Detection and prevalence of SARS-CoV-2 co-infections during the Omicron variant circulation in France

The emergence and global spread of SARS-CoV-2 variants of concern (VOCs) with increased transmissibility and/or immune escape, notably Delta (B.1.617.2) and Omicron (B.1.1.529), created periods of intense co-circulation of genetically divergent lineages. In late 2021 and early 2022, France experienced substantial overlap between Delta and Omicron BA.1, followed by co-circulation of Omicron sublineages BA.1 and BA.2. This context raised critical questions about the occurrence and detection of co-infections with distinct lineages and their clinical impact. Given that Delta and Omicron have been associated with different disease severities, characterizing the prevalence and clinical spectrum of co-infections, including BA.1/BA.2, was a key objective. The study aimed to (i) evaluate technical and bioinformatic approaches for detecting co-infections by WGS, and (ii) estimate the prevalence and clinical outcomes of co-infections during the fifth wave in France.

The authors note that detecting SARS-CoV-2 co-infections via WGS is challenging due to primer amplification biases, potential contamination, and the lack of unbiased, validated bioinformatic methods. Prior reports of Delta/Omicron co-infections relied largely on manual inspection of mutation frequencies or de novo assemblies, the latter being computationally intensive. Literature also highlights known amplicon-related artifacts affecting variant calls in certain regions and the risk of chimeric consensus sequences due to primer biases. Few studies had systematically quantified co-infection prevalence, particularly BA.1/BA.2, on large datasets at the time of this work. This study builds on those observations by benchmarking primer sets and introducing an agnostic lineage-based co-infection detection pipeline.

Study design comprised two components: (1) experimental mixes of cultured Delta (B.1.617.2) and Omicron (BA.1) isolates at defined ratios to benchmark primer sets and assess detection biases; and (2) population genomic surveillance to estimate co-infection prevalence and clinical impact. Experimental benchmarking: - Prepared 11 Delta:Omicron mix ratios: 0:100, 10:90, 20:80, 30:70, 40:60, 50:50, 60:40, 70:30, 80:20, 90:10, and 100:0. Viral loads had median 4.2 log copies/mL (RQ 0.4). - Sequencing used four amplicon primer sets: ARTIC V4, ARTIC V4.1, Midnight V1, Midnight V2. Each mix was extracted and sequenced in duplicate; genome coverage >98% with median depth ~2276×. - Evaluated detection rates of Delta- and Omicron-specific mutations using a curated list (based on covariants.org). Assessed primer biases and frequency estimation accuracy. Consensus sequence artifacts (chimeras) were characterized by lineage-specific mutation patterns depending on primer set. - Additional low-percentage mixes (Delta:Omicron 1:19 and 5:95) were sequenced in 10 replicates to test co-infection detection at low minor lineage frequencies and propensity for chimeric consensus sequences. Population sampling and case selection: - Samples sequenced at the National Reference Center (Hospices Civils de Lyon, HCL) originated from: (i) systematic sequencing of hospitalized patients and HCL healthcare workers; and (ii) random weekly nationwide Flash surveys coordinated by EMERGEN. Flash surveys request a fraction of positives from laboratories for one day per week. - Timeframe: December 6, 2021 (week 49-2021) to February 27, 2022 (week 08-2022). A total of 23,242 samples were sequenced with ARTIC V4/V4.1; high-quality WGS (coverage >90%) were obtained for 21,387 samples (16,230 Flash; 5,167 HCL/peripheral hospitals). - To assess clinical correlates, demographic and clinical data were compiled for outpatients, hospitalized patients, and healthcare workers. Bioinformatics pipeline and co-infection detection: - Reads processed with a standardized pipeline: adapter/quality trimming using cutadapt (reads >30 bp retained); alignment to the SARS-CoV-2 reference genome with Minimap2; variant calling and post-processing; primer-end cleaning with samtools mpileup. - Generated a lineage mutation database from 100 randomly sampled sequences per Pangolin lineage from GISAID (snapshot Feb 2022). Two detection strategies were used: (i) curated lists of clade-defining mutations for Delta and Omicron; and (ii) an agnostic approach identifying a dominant (main) and a minor (secondary) lineage by matching observed variants to lineage-specific mutation sets while excluding shared mutations. - The co-infection detection script quantified main and secondary lineage mutation match ratios; samples with a positive secondary lineage ratio were flagged as potential co-infections. - Quality assurance: All samples with a detected secondary lineage (n=61; 0.29%) were re-extracted and resequenced in duplicate to rule out contamination. Eight were not confirmed (no secondary lineage on duplicate) and excluded; 53 were confirmed co-infections. Statistical analysis: - Continuous variables summarized as mean±SD or median (IQR) and compared with Kruskal–Wallis or Mann–Whitney tests. Proportions compared with chi-squared or Fisher’s exact tests. ICU admission related to VOC, sex, and age analyzed by binomial logistic regression (Wald Chi-square). Significance threshold p<0.05. Analyses performed in R 4.0.5.

