This was a prospective cohort study conducted among residents of Jidong Community in Caofeidian District, Tangshan City, in north China. A total of 10,043 participants aged 18 years and above were recruited in 2014, of whom 9,078 participants signed informed consent forms and completed the baseline data and the first bio-specimen collection.

A preliminary experiment based on randomly selected 100 samples was conducted, and the seroprevalence rate of GI.2, GI.3, and GI.9 in participants was primarily determined to be 28.0%, 33.0%, and 29.0%, respectively. As a result, the minimal sample size of GI.2, GI.3, and GI.9 strains was respectively estimated to be 439, 347, and 418 using the following equation: [n = (Z_α ²)P(1 − P)/d²], in which P = 28.0%, 32.8%, and 29.2%, α = 0.05, Z_α = 1.96, and allowable error d = 0.15P. In the present study, 456 community residents were randomly selected for follow-up and biological sample collection until 2018. After excluding those lost in the follow-up, a representative sample of 449 individuals was included in the final analysis ( Figure 1 ).

Demographic data were collected by completing a set of combined self-administered questionnaires. The collected variables were categorized as follows: marital status, educational status, and monthly income per capita ( Table 1 ). Saliva samples were collected in 2016, while serum samples were collected annually from 2014 to 2018. All of the collected specimens were kept at -80°C for further experimental procedures.

The HBGA phenotypes of blood types (A, B, and O blood type), Lewis types (Le^a, Le^b, Le^x, and Le^y) of the saliva samples were determined by enzyme immunoassays (EIAs) using the corresponding monoclonal antibodies against individual HBGAs, as described earlier (Huang et al., 2005; Zhang et al., 2015). Briefly, boiled saliva samples at a dilution of 1:1,000 were coated on 96-well plates (Costar, Corning, NY, USA) overnight at 4°C. After blocking with 5% nonfat milk in phosphate buffered saline (PBS), the monoclonal antibodies specific for A (Z2A), B (Z5H-2), H (87-N) (Santa Cruz, CA, USA), Le^a (BG-5), Le^b (BG-6), Le^x (BG-7), and Le^y (BG-8) antigens (Signet Laboratories Inc., Dedham, MA, USA) were added. Then, horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG or IgM (Abcam, UK; 1:3,000) was added, followed by signal development with 3,3’,5,5’-tetramethylbenzidine (TMB). The cutoff value was set at OD450 = 0.2, a value of the mean of the background/blank wells adding a triple standard deviation (Huang et al., 2005; Zhang et al., 2015). The control and experimental wells in the 96-well costar plates were arranged in parallel repeats. As a measure of quality control, well-defined positive and negative saliva samples were added to each plate.

Due to the lack of cell culture and small animal infection models, an HBGA-based blockade assay, showing the ability of serum antibody to prevent NoV proteins binding to HBGA, has been accepted as a surrogate for neutralization worldwide (Cannon et al., 2009; Reeck et al., 2010; Malm et al., 2014; Jin et al., 2016; Sharma et al., 2017). HBGA blockade assays were based on a saliva-based HBGA P protein-binding assay (Yang et al., 2010). Saliva samples that demonstrated strong binding signals were selected as described previously. Saliva from Volunteer 23 (V23, positive to phenotype A) was used for GI.2 and GI.3 NoV, and saliva from Volunteer 36 (V36, positive to phenotype O) was used for GI.9 NoV (Dai et al., 2017; Xie et al., 2020). In brief, saliva samples (1:1,000) were coated onto plates. Then, GI.2, GI.3, or GI.9 P proteins were added at the concentration of 0.2–0.5 µg/mL. Anti-GI.2, GI.3, and GI.9 in-house-made mouse sera (1:3,000) were then used to detect the bound P proteins. HRP-conjugated goat anti-mouse IgG (1:3,000) and TMB substrate kit were used to detect the binding of P proteins.

An extra step was performed for HBGA blockade assay. For the blocking effects, a preincubation of P protein with serum samples for 1 h at 37°C was added to the saliva-coated plates. P proteins without preincubation with serum samples, which demonstrated OD450 values approximately 0.7–1.3, were used as a positive control. The blocking index was calculated using the following equation: 100 – (OD for wells with serum/OD for wells without serum) × 100. Finally, OD450 values were measured for each sample, and those samples that showed at least 50% of blockade were identified as positive for blockade antibody (Cannon et al., 2009; Reeck et al., 2010; Xie et al., 2020). Well-defined positive and negative serum samples for blockade assay were also included in each plate.

The distribution characteristics of seroprevalence and seroincidence were revealed using Pearson chi-square and Fisher’s precision probability tests. The seroprevalence and seroincidence were estimated using the exact Clopper–Pearson with 95% confidence intervals (95% CIs) of MedCalc software. The person-years for seroincidence of GI NoV depended on the years contributed by those participants from serum antibody negative to positive. The associated factors with seroprevalence of GI NoV were determined using the generalized estimating equation (GEE), while correlations between HBGA phenotypes and different GI NoV blockade antibody were analyzed using logistic regression. Pairwise comparison was used for comparison between two groups. The duration persistence and kinetic of anti-GI NoV blocking antibodies were estimated by a generalized linear model (GLM) based on the observed data from 2014 to 2018. Differences in antibody decay rates were shown as empirical P-values, calculated by the Fisher’s permutation test in the Stata software. Statistical analyses were conducted with IBM SPSS software version 26.0 (SPSS Inc., Chicago, IL, USA). A P-value <0.05 (two-sided) was considered to be statistically significant.

