Abstract

Because few studies have focused on recurrent Campylobacter bacteremia, we investigated two clinical cases of patients with common variable immunodeficiency and repeated Campylobacter bacteremia over a period of 6–10 years. We analyzed and compared genomes from isolates obtained from both patients during follow-up. For patient #1, 18 isolates of Campylobacter coli and 17 isolates of Campylobacter jejuni were obtained from 2014 to 2024. For patient #2, 10 isolates of C. coli were obtained from 2019 to 2024. Next-generation sequencing was used to identify species, characterize antimicrobial resistance, perform multilocus sequence typing, and analyze core-genome single-nucleotide polymorphisms, as well as to uncover potential sources of contamination. For patient #1, all 18 C. coli isolates obtained from 2022 to 2024 were from the same clonal complex and source of contamination (chicken) and exhibited high levels of genomic resemblance based on core-genome single-nucleotide polymorphism analysis. Each C. coli isolate probably originated from the same initial strain. However, two clusters of C. jejuni were identified: one consisting of isolates from 2014 and the other consisting of the remaining isolates from 2022 to 2024. A 16S rRNA mutation in position A1387G was present in four C. coli isolates from 2022 and 2023, and this was associated with gentamicin resistance. One C. coli isolate was also resistant to ertapenem and exhibited an amino acid duplication within the PorA protein sequence. For patient #2, each C. coli isolate was from the same clonal complex, which was of porcine origin. Similar to patient #1, three of the isolates from 2023 had an A1464G 16S rRNA mutation and were gentamicin resistant. Retrospective analyses of antimicrobial use for both patients highlighted an association between antimicrobial selection pressure and the emergence of resistance markers, suggesting in vivo selection.

Introduction

Campylobacter species, mainly Campylobacter jejuni and Campylobacter coli, are responsible for the highest number of bacterial gastroenteritis cases worldwide, more than those caused by Salmonella species (Chlebicz and Śliżewska, 2018; European Food Safety Authority and European Centre for Disease Prevention and Control, 2025). In 2023, a total of 148,181 cases of campylobacteriosis were reported in Europe, whereas there were only 77,486 cases of salmonellosis (Authority (EFSA) EFS and European Centre for Disease Prevention and Control, 2024). However, mortality rates remain low. In fact, symptoms induced by campylobacteriosis are generally mild; these include abdominal cramps, diarrhea, and fever but healthy individuals usually recover spontaneously (Kaakoush et al., 2015). Nevertheless, rare complications may occur in patients with concomitant diseases such as immunosuppression, diabetes, or cancer, and intestinal infections progress to bacteremia in approximately 1% of cases (Fernández-Cruz et al., 2010; Tinévez et al., 2022).

Immune system disorders are major risk factors for chronic Campylobacter infections, even without any gastrointestinal symptoms. This is especially true for primary immunodeficiency diseases such as common variable immunodeficiency (CVID) or, more frequently, X-linked agammaglobulinemia (XLA). In fact, a recent study in the United Kingdom showed that in patients with primary immunodeficiency and chronic or recurrent Campylobacter infections, strains can persist for years in the intestinal tract and reinfect the patient (Grammatikos et al., 2023). CVID is characterized by a deficiency in immunoglobulin production and is the leading cause of primary immunodeficiency, with a prevalence of approximately 1 case per 30,000 adults worldwide. CVID includes a range of conditions defined by distinctive genetic and immunological characteristics (Warnatz et al., 2002; Oksenhendler et al., 2008). XLA is a primary immunodeficiency caused by a mutation in the Bruton tyrosine kinase gene and characterized by a deficiency of circulating B lymphocytes and immunoglobulins (Conley and Cooper, 1998). It may be the most common disease associated with recurrent Campylobacter infections and affects individuals across a wide range of age groups (Grammatikos et al., 2023).

