Comparable Results of Helicobacter pylori Antibiotic Resistance Testing of Stools vs Gastric Biopsies Using Next-Generation Sequencing

Abstract

Helicobacter pylori cure rates have declined worldwide, making pretreatment susceptibility testing increasingly important.1Savoldi A. et al.Gastroenterology. 2018; 155: 1372-1382.e17Abstract Full Text Full Text PDF PubMed Scopus (449) Google Scholar, 2Graham D.Y. et al.Clin Gastroenterol Hepatol. 2021; https://doi.org/10.1016/j.cgh.2021.03.026Abstract Full Text Full Text PDF Scopus (12) Google Scholar, 3Shah S.C. et al.Gastroenterology. 2021; 160: 1831-1841Abstract Full Text Full Text PDF PubMed Scopus (45) Google Scholar Many US diagnostic laboratories now offer culture and susceptibility testing. In addition, molecular-based testing using next-generation sequencing (NGS) is now available in the United States.4Graham D.Y. et al.Am J Gastroenterol. 2022; https://doi.org/10.14309/ajg.0000000000001659Crossref PubMed Scopus (4) Google Scholar NGS provides susceptibility profiling based on detection of mutations in the H pylori genome associated with phenotypic resistance.5Saracino I.M. et al.Antibiotics. 2021; 10: 437Crossref Scopus (12) Google Scholar Advantages of NGS include detecting resistance to multiple antibiotics simultaneously and rapidly4Graham D.Y. et al.Am J Gastroenterol. 2022; https://doi.org/10.14309/ajg.0000000000001659Crossref PubMed Scopus (4) Google Scholar and providing results comparable with culture-based methods using either fresh or formalin-fixed gastric biopsies.6Argueta E.A. et al.Gastroenterology. 2021; 160: 2181-2183Abstract Full Text Full Text PDF PubMed Scopus (21) Google Scholar,7Hulten K.G. et al.Gastroenterology. 2021; 161: 1433-1442Abstract Full Text Full Text PDF PubMed Scopus (5) Google Scholar Here, we evaluated novel stool-based molecular methods for H pylori detection and, for the first time, resistance profiling by NGS from stool, comparing results obtained from stool with gastric tissue from the same patients. Stool-based susceptibility testing could largely obviate the need for endoscopy and gastric biopsies.8Marrero Rolon R. et al.J Clin Microbiol. 2021; 59 (e03040-20)Crossref PubMed Scopus (12) Google Scholar Two hundred sixty-two patients scheduled for diagnostic upper endoscopy from 4 practices (in Rhode Island, Florida, Illinois, and California) were recruited after providing informed written consent and without consideration of prior H pylori therapy. In addition to biopsies for clinical purposes, antral and corpus biopsies were taken specifically for research. A spontaneously passed stool specimen was also collected within 2 weeks of the endoscopy before any anti–H pylori therapy. Agreement between stool and gastric tissue results was evaluated by kappa calculation (http://vassarstats.net/kappa.html). For details of specimen collection and laboratory analyses see Supplementary Methods. First, the novel fecal polymerase chain reaction (PCR) assay (PyloriDx; American Molecular Laboratories) designed for the detection of H pylori DNA was compared with fecal H pylori antigen testing in 254 patients. Assuming the antigen test was 100% accurate, the sensitivity, specificity, and accuracy of the stool PCR were 98.5%, 98.9 %, and 98.9%, respectively (Supplementary Table 1). Stool PCR results were also in high agreement with those of fresh gastric tissue samples (κ = 0.98). Of 262 patients, 72 (27.5%) were dual H pylori positive by both stool and gastric PCR, 1 patient by stool PCR only, 1 patient by gastric PCR only, and 188 patients negative by both. Of the 72 dual H pylori–positive patients, 2 had insufficient gastric DNA for NGS. Paired stool and gastric tissue NGS analyses were conducted in 70 patients (mean age, 60 years [range, 18-83]; 31 men). Six had gastric NGS on formalin-fixed paraffin-embedded tissue sections and 64 from fresh, unfixed biopsies. NGS results were highly comparable for each antibiotic tested, with κ coefficient values ranging from 0.88 to 1.00 (Table 1). In the 6 patients with formalin-fixed paraffin-embedded tissue sections, agreement was also very close, with only 1 antibiotic mismatch identified.Table 1Comparison of NGS Analysis of Antibiotic Resistant Mutations in 6 Antibiotics Between Stool and Fresh Gastric Tissue SamplesAntibioticGene evaluatedResistance-associated mutations by NGSaValues are n (%).Agreement between testsbValues in parentheses are 95% confidence intervals. (κ)GastricStoolClarithromycin23S rRNA34 (53.1)34 (53.1)0.94 (0.90–1.00)LevofloxacingyrA19 (29.7)16 (25.0)0.88 (0.75–1.00)MetronidazolerdxA20 (31.3)17 (26.6)0.89 (0.76–1.00)Tetracycline16S rRNA6 (9.4)6 (9.4)1.00Amoxicillinpbp14 (6.3)4 (6.3)1.00RifabutinrpoB001.00NGS data were compared from 64 patients with positive PCR for H pylori in both