg, COPB1, CTSL, and HNRNPA1) as “coronavirus specific” genes (Table 2). However, in the nFerX analysis, only 2 of these 7 genes (ie, CTSL and HNRNPA1) had a local score greater than 3. For example, the expression of COPB1, which is a protein coding gene that transports proteins and lipids from the Golgi apparatus to the endoplasmic reticulum, has been implicated in the replication of SARS-CoV not SARS-CoV-2, which might explain why the local score for this gene is less than 3 (38).

Antiviral Response

Several genes that participate in the host antiviral response to COVID-19 were upregulated in diabetic gastroenteropathy. For some of these genes (eg, GZMB and LCN2, which had local scores of 3.8), the local scores were greater than 3.0. GZMB, which is involved in the activation cascade of cysteine proteases and immune mediated cytotoxicity, contributes to the antiviral response (39). The reduced mucosal expression of B-cell lymphoma 3-encoded protein (BCL3), which plays a critical role in counter-regulating inflammatory responses by limiting the transcription of NF-κB–dependent genes (40), may accentuate the inflammatory response. For others, (eg, HAT1) which is an induced gene that enhances cell proliferation (41), the local score for infection is less than 3.0.

Inflammation

There was increased duodenal mucosal expression of CCR2, CXCL2, and LCN2. CCR2 is a key receptor for the monocyte chemoattractant proteins (subtypes 1-4) that induce monocytes to leave the circulation and differentiate into tissue macrophages (42). CXCL2 induces the migration of neutrophils. LCN2 is an antimicrobial peptide that is released by intestinal epithelium, adipose tissue, and neutrophils (43, 44). The small intestinal mucosal expression of LCN2 genes was also increased in patients with SARS-CoV2 infection vs controls (9). Several interferon response genes (eg, TXNIP) were upregulated in diabetic gastroenteropathy vs controls. Thioredoxin-interacting protein (TXNIP) is a pro-apoptotic protein that is strongly induced by glucose and is a key regulator of diabetic beta-cell apoptosis and dysfunction (45). For both overexpressed genes (ie, HAX1 and INHBE) that are associated with IL-2 and IL-6 signaling pathways in the IPA analysis, the local score was less than 3.0, which suggests that these genes are not strongly connected with these pathways. Taken together, these findings suggest that diabetic gastroenteropathy is associated with upregulation of selected chemokines (eg, CCR2, CXCL2, and LCN2) and selected genes in IFN-gamma and IFN-alfa pathways, which have not been strongly implicated in SARS-CoV2-related immune response (46). However, the mucosal expression of genes that are associated with other acute inflammation pathways (eg, IL-2 and IL-6 signaling) that mediate the cytokine storm in SARS-CoV2 infection was not different in diabetic gastroenteropathy vs healthy controls (16). Similarly, in mice with MERS-CoV infection, diabetes was associated with less severe pulmonary inflammation (ie, monocytes, macrophages, and CD4 + T cells) and reduced expression of cytokines (eg, CCL2, CXCL10, TNFα, and IL6) (47) but more severe and prolonged lung involvement. Hence, less inflammation is not necessarily associated with less severe disease.

Dysmetabolism

As detailed previously, there is reduced expression of mitochondrial DNA–encoded mitochondrial genes (eg, MT-ND1, MT-ND2, and MT-CYB) and increased expression of nDNA-encoded OXPHOS genes in diabetic gastroenteropathy (25). Arguably, the upregulated expression of nuclear oxidative phosphorylation genes may fuel (ie, provide the energy and metabolites) necessary for viral replication (34, 48). The G0/G1 switch gene 2 (G0S2), which is elevated in various inflammatory disorders, such as rheumatoid arthritis (49), vasculitis (50), and psoriasis, is involved in the energy metabolism and proliferation of CD8 + T cells (51). Of note, this gene was also significantly overexpressed in diabetic gastroenteropathy with delayed gastric emptying when compared to normal gastric emptying (FC = 1.10), alluding to an association between mucosal inflammation and delaying of gastric emptying in diabetic gastroenteropathy.

The increased expression of several genes involved in cholesterol synthesis, including DHCR-7, is interesting since “emerging evidence suggests a critical link between innate immunity and cholesterol metabolism” (52). Inhibition of DHCR7 activity enhances macrophage-mediated antiviral function in vitro and in vivo (52). FDPS is involved in the synthesis of cholesterol and virus internalization (53).

Several genes that predispose to oxidative stress, which has been postulated to predispose to alveolar damage, thrombosis, and RBC dysregulation in SARS-COV-2 infection, were increased in diabetic gastroenteropathy (54). These genes include RBP4, which is an adipokine that is associated with insulin resistance and cardiovascular risk factors in DM (55, 56). However, the activity of other genes (eg, NOS2) in these pathways were downregulated in diabetic gastroenteropathy.

Limitations

Compared to healthy controls, the duodenal mucosal expression of genes was different in patients with diabetic gastroenteropathy but not in functional dyspepsia, even though gastrointestinal symptoms were more severe in functional dyspepsia than in diabetic gastroenteropathy. These findings are consistent with the hypothesis that the observed differences in gene expression between diabetic gastroenteropathy and healthy controls are related to DM per se rather than gastrointestinal involvement. However, this needs to be confirmed by studying duodenal mucosal gene expression in DM patients without gastrointestinal symptoms. In many circumstances, transcript levels are not sufficient to predict protein levels (57). Hence, these findings need to be confirmed by measuring protein levels.

Implications

In summary, patients with diabetic gastroenteropathy have increased duodenal mucosal expression of genes that predispose to SARS-CoV-2 infection and to viral replication. Further studies of the natural history of SARS-CoV2 infection are necessary to determine whether prolonged viral shedding and/or post-acute gastrointestinal sequelae are more common in diabetes patients (58, 59). The mucosal expression of selected inflammation-related genes (eg, CXCL2 and LCN2) was modestly increased, which argues against the hypothesis that patients with diabetic gastroenteropathy infected with SARS-CoV2 are predisposed to more severe gastrointestinal inflammation.