Relapses of campylobacteriosis are rare in immunocompetent individuals (García-Sánchez et al., 2024). Therefore, it is advisable to look for evidence of immunodepression in cases of recurrent Campylobacter infections (Gharamti et al., 2020). These infections can sometimes be complicated by reactive arthritis (Arai et al., 2007), deep-seated infections such as osteomyelitis (Hartman et al., 2020), or cellulitis (Tokuda et al., 2004). The characteristics of both the patient and the strain may strongly influence the duration of infections, and antimicrobial administration combined with immunoglobulin replacement therapy is often necessary to eradicate the problem (Najjar et al., 2020). Although immunoglobulin replacement therapy can counteract Campylobacter infections, it does not appear to influence the chronic asymptomatic presence of the bacteria. IgA deficiency may be a risk factor for campylobacteriosis in individuals with CVID or XLA (Oksenhendler et al., 2008; Dion et al., 2019; Roa-Bautista et al., 2023), but some data have highlighted other potential risk factors (Dion et al., 2019; Grammatikos et al., 2023; Markocsy et al., 2024). IgM deficiency may reduce the clearance of Campylobacter from the blood, but can also alter intestinal permeability and thus contribute to an increased risk of bacteremia (Merrick et al., 2022). Indeed, standard preparations of polyvalent immunoglobulins contain only very small amounts of IgA and IgM, which may explain recurrent campylobacteriosis even in patients receiving immunoglobulin replacement therapy.

In France (Lehours et al., 2023) and throughout the European Union, Campylobacter resistance rates to ciprofloxacin and tetracycline had reached alarming levels by 2023 in human isolates (Authority EFS, European Centre for Disease Prevention and Control, 2025). For C. jejuni and C. coli species, 71.9 and 75% resistance to ciprofloxacin, respectively, was observed in the European Union, compared to 64.8 and 66.2% in France. Resistance to ciprofloxacin is associated with mutations at position 86 or 90 in the Gyrase A protein (Tang et al., 2017). In total, 47.9% of C. jejuni isolates and 68.2% of C. coli isolates were resistant to tetracycline in the European Union, compared to 44.1 and 78.9% in France; the main resistance mechanism involved was tet(O) or sometimes tet(O-32-O) (Sougakoff et al., 1987). In patients with primary immunodeficiency, resistance to antimicrobials such as carbapenems and macrolides may drastically reduce the impact of antimicrobial therapy alone (Ongen et al., 2023; van der Meer et al., 1986; van den Bruele et al., 2010; Zhuo et al., 2024), sometimes proving fatal (Merrick et al., 2022).

In the present study, we describe the clinical and microbiological characteristics of two patients with CVID and recurrent Campylobacter infections. A total of 45 isolates were obtained from the two patients and sent to the French National Reference Center for Campylobacters and Helicobacters (NRCCH) for analysis. These included 18 isolates of C. coli and 17 isolates of C. jejuni from the first patient and 10 isolates of C. coli from the second patient. Whole-genome sequencing (WGS) was used to evaluate the adaptation of these strains to antimicrobial therapies. The study highlighted strong selective pressure among the Campylobacter populations obtained from these patients and noted the emergence of highly resistant bacterial strains with rare or novel genomic resistance mechanisms.

Materials and methodsClinical cases descriptions

This study describes recurrent infections of C. jejuni and C. coli in two female patients with CVID from the University Hospital of Lyon (Hospices Civils de Lyon, Lyon, France). Patient #1 was born in 1954 and, in the 1990s, after multiple episodes of otitis media, the appearance of hypogammaglobinemia, and bronchial dilatation, she was diagnosed with CVID. Immunoglobulin (Ig) replacement therapy consisted of Gammagard (Takeda Pharmaceuticals, Cambridge, MA, United States) due to suspected hypersensitivity to IgA. Other relevant comorbidities were as follows: villous atrophy leading to chronic malnutrition; inflammatory-bowel-disease-like enteritis and colitis (treated with adalimumab from June to July 2020 and with ustekinumab from June 2023 to July 2024); and regenerative nodular hyperplasia of the liver, which later progressed to cirrhosis. Additionally, in 2014 she was diagnosed with a gastric adenocarcinoma, which was treated by partial gastrectomy.