fresh gastric and stool samples. A κ greater than 0.81 indicates very good agreement. For clarithromycin the numbers (and percentages) of cases was the same for gastric and stool samples but the κ coefficient was less than 1.00 because 1 patient had a mutation identified in the stool but not in the gastric specimen, whereas another had the mutation identified in the gastric sample but not in the stool.a Values are n (%).b Values in parentheses are 95% confidence intervals. Open table in a new tab NGS data were compared from 64 patients with positive PCR for H pylori in both fresh gastric and stool samples. A κ greater than 0.81 indicates very good agreement. For clarithromycin the numbers (and percentages) of cases was the same for gastric and stool samples but the κ coefficient was less than 1.00 because 1 patient had a mutation identified in the stool but not in the gastric specimen, whereas another had the mutation identified in the gastric sample but not in the stool. Overall, NGS results were concordant between stool and gastric biopsy samples in 64 of 70 patients (91.4%) and in 59 of 64 patients (92.2%) with fresh gastric tissue samples. In 6 patients there was mismatch because of a single antibiotic in 4 patients. In 1 patient 2 mutations were identified in the gastric sample with none in the stool, and in 1 patient 3 mutations were noted in gastric tissue, but only 1 was in stool. In addition to these clinically relevant mismatches in 6 subjects, 5 others had differences in the types and frequencies of mutations identified in their stool and gastric samples that did not influence determination of antimicrobial resistance. Discordance between the stool and gastric samples might have been due, at least in part, to heteroresistance of minority substrains of H pylori9Ailloud F. et al.Nat Commun. 2019; 10: 2273Crossref PubMed Scopus (58) Google Scholar that were detected from stool but not gastric samples, or vice versa. The results of stool NGS resistance testing correlated closely with those from gastric NGS regarding the number of antibiotics predicted to be resistant for each subject (Supplementary Table 2). Of 70 patients, no antibiotic resistance mutation was found in 21 patients (30%) overall, with at least 1 gene with mutations associated with resistance in 49 samples (70%). Our study is the first to report the use of NGS to determine the H pylori antimicrobial susceptibility profile from feces, 1 of the most complex biologic materials. The SuperFecal (American Molecular Laboratories) method was specifically developed for NGS, including reagents and steps to remove any PCR inhibitors and to obtain sufficient high purity and integrity DNA from stool. Because NGS results from stool specimens correlated very closely with gastric biopsy NGS in the same patients, we conclude that accurate antibiotic susceptibility data can now be obtained without the inconvenience, costs, and risks of endoscopy. Stool NGS testing, which is now commercially available in the United States with a Proprietary Laboratory Analysis code, will be especially valuable for refractory cases, where regimen selection by empiric guesswork is no longer recommended.3Shah S.C. et al.Gastroenterology. 2021; 160: 1831-1841Abstract Full Text Full Text PDF PubMed Scopus (45) Google Scholar,4Graham D.Y. et al.Am J Gastroenterol. 2022; https://doi.org/10.14309/ajg.0000000000001659Crossref PubMed Scopus (4) Google Scholar Our data also highlight the high rates of antibiotic resistance among H pylori from patients in 4 widely separated clinical sites in the United States and confirm the view that clarithromycin, metronidazole, and levofloxacin should now only be used in triple therapy after confirmation of susceptibility.2Graham D.Y. et al.Clin Gastroenterol Hepatol. 2021; https://doi.org/10.1016/j.cgh.2021.03.026Abstract Full Text Full Text PDF Scopus (12) Google Scholar,4Graham D.Y. et al.Am J Gastroenterol. 2022; https://doi.org/10.14309/ajg.0000000000001659Crossref PubMed Scopus (4) Google Scholar Molecular-based susceptibility testing, such as NGS, is best suited to predicting H pylori resistance to antibiotics in which the molecular underpinnings of resistance are well understood, such as clarithromycin and levofloxacin. The protocols were designed to capture the major known clinically relevant antibiotic mutations but may under-report resistance by missing rare or unknown mutations or nongenetic mechanisms.10Tshibangu-Kabamba E. et al.Nat Rev Gastroenterol Hepatol. 