Increased GII.3[P12] norovirus outbreaks and viral whole genome analysis in Beijing, China during 2021 and 2023

In this study, samples and epidemiological information were collected from the AGE outbreaks surveillance network in Beijing, along with laboratory data, including pathogen detection results, pathogen grouping, and cycle threshold (Ct). AGE cases were defined as patients with diarrhea (≥ three loose stools within 24 h) and/or vomiting (≥ 1 episode). An outbreak was defined as ≥ 3 AGE cases caused by NoV within 72 h that were epidemiologically linked (through close contact or common exposure), with ≥ 2 laboratory-confirmed. All the samples collected were stored at − 80 ℃.

NoV genomic RNA was extracted from 140 µl of a 10% (w/v) fecal suspension in phosphate-buffered saline using the QIAamp Viral RNA Mini kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. Approximately 60 µl of nucleic acid eluent was collected into a nuclease-free tube. Real-time RT-PCR was employed to group pathogens and determine Ct values.

The QIAGEN One-Step RT-PCR kit (Qiagen, Hilden, Germany) was used to amplify the partial VP1 and RdRp regions of NoV, with primers MON432/GISKR for GI and MON431/G2SKR for GII [20, 21]. The obtained sequences were assembled using Seq Man 12.2, and genotypes were determined with NoV Genotyping Tool 2.0 (http://www.rivm.nl/mpf/Norovirus/Genotyping Tool).

Sequences with the GII.3[P12] (only one per outbreak) genotyping were selected, capturing their identifiers, sampling times, and regions. Reference sequences were downloaded from the National Center for Biotechnology Information (NCBI). The best nucleotide substitution model was assessed using ModelFinder 2.2.0 [22]. We employed three tree priors (constant-size, Bayesian skyline, and exponential growth) and two clock models (strict and relaxed molecular clock) to calculate marginal likelihoods with path sampling [23] (Table S2, S3). We performed phylogenetic analysis to identify GII.3[P12] subtypes circulating in Beijing, and date the most recent common ancestor (TMRCA) by using BEAST v1.10.4 [24]. Markov chain Monte Carlo (MCMC) chains were iterated 2 × 10⁸ times, sampling every 20,000 steps. After discarding the first 10% as burn-in, the convergence of continuous parameters was evaluated by effective sample size (ESS) in Tracer 1.6, and a value greater than 200 was acceptable. The Maximum Clade Credibility (MCC) tree was visualized using FigTree 1.4.2 [25]. Nucleotide intra-sequence similarity and inter-subtype similarity were computed using BioEdit 7.1.3.

To investigate the spatial dynamics of GII.3[P12] NoV in Beijing, we conducted a phylogeographic analysis using an asymmetric continuous-time Markov chain with a Bayesian stochastic search variable selection (BSSVS) model in BEAST 1.10.4, based on GII.3[P12] sequences clustered into different clades [24, 26]. The regions of these sequences were chosen and coded as discrete states. Phylogeographic analysis, using posterior-simulation estimates, demonstrated that both clade 2 and clade 3 exhibited the best-fit geographical diffusion under the HKY substitution model with a strict molecular clock. The diffusion pathways were summarized using spreaD3 0.9.6 [27].

Samples from different evolutionary branches were selected for whole-genome sequencing based on subtype identification results. In each outbreak, the sample with the lowest Ct value was prioritized for testing, but multiple samples might be sequenced to ensure whole-genome acquisition. Viral RNA was extracted from 10% (w/v) stool suspensions using the QIAamp Viral RNA Mini kit (Qiagen, Hilden, Germany). RNA from each stool sample was reverse transcribed using LunaScript^® Reverse Transcription SuperMix Kit (New England Biolabs, Ipswich, MA, USA), and complementary DNA (cDNA) was amplified using 22 newly designed overlapping PCR primer sets (Table S1) to cover the whole genome. Each reaction included 12.5 µL Q5^® Hot Start High-Fidelity 2× Master Mix (New England Biolabs, Ipswich, MA, USA), 3.0 µL F1(odd number)mix/F2༈even༉mix, 5.9 µL H₂O, and 3.6 µL cDNA. After incubating at 98 °C for 30 s to inactivate RNase, each reaction underwent 35 cycles of denaturation at 98 °C for 15 s, annealing at 65 °C for 5 min, and then saved at 4 °C. Amplifications were purified using AMPure XP Beads (Beckman Coulter, Brea, CA, USA) and quantified using the Qubit^® 3.0 Fluorometer (Qiagen, Hilden, Germany). Genomic DNA tagging and fragmentation were performed with the Nextera^® XT Library Prep Kit (Illumina, San Diego, CA, USA), followed by library amplification using the Nextera^® XT Index Kit v2 (Illumina, San Diego, CA, USA). The amplified library was purified, quantified, and combined into one tube, diluted to a final concentration of 1.0 ng/µL. Sequencing was performed on the Illumina MiniSeq platform. Based on the downloaded reference sequences (GenBank accession number: OP901693), the CLC Genomics Workbench 23.0 was used to assemble short reads to obtain the whole-genome sequence. In this study, all nucleotide and amino acid positions were referenced to the same sequence (GenBank accession number: OP901693).