From 2014 onward, patient #1 (born in 1954) experienced over 40 episodes of C. jejuni or C. coli bacteremia despite receiving polyvalent immunoglobulin replacement therapy with IgG trough levels ranging from 9 to 11 g/L. These episodes were either clinically asymptomatic or only associated with mild symptoms (e.g., cellulitis of the limbs, diarrhea, or occasional fevers) and inflammatory biologic syndrome. The bacteremia episodes were initially several months apart but gradually increased in frequency. In 2024, the first episode of ascites occurred, with fluid that contained C. jejuni. Asymptomatic bacteremia episodes were not systematically treated with antibiotics. Symptomatic episodes were treated using antibiotics, including amoxicillin-clavulanic acid, azithromycin, or ertapenem, sometimes in combination with gentamicin. The antibiotics chosen were based on antimicrobial susceptibility testing results and/or ease of administration for outpatient ambulatory therapy. Treatment duration varied from 7 days to 6 weeks. Antibiotic suppressive treatment with azithromycin or oral decontamination capsules did not prevent recurrent Campylobacter bacteremia. All bacteremias were monomicrobial.

Patient #2 was born in 1947 and had been monitored for CVID since 2014. Her first episode of C. coli bacteremia occurred in 2016. Since her CVID diagnosis, patient #2 had undergone subcutaneous and intravenous substitution therapy to maintain IgG levels at 8–9 g/L. She received no additional treatment during the bacteremia episodes. She had a fractured left shoulder in 2017 (prosthesis inserted), a fractured left femur and humerus in 2020 (gamma nail and prosthesis inserted, respectively), and pain in her right hip with associated swelling since September 2021. Subcutaneous samples from the hip were positive for C. coli. Arthritis in the right hip was treated by removal of the gamma nail and two courses of ertapenem followed by imipenem. A fecal microbiota transplant was performed in July 2021 but stools have remained positive for C. coli since then. Similar to patient #1, clinical stabilization has occurred many times since 2016 but samples have remained positive for C. coli. Symptomatic episodes have been treated with antibiotics, including amoxicillin-clavulanic acid, azithromycin, gentamicin, and carbapenems (i.e., ertapenem and imipenem). As for patient #1 all bacteremias were monomicrobial.

To date, patient #2 still has C. coli infections; patient #1 died from pneumonia in December 2024.

Clinical sampling and antimicrobial susceptibility testing

A total of 35 isolates were obtained from patient #1 between 2014 and 2024, including 18 isolates of C. coli (designated the 1C pool of isolates in the present study) and 17 isolates of C. jejuni (designated the 1J pool of isolates). A total of 10 isolates of C. coli were obtained from patient #2 between 2019 and 2024 (designated the 2C pool of isolates). Each isolate was obtained from blood culture except for isolate 2C03 from patient #2, which was obtained from stools in 2022. All C. coli and C. jejuni were isolated on Columbia blood agar plates with 5% sheep blood (Thermo Fisher Scientific, Waltham, MA, United States) and incubated at 37 °C, following the Referentiel de Microbiologie Clinique (REMIC V7) published in 2022 by the French Society for Microbiology. A microaerobic atmosphere was maintained (79.7%N2, 7.1%CO2, 6.1%O2, and 7.1%H2) using an Anoxomat microprocessor (Mart Microbiology BV, Lichtenvoorde, the Netherlands). Species identities were confirmed using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (Bessède et al., 2011). Susceptibility to five routinely monitored antimicrobials (ampicillin, ciprofloxacin, erythromycin, tetracycline, and gentamicin) was assessed using the disk diffusion method, and resistance was also detected using WGS. In addition, minimum inhibitory concentrations (MICs) of three carbapenems (ertapenem, meropenem, and imipenem) were determined using Etest strips (bioMérieux, Marcy-l’Étoile, France). For both analyses, bacterial inoculums at 0.5 McFarland standard were subcultured on Mueller–Hinton agar supplemented with 5% defibrinated horse blood and 20 mg/L nicotinamide adenine dinucleotide (β-NAD; bioMérieux). Cultures were incubated for 24–48 h at 36 °C in a microaerobic environment, and data were collected based on the CASFM/EUCAST 2022 guidelines (Comité de l’antibiogramme de la Société Française de Microbiologie, 2022). MICs were measured in mg/L at the point where growth inhibition intersected the strip by two independent readers. For the disk diffusion method, inhibition zone diameters were measured via the SIRscan Auto (i2A, Montpellier, France) automatic system and confirmed manually. The C. jejuni ATCC 33560 reference strain was used for quality control.