2021; 18: 613-662Crossref PubMed Scopus (55) Google Scholar Limitations of our study include concerns over generalizability to other clinical settings or geographic regions. Also, whether patients had undergone H pylori eradication therapy and their antibiotic exposure history were not known. This may have provided insight into the high rates of resistance observed. Further limitations are that the stool antigen test alone was used as the reference standard of infection presence rather than a composite of biopsy-based H pylori tests and that no “gold standard” comparison with phenotypic resistance testing was performed. In a prior study,7Hulten K.G. et al.Gastroenterology. 2021; 161: 1433-1442Abstract Full Text Full Text PDF PubMed Scopus (5) Google Scholar clarithromycin and levofloxacin resistance evaluated by NGS in gastric biopsies were well correlated with phenotypic testing, whereas metronidazole and amoxicillin resistance were less so, and require further detailed investigation. We have previously shown6Argueta E.A. et al.Gastroenterology. 2021; 160: 2181-2183Abstract Full Text Full Text PDF PubMed Scopus (21) Google Scholar that clarithromycin and levofloxacin resistance evaluated by NGS from formalin-fixed paraffin-embedded tissues were predictive of eradication failure with triple therapy, whereas the assessment of metronidazole, in which resistance is not “all or none,” was not. Ultimately, prospective comparisons of NGS and phenotypic testing on clinical outcome will be necessary to determine which method (phenotypic or genotypic, gastric or stool) is more cost-effective and practical in real-world situations outside research studies.∗ The authors appreciate the assistance of Jason Machan, PhD, at Brown University for statistical advice and Hongjun Zhang (American Molecular Laboratories) for manuscript review. Steven F. Moss, MD (Conceptualization: Lead; Formal analysis: Lead; Investigation: Equal; Resources: Equal; Writing – original draft: Lead). Long P. Dang, MD (Investigation: Supporting; Resources: Equal). David Chua, MD (Investigation: Supporting; Resources: Equal). Javier Sobrado, MD (Investigation: Supporting; Resources: Equal). Yi Zhou, PhD (Methodology: Supporting). David Y. Graham, MD (Conceptualization: Supporting; Writing – review & editing: Lead). At the Rhode Island site, research biopsies were formalin-fixed and paraffin-embedded as for routine gastric histopathology specimens. At the other 3 sites fresh gastric tissue biopsies were placed in AmT™ Tissue Buffer, a tissue preservation and transportation buffer developed by American Molecular Laboratories, and shipped to American Molecular Laboratories within 7 days. Stool specimens were collected using an in-toilet stool collection device (Protocult; Ability Building Community, Rochester, MN). The stool was shipped with ice packs for analysis at American Molecular Laboratories. For the Rhode Island site, the study was approved by Lifespan’s Institutional Review Board. At the other sites approval was by Advarra, a commercial Institutional Review Board accredited by the Association for the Accreditation of Human Research Protection Programs. A US Food and Drug Administration–approved monoclonal stool antigen test (H pylori Quik Chek; TechLab, Blacksburg, VA) was used, following the manufacturer’s instruction. This is a rapid membrane enzyme immunoassay qualitative detection test requiring 50 mg of raw stool, with a reported sensitivity of 97% and specificity of 100% compared with a composite reference method of gastric biopsy histology and rapid urease testing.1Savoldi A. et al.Gastroenterology. 2018; 155: 1372-1382.e17Abstract Full Text Full Text PDF PubMed Scopus (449) Google Scholar The results are reported as positive or negative for H pylori antigens. H pylori DNA was either extracted from the formalin-fixed paraffin-embedded tissues as previously described2Graham D.Y. et al.Clin Gastroenterol Hepatol. 2021; https://doi.org/10.1016/j.cgh.2021.03.026Abstract Full Text Full Text PDF Scopus (12) Google Scholar or from fresh tissue stored in AmT™. The fresh tissues were homogenized and then digested with proteinase K for 30 minutes at 56°C, followed by 5 minutes at 90°C. The lysed fresh tissues were loaded onto QIACube for DNA purification with QIAamp MiniRotor Adapters (Qiagen, Germantown, MD). The double-stranded DNA content was quantified by Qubit (Thermo Fisher Scientific, Waltham, MA). Clean and inhibitor-free DNA including adequate H pylori DNA from fecal samples suitable for real-time PCR and NGS was obtained using the SuperFecal method developed by American Molecular Laboratories (World Intellectual Property Organization patent no. WO 2020/237237). Briefly, approximately 0.5 g of solid or semisolid fecal samples were homogenized with SuperFecal lysis buffer. After centrifugation, the supernatant was applied onto a SuperFecal column device and washed twice with SuperFecal wash buffers I and II. The DNA was then eluted from the column and DNA quality and quantity measured using a Nanodrop spectrophotometer (Thermo Fisher Scientific). Purified fecal DNA was either kept at 4°C temporarily for immediate downstream analysis or stored at –20°C for longer term storage. To detect H pylori DNA in gastric tissues, a real-time PCR targeting the H pylori 23S rRNA gene was used as described previously.2Graham D.Y. et al.Clin Gastroenterol Hepatol. 