For samples failing to obtain whole-genome sequences via NGS, we selected those with fewer gaps, redesigned primers for gaps, and amplified them with the QIAGEN^® OneStep RT-PCR Kit (Qiagen, Hilden, Germany), acquiring the missing sequence fragments through Sanger sequencing.

The whole-genome sequences of GII.3[P12] NoV downloaded from the NCBI database were used as the reference and integrated with those obtained in this study to form Dataset 3. Information was collected, including sequence identifiers, sampling times, and locations. The optimal nucleotide substitution model was evaluated using ModelFinder 2.2.0. The BEAST 1.10.4 was utilized to construct a MCC Tree for phylogenetic analysis and estimation of TMRCA [24]. BioEdit 7.1.3 was used to calculate the intra-sequence and inter-genotype nucleotide similarities. We constructed maximum likelihood trees to enhance the reliability of our cluster classification, assessing their robustness with 1000 bootstrap replicates. Additionally, to investigate the possibility of intra-genotypic recombination among different clades, we also built trees based on the complete VP1 and RdRp sequences, respectively.

Both the obtained nucleotide sequences and reference strains were translated into amino acid sequences for mutation analysis using MEGA X, and amino acid mutation sites were visualized using WebLogo (https://weblogo.berkeley.edu/logo.cgi).

SPSS 19.0 was used to compare differences in season, region, setting, and cases’ gender distribution caused by various subtypes of GII.3[P12] via χ² Test or Fisher’s exact method. The Bonferroni correction was applied for further pairwise comparisons when p < 0.05. All tests were conducted at a significance level of α = 0.05, and results were considered statistically significant if p < 0.05. According to the climatic characteristics of Beijing, the seasons are divided as follows: spring (from March to May), summer (from June to August), autumn (from September to November), and winter (from December to February) [28]. Under the education system in China, the various educational stages are divided as follows: kindergarten (ages 3–6), primary school (ages 6–12, Grades 1–6), and comprehensive schools, which refer to the “nine-year consistent program” schools that includes both primary and junior secondary education (ages 6–15, Grades 1–9). Secondary education is divided into junior secondary (ages 12–15, Grades 7–9) and senior secondary (ages 15–18, Grades 10–12).

In Beijing, a total of 665 NoV AGE outbreaks were reported from August 2021 to July 2023, of which 343 were genotyped. GII.3[P12] (34.40%, 118/343) was the most dominant. GII.4[P16], GII.4[P31], GII.17[P17], mixed, GII.2[P16] were also prevalent, accounting for 13.12% (45/343), 12.84% (44/343) 8.16% (28/343), 7.30% (25/343), 7.00% (24/343) separately. When analyzed by surveillance year, GII.3[P12] accounted for 40.00% (42/105) genotyped outbreaks in 2021–2022 and 31.93% (76/238) in 2022–2023 (Fig. 1).

The MCC tree was constructed based on 118 VP1-RdRp sequences from 118 GII.3[P12] outbreaks (Fig. 2). Phylogenetic analysis revealed three evolutionary clades, with the dominant GII.3[P12] norovirus strains circulating in Beijing clustered in clade 2 and clade 3. Clade 1 contained only one sequence detected in 2021. Clade 2 included sequences from 2021 to 2023, with those detected in 2022 being dominant (62.71%, 37/59). In clade 3, 57 sequences were detected in 2023, with only one in 2022. The internal nucleotide similarity of the 59 sequences in clade 2 was 98.67%–100.00%, that of the 58 sequences in clade 3 was 97.91%–100.00%, and the nucleotide similarity between clade 2 and clade 3 was 96.39%–98.29%. Bayesian inference showed that the TMRCA of clades 1, 2, and 3 was 2011.45, 2013.46, and 2012.17, respectively. In clade 2 and clade 3, the divergence times of the Beijing strains are 2017.14 and 2016.65, respectively.