WGS and molecular characterizations

WGS of the isolate genomes obtained from patients #1 and #2 were performed using Illumina sequencers (iSeq 100, HiSeq 4,000, NextSeq 500, and NovaSeq 6,000; Illumina, San Diego, CA, United States). Quality control of raw reads and read trimming were performed using FastQC v0.12.0 (Wingett and Andrews, 2018) and Sickle v1.33 (Joshi and Fass, 2011), respectively. Genomes were then assembled using SKESA v2.5.1 (Souvorov et al., 2018). The assembly qualities were checked based on the ECDC recommendations for Campylobacter species. Specifically, the assembled genome size should be within the range of 1.5–1.9 Mb, the N50 value should be 30,000 bp (there is no defined threshold), and the total number of contigs should be less than 500, each with a sequence length greater than 300 bp.

Molecular typing of isolates was performed using multilocus sequence typing (MLST) and core genome multilocus sequence typing (cgMLST) methods. Oxford PubMLST schemes (Cody et al., 2017) were downloaded (i.e., allele sequences and profiles), and the Nucleotide-Nucleotide BLAST v2.12.0 + command line tool (Altschul et al., 1997) was used to extract all loci. BLAST was also used to determine molecular antimicrobial resistances, together with various gene and mutation databases (e.g., NCBI, CARD, and ResFinder, as well as the in-house NRCCH Campylobacter resistance database). In the context of carbapenem resistance, we performed PorA protein structure homology modeling using the AlphaFold prediction tool (Jumper et al., 2021). Antimicrobial resistance-carrying plasmid sequences were identified using RFPlasmid tools v1.0 (van der Graaf- Bloois et al., 2021). Potential sources of contamination were identified using previously published data for C. jejuni and C. coli, as well as STRUCTURE software for population genetics inference (Hubisz et al., 2009). Briefly, 15 C. jejuni host-segregating markers of the poultry, ruminant and environment reservoirs and 259 C. coli SNPs of the poultry, ruminant and pig reservoirs were extracted from each clinical isolate in order to compute a score of attribution (Thépault et al., 2017; Jehanne et al., 2020).