A total of 449 participants were included in the final analysis, of whom 200 (44.5%) were women, with a mean age of 40.3 ± 12.7 years. Most were Han (97.1%) and married (88.9%); 269 participants (59.9%) having a degree or higher were educated. In addition, 244 (54.3%) earned a monthly income per capita of between 3,001 and 5,000 RMB ( Table 1 ).

The seroprevalence of GI.2 varied from 14.7% to 18.5% as follows from 2014 to 2018: 18.0% (81/449, 95% CI: 14.6%–21.9%), 15.1% (68/449, 95% CI: 12.0%–18.8%), 16.0% (72/449, 95% CI: 12.8%–19.8%), 15.4% (69/449, 95% CI: 12.2%–19.0%), and 14.7% (66/449, 95% CI: 11.6%–18.3%), with the highest in 2014 and lowest in 2018. In contrast, seroprevalence of GI.3 varied from 35.0% to 38.3% as follows: 35.0% (157/449, 95% CI: 30.6%–39.6%), 37.0% (166/449, 95% CI: 32.5%–41.6%), 38.1% (171/449, 95% CI: 33.6%–42.8%), 38.3% (172/449, 95% CI: 33.8%–43.0%), and 38.8% (174/449, 95% CI: 34.2%–43.4%), with the highest in 2018 and the lowest in 2014. For GI.9, it varied from 17.6% to 22.0% as follows: 22% (99/449, 95% CI: 18.3%–26.2%), 19.8% (89/449, 95% CI: 16.1%–23.8%), 17.8% (80/449, 95% CI: 14.4%–21.7%), 17.6% (79/449, 95% CI: 14.2%–21.4%), and 18.7% (84/449, 95% CI: 15.2%–22.6%), with the highest in 2014 and the lowest in 2017 ( Figure 2 ). There was no significant variation in the seroprevalence of the three strains during the 5-year period (P > 0.05, Figure 2 ). The demographic characteristics of the seroprevalence of GI.2, GI.3, and GI.9 NoVs in the study population were not statistically significant at baseline in 2014 (P > 0.05, Table 1 ).

Individuals aged 50–59 years were significantly associated with the prevalence of GI.3 (OR = 1.757, P = 0.033) and GI.9 NoV (OR = 2.039, P = 0.022) during 2014–2018. In addition, individuals with married status were negatively associated with the prevalence of GI.3 (OR = 0.461, P = 0.002) as shown in Table 2 .

During the 4-year follow-up, 1,506, 1,130, and 1,449 person-years for participants with initial antibody negative were followed, and a total of 149, 241, and 152 participants showed NoV seroconversion for GI.2, GI.3, and GI.9 NoV, respectively. Thus, the seroincidences of the studied GI NoVs were 10 (95% CI: 8–12), 21 (95% CI: 19–24), and 11 (95% CI: 9–12) per 100 person-years for GI.2, GI.3, and GI.9 NoV between 2014 and 2018, respectively. Each observational year showed a stable seroincidence over the period ( Table 3 ).

A significant difference was found in seroincidence of GI.9 when all age groups were compared (P = 0.012, Table 3 ). It was relatively low at 5.0/100.0 person-years in those aged 18–29 years but gradually increased to reach a peak of 12.0/100.0 person-years in the age group 50–59 years, which later declined to 8.0/100.0 person-years in those aged 60–80 years ( Table 3 ). The seroincidence of GI.2 and GI.3 also exhibited similar age trends but without statistical significance. The seroincidence of GI.3 was significantly low in the married population compared to the unmarried (19.0/100.0 person-years vs. 37.0/100.0 person-years, P < 0.001). However, the seroincidence of GI.2 exhibited no statistical significance with any of the sociodemographic variables, as illustrated in Table 3 .

Among 449 participants, 196 (43.7%), 302 (67.3%), and 182 (40.5%) respectively tested at least once positive to GI.2, GI.3, or GI.9 during 2014–2018. There was a significant difference between marital status for GI.3 NoV between GI NoV antibody positive and negative groups. No other significant difference of demographic distribution was observed ( Table 4 ).

The distribution of HBGA phenotypes in the participants was as follows: phenotype A, 31.2%; phenotype B, 18.0%; phenotype AB, 19.4%; phenotype O⁺ (group O with Rh antigen positive), 18.5%; and 12.9% for O^- phenotype (group O with Rh antigen negative). In terms of the Lewis phenotype, 12.9% of the individuals were Le^a+/Le^x+, 56.1% were Le^b+/Le^y+, and 31.0% were Le^a+b+/Le^x+y+. Majority of the participants (87.1%, 391/449) were classified as secretors ( Table 5 ).