GII.3[P12] NoV outbreaks predominantly occurred in spring (65.26%, 77/118), followed by winter (25.42%, 30/118), summer (8.47%, 10/118), and autumn (0.85%, 1/118). These outbreaks were distributed across 10 districts in Beijing, with 67 (56.78%) in urban areas (Chaoyang, Dongcheng, Fengtai, and Xicheng) and 51 (43.22%) in suburban areas (Changping, Daxing, Fangshan, Jingkai, Shunyi, and Tongzhou). Most GII.3[P12] NoV outbreaks occurred in kindergartens (78.81%, 93/118) and primary schools (18.64%, 22/118), while fewer occurred in comprehensive schools (1.70%, 2/118) and middle schools (0.85%, 1/118). The transmission modes of 116 GII.3[P12] NoV outbreaks were identified: person-to-person transmission was most common (99.14%, 115/116), followed by foodborne transmission (0.86%, 1/116).

Characteristics of outbreaks caused by strains of clade 2 and clade 3 were compared, except for clade 1 (containing only one strain) (Table 1). All clade 2 outbreaks occurred in winter-spring (100%, 59/59), whereas 17.24% (10/58) of clade 3 outbreaks occurred in summer-autumn. Clade 2 outbreaks predominantly occurred in kindergartens (89.83%, 53/59), while clade 3 outbreaks occurred in kindergartens (68.97%, 40/58) and primary schools (27.59%, 16/58) (p_seasons= 0.001, p_settings = 0.009 and p_age< 0.001).

The results of spatiotemporal analysis revealed that from August 2021 to July 2023, the primary migration hubs for clade 2 were Chaoyang (4 out-migrations, 3 in-migrations), Tongzhou (4 out-migrations, 2 in-migrations), and Daxing (2 out-migrations, 2 in-migrations) (Fig. 3A). For clade 3, key transmission centers were Chaoyang (3 out-migrations, 1 in-migration) and Changping (3 out-migrations, 1 in-migration) (Fig. 3B).

We obtained 82 GII.3[P12] NoV whole-genome sequences (7444 bp) from 72 outbreaks, with ORF1 (5049 bp), ORF2 (1647 bp), and ORF3 (757 bp). The MCC tree showed that these sequences were clustered into three clades (2 of clade 1, 38 of clade 2, and 42 of clade 3) (Fig. 4), corresponding to the clustering in the MCC tree based on partial VP1-RdRp sequences. For clade 1, all sequences detected in 2021 and showed the highest sequence similarity to strains from Russia in 2022 (GenBank: OP901693), with nucleotide identities ranging from 98.49% to 98.53%. In clade 2, most sequences (20/38, 52.63%) were detected in 2022, sharing 97.60%–98.10% with strains from Beijing, China in 2018(GenBank: OM373562). For clade 3, all sequences detected in 2023, and showed the highest sequence similarity to strains from Taiwan, China in 2020 (GenBank: ON569431), with nucleotide identities ranging from 98.00% to 100.00%. Among the three clades, nucleotide similarity varied from 96.30% to 96.90% between clades 2 and 3, 96.50% to 96.90% between clades 1 and 2, and 96.90% to 97.10% between clades 1 and 3. Bayesian inference showed that the TMRCA of clades 1, 2, and 3 was 2015.66, 2016.56, and 2017.71, respectively. In clade 1, clade 2 and clade 3, the divergence times of the Beijing strains are 2018.81, 2019.61 and 2019.17, respectively.

The 82 whole-genome nucleotide sequences from this study were converted into amino acid sequences for comparison of the VP1 and RdRp genes with reference sequences. The results showed that the three known HBGA binding sites of GII.3 remained conserved (Table 2). In the ORF1 region of GII.3[P12] NoV, 13 amino acid mutations were identified: 3 in P48 gene (aa 1–330), 5 in P22 gene (aa 697–875), 1 in VPg gene (aa 876–1008), 2 in Pro gene (aa 1009–1189), and 2 in RdRp gene (aa 1190–1699) (Table 3). Sequences in clade 1 exhibited 5 amino acid mutations: V779I, D870G, K1004R, I1057V, and I1521V (Fig. 5A). Sequences in clade 2 exhibited 8 amino acid mutations: A21V, S195L, R278K, V779I, A782V, A791V, I850T, P1051S, V1091A, and S1571T (Fig. 5B). Sequences in clade 3 exhibited only 1 amino acid mutation: T701I (Fig. 5C).