Core genome comparisons of isolated strains

To obtain an overview of strain adaptation over time, core genomes from 1C, 1J, and 2C isolates were generated, and pairwise comparisons of single-nucleotide polymorphisms (SNPs) were performed. First, each genome was annotated using Prokka v1.14.6 (Seemann, 2014), and pangenomes were generated from annotations outputs using Roary v1.7.8 (Page et al., 2015). Core genomes were constructed from pangenomes by selecting genes shared by all isolates. Raw sequencing data from each isolate from the 1C, 1J, and 2C pool of isolates was aligned against the 1C, 1J, and 2C constructed core genome, respectively, using bwa v0.7.17 (Li, 2013) and SAMtools v1.19.2 (Li, 2011) with default parameters. BCFtools v1.19 (Li, 2011) was used to call variants with the default parameters. The pairwise comparisons were performed for each nucleotide position where a SNP was detected. Distances between each isolate (i.e., the number of different variants between two isolates) were computed using MEGA v11.0.13 (Tamura et al., 2021), and trees were visualized using iTOL v7 (Letunic and Bork, 2024). Finally, EggNOG-mapper v2.1.12 (Cantalapiedra et al., 2021) was used to classify each annotated gene into functional categories and identify biological mechanisms with the most genomic variability. In order to confirm that isolates from the present study originated from the same strains, core genomes were compared with a random selection of 160 additional genomes, namely 10 C. jejuni and 10 C. coli genomes from each year between 2017 and 2024. Raw reads were aligned against cgMLST Oxford scheme v1 (n = 1,343 genes), and neighbor-joining trees of SNP distances were constructed as previously described.

ResultsGenomic characteristics of isolates

Genomic data from all 45 isolates from both patients were obtained (summarized in Table 1 and complete data in Supplementary Table 1). For the C. coli isolates obtained from patient #1 (n = 18) and patient #2 (n = 10), the genomes had an average size of 1,715,904 bp (minimum genome size of 1,710,634, maximum genome size of 1,718,195 and GC = 30.7%) and 1,772,399 (minimum genome size of 1,761,156, maximum genome size of 1,776,391 and GC = 29.4%), respectively, similar to the C. coli reference strain ATCC 51798. The average number of contigs was 33 ± 7 and 61 ± 27, respectively. For patient #1, the total number of coding sequences from Prokka annotations of C. coli isolates was 1,775 ± 5, with a core genome of 1,709 genes and an accessory genome of 133 genes; for patient #2, there were 1,833 ± 18 coding sequences from C. coli isolates, with a core genome of 1,751 genes and an accessory genome of 146 genes. All C. coli isolates from this study were identified as complex clonal 828 strains (CC828) (Table 1). However there were minor changes in sequence types (ST) due to mutations in the aspA and gltA genes (Supplementary Table 1). In addition, patient #1 C. coli isolates were attributed to the poultry reservoir, whereas patient #2 isolates were attributed to the pig reservoir.