2021; https://doi.org/10.1016/j.cgh.2021.03.026Abstract Full Text Full Text PDF Scopus (12) Google Scholar PyloriDx was used to detect H pylori DNA in fecal samples. PyloriDx is a real-time PCR assay designed to target the 23S rRNA gene for H pylori in fecal samples using the same 23S rRNA primer/probe set. For stool samples, PCR was performed on a QuantStudio 3 real-time PCR system (Thermo Fischer Scientific) with initial denaturation at 95°C for 3 minutes followed by 40 cycles of denaturation at 95°C for 3 seconds and annealing/extension at 60°C for 30 seconds. QuantStudio Design & Analysis Software, version 1.4.3, installed on the Applied Biosystems device, was used for data analysis. For both tissue and stool sample real-time PCR, H pylori 26695 strain genomic DNA (ATCC, Manassas, VA) served as a positive control; nuclease-free water was used for the no template negative control. DNA mutations within the 6 H pylori genes associated with most cases of antibiotic resistance (23S rRNA, gyrA, 16S rRNA, pbp1, rpoB, and rdxA)3Shah S.C. et al.Gastroenterology. 2021; 160: 1831-1841Abstract Full Text Full Text PDF PubMed Scopus (45) Google Scholar,4Graham D.Y. et al.Am J Gastroenterol. 2022; https://doi.org/10.14309/ajg.0000000000001659Crossref PubMed Scopus (4) Google Scholar were analyzed and determined by PyloriAR as described previously.2Graham D.Y. et al.Clin Gastroenterol Hepatol. 2021; https://doi.org/10.1016/j.cgh.2021.03.026Abstract Full Text Full Text PDF Scopus (12) Google Scholar For library preparation, sequencing, and data analysis of the gastric tissue samples, the procedure was followed as previously described.2Graham D.Y. et al.Clin Gastroenterol Hepatol. 2021; https://doi.org/10.1016/j.cgh.2021.03.026Abstract Full Text Full Text PDF Scopus (12) Google Scholar However, for the stool samples, new procedures were specially designed (PyloriAR NGS) to process stool DNA samples, including stool DNA pretreatment and H pylori–targeted gene/DNA region enrichment steps in the preparation of the library. After this, for both stool and gastric sample, adapters and indexes were added according to the manufacturer’s protocol (Illumina Inc, San Diego, CA). The quality of libraries was assessed using the Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA) and a Qubit 3.0 fluorometer (Invitrogen, Waltham, MA). The libraries were normalized, pooled, and diluted. The 600 μL of the final 6 pM pooled library and 5% PhiX were loaded on a MiSeq platform (Illumina Inc) to sequence with a MiSeq Reagent Micro Kit V2 (Illumina Inc). When the MiSeq sequencing run was completed, FASTQ files were analyzed by NextGENe software (v 2.4.1.1; SoftGenetics, State College, PA). In format conversion (FASTQ to FASTA), the adapter sequence and low-quality reads (<100 base pair long) were trimmed before analysis. The high-quality reads were aligned and mapped against the reference genome of H pylori 26695 (NC_000915.1) to analyze and identify mutations or variances associated with conferring resistance. To be named as variants for all positions in the target region, the following criteria had to be met: at least 500× coverage at each position, at least 25 occurrences of variant, and at least 5% allele frequency.Supplementary Table 1Performance Characteristics of Stool PyloriDx PCR TestCharacteristicsValue (%)95% confidence interval (%)Sensitivity98.591.8–100.0Specificity98.996.2–99.9Positive predictive value83.055.1–95.1Negative predictive value99.999.4–100.0Accuracy98.996.7–99.8Characteristics were based on comparison with assumed 100% accuracy of a fecal antigen test. Open table in a new tab Supplementary Table 2Comparison Between Stool and Gastric NGS for Number of Antibiotics Predicted to be ResistantPredicted no. of resistant antibioticsNo. of cases in stool samplesNo. of cases in gastric samplesAgreement (κ)No antibiotic resistance21200.97 (0.90–1.00)≥1 antibiotic43440.97 (0.90–1.00)1 antibiotic16150.87 (0.73–1.00)2 antibiotics16180.84 (0.69–0.99)3 antibiotics9100.94 (0.82–1.00)4 antibiotics111.00Values in parentheses are 95% confidence intervals. NGS data compared were from 64 patients with positive PCR for H pylori in both fresh gastric tissue and stool samples. Open table in a new tab Characteristics were based on comparison with assumed 100% accuracy of a fecal antigen test. Values in parentheses are 95% confidence intervals. NGS data compared were from 64 patients with positive PCR for H pylori in both fresh gastric tissue and stool samples.