Compared to the group with antibody negative to GI.2 NoV, the positive group had a significantly larger proportion of A phenotype (31.1%) but a lower proportion of O^- phenotype (7.1%; P = 0.001), and similar results were observed for GI.3 NoV (P < 0.001, Table 5 ). For antibody-positive group to GI.2 and GI.3 NoV, a higher proportion of Le^b+/Le^y+ (P_GI.2 = 0.005, P_GI.3 = 0.010) and a lower proportion of nonsecretor (P_GI.2 = 0.001, P_GI.3 = 0.003) were observed, indicating the individuals with Le^a+/Le^x+ or the nonsecretor less susceptible to GI.2 or GI.3 NoV ( Table 5 ). However, no statistical difference of HBGA distribution was found across groups for GI.9 (P = 0.245, Table 5 ).

Significant differences were observed in positive rates of blockade antibody among participants with different HBGA phenotypes for GI.2 and GI.3, but not for GI.9 NoV. For GI.2 NoV, the positive rate of blockade antibody of individuals with O^- phenotype was significantly lower than that of individuals with B (24.1% vs. 56.8%, P < 0.05) and O⁺ phenotype (24.1% vs. 51.8%, P < 0.05). For GI.3 NoV, blockade antibody positive rate of individuals with O⁺ phenotype was the highest (88.0%) and followed with those with A (73.6%), AB (59.8%), B (55.6%), and O^- phenotype (50.0%) with significant differences ( Figure 3A ). Positive rates were significantly higher for both individuals with Le^b+/Le^y+ (GI.2: 45.6% and GI.3: 69.0%) and Le^a+b+/Le^x+y+ (GI.2: 48.2%, GI.3: 71.2%) phenotypes than that for the Le^a+/Le^x+ phenotype (GI.2: 24.1%, GI.3: 50.0%, Figure 3B ), and similar findings were observed for secretor status ( Figure 3C ).

Compared to the O^- phenotype, participants with A and O⁺ for GI.2 and GI.3 NoV and phenotype B for GI.2 NoV showed elevated susceptibility. Infection risk of GI.2 and GI.3 strains was also significantly higher among those with Le^b+/Le^y+ (GI.2: OR = 2.843; GI.3: OR = 2.363) and those with Le^a+b+/Le^x+y+ (GI.2: OR = 3.139; GI.3: OR = 2.371). Secretors had a significantly higher risk of infection with GI.2 (OR = 2.950) and GI.3 (OR = 2.366) NoV than that of the nonsecretors. However, no significant association susceptibility to GI.9 strain was observed among individuals with different HBGA phenotypes ( Table 6 ).

In the initial antibody positive population, repeated seroconversion (initially positive to negative and turned positive again) during follow-up was defined as reinfection. A total of 81, 157, and 99 individuals were respectively found to be seropositive of anti-GI.2, GI.3, and GI.9 NoV blockade antibodies in 2014. Among them, 66, 104, and 72 entered the anti-GI.2, GI.3, and GI.9 NoV blockade antibody persistence and seronegative conversion rate analysis, respectively, and the remaining 15, 53, and 27 participants with sero-reinfection were excluded to avoid reinfection affecting the accuracy of the results. During the follow-up period, 5-year persistent positive rates of 29.9% (47/157), 29.3% (29/99), and 12.3% (10/81) against GI.3, GI.9, and GI.2 NoV were observed ( Figure 4 ).

The overall negative conversion rates for GI.2, GI.3, and GI.9 strains of blockade antibodies in the initially seropositive population were 84.8% (55/66), 54.8% (57/104), and 59.7% (43/72), respectively. Most of the antibody reversal occurred within the first year after seropositivity, with anti-GI.2 and GI.9 NoV antibodies having the highest serum reversal rate of 50% (33/66) and 34.7% (25/72) after 1 year in 2015. However, anti-GI.3 NoV antibodies had the highest negative conversion rates of 34.7% (25/72) in the fourth year. The three strains exhibited statistically significant differences in serum reversal rates across years ( Supplementary Table S1 ).

The blockade rates of GI NoVs antibodies showed an overall decreasing trend from year to year with an overall mean decay rate of -5.9%/year, -3.6%/year, and -4.0%/year for GI.2, GI.3, and GI.9, respectively. Notably, the antibodies of GI.2 and GI.9 strains both declined more rapidly in the first 1–2 years after positivity, with decay rates of -10.2%/year (95% CI: -12.6%/year, -7.8%/year) for GI.2 and -6.6%/year (95% CI: -8.9%/year, -4.3%/year) for GI.9 with a statistically significant trend (P < 0.001, Figure 5 ). However, for the GI.3 strain, the decay rate of the blocking effect after infection in the first 2 years was similar to that second 2 years, at -4.7%/year and -3.9%/year (95% CI: -5.4%/year, -2.3%/year). The least persistent antibody among the three GI NoV strains was estimated to be GI.2 at approximately 2.3 years, while anti-GI.3 and GI.9 NoV antibodies lasted for approximately 4.2 and 4.8 years, respectively ( Figure 5 ).