IdSampling dateSpeciesOrigin (a)ST/CC (b)Source (c)Molecular antimicrobial resistance detection (d)Carbapenem susceptibility (MIC in mg/L) (e)1C012022-02-14C. coliBlood825/828ChickenGyrA-T86I; tet(O)S1C022022-04-25C. coliBlood825/828ChickenGyrA-T86I; 23S-A2075G; tet(O)S1C032022-05-16C. coliBlood825/828ChickenGyrA-T86I; tet(O)S1C042022-11-25C. coliBlood825/828ChickenGyrA-T86I; tet(O); 16S-A1387GERT (1.5)1C052022-11-25C. coliBlood825/828ChickenGyrA-T86I; tet(O)ERT (1.5)1C062022-12-16C. coliBlood825/828ChickenGyrA-T86I; tet(O); 16S-A1387GERT (2)1C072022-12-16C. coliBlood825/828ChickenGyrA-T86I; tet(O); 16S-A1387GS1C082023-01-06C. coliBlood825/828ChickenGyrA-T86I; tet(O); 16S-A1387GS1C092023-03-10C. coliBlood825/828ChickenGyrA-T86I; tet(O)S1C102023-04-26C. coliBlood825/828ChickenGyrA-T86I; tet(O)ERT (2)1C112023-11-28C. coliBlood14,736/828ChickenGyrA-T86I; tet(O)S1C122024-01-09C. coliBlood14,732/828ChickenGyrA-T86I; tet(O)S1C132024-01-09C. coliBlood14,732/828ChickenGyrA-T86I; tet(O)S1C142024-01-09C. coliBlood14,732/828ChickenGyrA-T86I; tet(O)S1C152024-03-15C. coliBlood14,737/828ChickenGyrA-T86I; tet(O)S1C162024-03-15C. coliBlood14,737/828ChickenGyrA-T86I; tet(O)S1C172024-09-21C. coliBlood825/828ChickenGyrA-T86I; tet(O)S1C182024-09-21C. coliBlood825/828ChickenGyrA-T86I; tet(O)S1 J012014-03-26C. jejuniBlood21/21CattleGyrA-T86I + D90NS1 J022014-03-26C. jejuniBlood12,324/21CattleblaOXA581-G63T; GyrA-T86I + D90NS1 J032022-05-16C. jejuniBlood14,728/21CattleblaOXA193-G63T; GyrA-T86I; 23S-A2075G; tet(O)_plaS1 J042022-07-07C. jejuniBlood14,728/21CattleblaOXA193-G63T; GyrA-T86I; 23S-A2075G; tet(O)_plaS1 J052023-03-10C. jejuniBlood14,728/21CattleblaOXA193-G63T; GyrA-T86I; 23S-A2075G; tet(O)_plaS1 J062023-03-31C. jejuniBlood14,728/21CattleblaOXA193-G63T; GyrA-T86I; 23S-A2075G; tet(O)_plaS1 J072023-04-26C. jejuniBlood14,728/21CattleblaOXA193-G63T; GyrA-T86I; 23S-A2075G; tet(O)_plaS1 J082023-11-28C. jejuniBlood14,728/21CattleblaOXA193-G63T; GyrA-T86I; 23S-A2075G; tet(O)S1 J092024-01-09C. jejuniBlood14,728/21CattleblaOXA193-G63T; GyrA-T86I; 23S-A2075GS1 J102024-01-09C. jejuniBlood14,728/21CattleblaOXA193-G63T; GyrA-T86I; 23S-A2075G; tet(O)S1 J112024-01-09C. jejuniBlood14,728/21CattleblaOXA193-G63T; GyrA-T86I; 23S-A2075GS1 J122024-02-06C. jejuniBlood14,728/21CattleblaOXA193-G63T; GyrA-T86I; 23S-A2075G; tet(O)S1 J132024-02-06C. jejuniBlood14,728/21CattleblaOXA193-G63T; GyrA-T86I; 23S-A2075GS1 J142024-03-15C. jejuniBlood14,728/21CattleblaOXA193-G63T; GyrA-T86I; 23S-A2075GS1 J152024-04-03C. jejuniBlood14,728/21CattleblaOXA193-G63T; GyrA-T86I; 23S-A2075GS1 J162024-07-10C. jejuniBlood14,729/21CattleblaOXA193-G63T; GyrA-T86I; 23S-A2075GS1 J172024-09-21C. jejuniBlood14,728/21CattleblaOXA193-G63T; GyrA-T86I; 23S-A2075GS2C012019-06-07C. coliBlood14,731/828PigblaOXA193-G63T; GyrA-T86I; 23S-A2075G; tet(O)S2C022022-07-06C. coliBlood14,731/828PigblaOXA193-G63T; GyrA-T86I; 23S-A2075G; tet(O)S2C032022-10-04C. coliStools14,731/828PigblaOXA193-G63T; GyrA-T86I; 23S-A2075G; tet(O); 16S-G1464TS2C042023-01-05C. coliBlood14,731/828PigblaOXA193-G63T; GyrA-T86I; 23S-A2075G; tet(O); 16S-G1464TERT + MER (>32)2C052023-02-21C. coliBlood14,731/828PigblaOXA193-G63T; GyrA-T86I; 23S-A2075G; tet(O)S2C062023-03-27C. coliBlood14,730/828PigblaOXA193-G63T; GyrA-T86I; 23S-A2075G; tet(O); 16S-G1464TS2C072024-05-02C. coliBlood14,731/828PigblaOXA193-G63T; GyrA-T86I; 23S-A2075G; tet(O)S2C082024-06-04