- Primer performance and bias: >90% of Delta-specific mutations were detected across all mixes and primer sets, while Omicron-specific mutation detection was lower when Omicron was the minor component (e.g., 26% in 90:10 Delta:Omicron mixes with Midnight primers). ARTIC V4 was the least biased for Delta/Omicron co-infection detection and relative frequency estimation. Delta frequencies were overestimated in mixes where Omicron was below ~30%, particularly with Midnight primers. - Chimeric consensus sequences: Majority-rule consensus calling produced artifactual Delta–Omicron chimeras in several mixes, with primer-dependent patterns (e.g., Omicron sequences bearing Delta-specific mutations such as S:K1452R; 3′ genome regions from Delta with Midnight). In natural co-infections, 65% (69/106) of confirmed duplicates yielded chimeric consensus sequences; all samples with a minor lineage >38% were chimeric, with a median of 4 minor lineage-specific mutations present in the consensus. - Agnostic co-infection detection: The lineage-agnostic script correctly identified main and secondary lineages across mixes. Main lineage match ratios were >0.9; secondary ratios ranged 0.216–0.941, with the lowest ratios linked to primer bias. In additional 1:19 and 5:95 mixes, co-infections were detected in nearly all replicates; one 1:19 replicate with Midnight V1 was misclassified as non–co-infected. - Surveillance results and prevalence: Of 21,387 high-quality sequences, 61 (0.29%) initially showed a secondary lineage; after duplicate confirmation, 53 co-infections were validated: 28 Delta/Omicron (BA.1), 1 Delta/Omicron (BA.2), and 24 BA.1/BA.2. Estimated prevalence: 0.18% for Delta/Omicron and 0.26% for BA.1/BA.2 during periods of co-circulation. Peak Delta/BA.1 co-infection prevalence occurred in weeks 51–52 of 2021 (0.31% and 0.25%). BA.1/BA.2 co-infections reached 0.78% at peak co-circulation (week 08-2022). - Replicate concordance: Minor lineage relative abundance was strongly correlated between duplicates (Pearson r=0.92, p<2e-16); median minor lineage abundance was 20% (IQR 26.25). - Clinical outcomes: ICU admission rates among hospitalized patients were higher in Delta/Omicron co-infections (15.38%) compared with Omicron (1.64%–1.95%) and Delta (4.81%–5.56%). No ICU admissions were observed among BA.1/BA.2 co-infections. Among 53 co-infected patients, 21 (39.6%) were unvaccinated; among vaccinated (32/53, 60.4%), most had two doses (56.3%). - Recombinant suspicion: In 3/53 (5.6%) co-infected samples, lineage-specific allele frequency distributions suggested potential recombinants with differing relative abundances within samples. - Contamination control: 8/61 initially flagged co-infections were not confirmed on duplicate sequencing and were attributed to probable contamination during the first sequencing run.

This study demonstrates that SARS-CoV-2 co-infections occurred during periods of intense co-circulation of divergent lineages (Delta with Omicron BA.1; Omicron BA.1 with BA.2), albeit at low prevalence (<1%). Accurate detection is complicated by amplicon primer biases that can both skew variant frequency estimates and generate chimeric consensus sequences. By benchmarking four primer sets and deploying an agnostic lineage-based detection pipeline with duplicate confirmation, the authors provide robust estimates of co-infection prevalence and characterize technical pitfalls in WGS-based detection. Clinically, Delta/Omicron co-infections were associated with higher ICU admission rates than single-lineage Omicron or Delta, suggesting that concurrent infection with divergent variants could impact disease severity; however, BA.1/BA.2 co-infections did not show increased ICU admissions. The strong replicate concordance in minor lineage abundance supports the reliability of the detection pipeline. Evidence of potential within-sample recombinants in a minority of co-infections underscores the evolutionary risk posed by co-infection events. Overall, the findings stress the necessity of unbiased detection methods, careful QC (including duplicate sequencing), and context-aware interpretation of WGS data during variant transitions.

The study provides a validated framework for detecting SARS-CoV-2 co-infections during periods of variant co-circulation, combining primer set benchmarking with an unbiased, lineage-agnostic bioinformatic pipeline and duplicate confirmation. Co-infections were rare but present (0.18% Delta/Omicron; 0.26% BA.1/BA.2), with Delta/Omicron co-infections associated with higher ICU admission rates than single-variant infections. The work highlights primer-specific biases and the risk of chimeric consensus sequences, emphasizing the need for cautious interpretation of WGS results. Future research should refine detection thresholds for low-abundance minor lineages, incorporate sequencing methods less prone to amplification bias (e.g., hybrid capture or metagenomics), enhance contamination controls, and further investigate the clinical implications and evolutionary dynamics (including recombination) of co-infections.

- Sensitivity threshold: Co-infections with very low minor lineage abundance (<5%) or involving closely related lineages (fewer than ~6 specific mutations distinguishing them) may not be detected by the pipeline. - Short follow-up for BA.1/BA.2: The assessment period for BA.1/BA.2 was limited as BA.2 frequency was still rising in France, potentially underestimating prevalence. - Incomplete metadata: ICU admission and vaccination status were missing for some Flash survey cases, limiting clinical characterization. - Methodological constraints: Only amplicon-based WGS was used; metagenomic or hybrid-capture approaches (less prone to primer bias) were not evaluated. - Primer bias and chimeras: Amplicon primer sets can induce preferential amplification and chimeric consensus sequences, affecting frequency estimates and lineage calling despite mitigation strategies. - Potential contamination: Although duplicates were used to confirm co-infections, initial runs showed some contamination events, underscoring the need for stringent QC.