We used samples from the Community Awareness Resources and Education (CARE) project developed by the Ohio State University Center for Population Health and Health Disparities (CPHHD) funded by the National Institutes of Health (NIH) (32). Serum samples were collected from women scheduled for a routine Papanicolaou (Pap) smear test at 17 participating health clinics located throughout the Appalachian Ohio region between January 2006 and December 2008 (n=1131) (33). Appalachian Ohio lies in the southern and eastern parts of Ohio, comprising 32 counties. The area is known to have elevated levels of cancer incidence and mortality in contrast to non-Appalachian regions (34–36) and has been categorized as an underserved and unique demographic by the National Cancer Institute (NCI). For this study, a total of 136 serum samples were randomly selected to include 41 samples from each group and all 13 samples from women for which cervical intraepithelial neoplasia (CIN) grade is available. These included 82 samples from women with normal cervical cytology, of which 41 were positive for at least one of four high risk HPV types (HPV16, 18, 33, or 45), and 41 had no detectable HPV. The remaining samples (n=54) were from women with abnormal cervical cytology of which 13 underwent colposcopy and biopsy and for which CIN grade is available. The classification of abnormal cervical cytology was done according to the 2001 Bethesda System for Reporting Pap Smear Results (37). Cases with abnormal cervical cytology were classified as having atypical squamous cells of undetermined significance (ASC-US), atypical glandular cells (AGC), low-grade squamous intraepithelial lesions (LSIL), high-grade squamous intraepithelial lesions (HSIL), or carcinoma. Serum samples were collected using a standardized sample collection protocol and stored at -70°C until use. Written informed consent was obtained from all subjects under institutional review board approval. Age and race information were collected using a self-administered questionnaire prior to the Pap smear and following consent.

HPV DNA testing was done on all cytology samples (N=136) collected in specimen transport medium (Qiagen, Valencia, CA) after shipping to the Centers for Disease Control and Prevention (CDC) as previously described (37). Briefly, DNA was extracted from 150 µL of each sample using the MagNA Pure DNA kit III (Roche, Indiannapolis, IN). HPV typing was done using the Linear Array (LA) HPV Genotyping Assay (Roche, Indiannapolis, IN), which detects 37 HPV types (6, 11, 16, 18, 26, 31, 33, 35, 39, 40, 42, 45, 51, 52, 53, 54, 55, 56, 58, 59, 61, 62, 64, 66, 67, 68, 69, 70, 71, 72, 73, 81, 82, 83, 84, 89, and IS39). Because of known cross-reactivity between HPV33, 35, and 58 and the XR (52) probe, the presence of HPV52 was confirmed by an HPV52 quantitative PCR assay in XR-positive samples that are also positive for any of the cross-reactive HPV types (38).

Custom HPV high-density nucleic acid programmable protein arrays (HD-NAPPA) were produced as previously described (25, 31, 39) with modifications described here. HD-NAPPA is a silicon nanowell version of the NAPPA technology (40, 41) that allows high throughput and rapid in situ antigen expression and display for Ab detection in the settings of cancer (42–44), infectious diseases (45, 46), and autoimmunity (47, 48). In HD-NAPPA, cDNA plasmids encoding the antigens are spotted in individual nanowells that are then sealed, minimizing protein diffusion to neighboring spots following expression and enabling a higher throughput of more than 10,000 proteins per array (39) ( Figure 1 ).

SuperBlock (Thermo Fisher Scientific, Rockford, IL) was used to block the arrays before protein expression to reduce non-specific binding. Arrays were then rinsed with DI water and centrifuged before they were filled with human In Vitro Transcription and Translation (IVTT) coupled system (Thermo Fisher Scientific). The wells were sealed with a polystyrene membrane under a pressure of 200 PSI. Arrays were incubated in a custom reactor device at 30˚C for 2 hours for protein expression and at 15˚C for 30 minutes for protein capture by the anti-GST Ab. Wells were then blocked for 30 min with 5% skimmed milk in phosphate buffered saline with 0.2% tween-20 (PBS-T). For detection of protein expression levels, mouse anti-GST monoclonal Ab (Cell signaling technology, Danvers, MA) was added and Alexa Fluor 647-labeled goat anti-mouse IgG (H+L) secondary Ab was used for detection (Thermo Fisher Scientific).

HPV nanowell arrays were expressed on the day of the assay to be probed with serum samples. A custom 16-well gasket (GraceBio-Labs, Bend, OR) was used on each array to separate sub-arrays to allow the addition of a different serum sample to each one. Serum was diluted 1:100 in 5% skim milk in PBST and each individual serum sample was added to a sub-array. Arrays were incubated overnight at 4˚C with gentle shaking and then rinsed with 5% milk in PBS-T. Bound Abs were detected using Alexa-Fluor 647-conjugated goat anti-human IgG Ab (H-L). Arrays were rinsed to remove unbound secondary Ab and dried by centrifugation before scanning.

We determined if serum Abs were reactive to more than one homologous Ag from different HPV types (e.g. Abs in serum from one patient recognize the L1 Ag from multiple HPV types). This allows the detection of potential cross-reactivity of Abs to Ags with high sequence similarity. However, Ab reactivity to homologous Ags from multiple HPV types could be due to past exposure to a different HPV type or the presence of this different HPV type in a different site of the body such as the oropharynx. Seropositivity was determined using the same cutoff value used throughout the study (see statistical analysis below).

Arrays were scanned at 635 nm with a Tecan PowerScanner (Tecan Group, Männedorf, Switzerland). Images were analyzed using the ArrayPro Analyzer software (MediaCybernetics, Bethesda, MD) for the quantification of signal intensity of individual spots. Normalization of raw intensity values was performed by dividing each spot signal intensity value by the median intensity of all spots to calculate the signal/background ratio.

The correlation of raw signal intensities of protein expression between the sub-arrays on an array randomly selected for quality control was determined with scatter plots and the Pearson correlation coefficient (R) was calculated to assess consistency. Protein expression on the arrays was measured by calculating the mean values of raw signal intensities of duplicate spots printed on the array. Mean values (of duplicate spots for a given Ag) of normalized signal intensity were calculated and a normalized signal cutoff of 1.5 was used to define seropositivity for any given Ag. Seropositivity rates were calculated for specific HPV proteins or specific types and their 95% confidence intervals were calculated. Pairwise comparisons of age differences among the four groups of women were conducted using the Dunn’s test (R version 4.5.0). A false discovery rate (FDR) adjusted p-value less than 0.05 was considered statistically significant. A total of 136 samples were run on the arrays.

The age distributions were not normally distributed. To account for skewed data, we used the pairwise comparison of the Wilcoxon-rank sum test, resulting in no significant age difference between the 4 groups of women (FDR adjusted p-value=0.05).

The quality and reproducibility of the HPV HD microarray printing was evaluated by DNA staining with picogreen and by measuring protein levels expressed and displayed using anti-GST monoclonal Ab ( Figure 2 ). Sixteen identical sub-arrays were printed on each array. The correlation coefficients of anti-GST signals were determined for each two sub-arrays. They were found to be in the range of 0.89-0.99, with 85.4% of the calculated R values (n=120) ≥0.95 and 89.2% ≥0.93 ( Figure 2 ), reflecting high correlation of protein display between the sub-arrays.

In this study, we aimed to determine the frequency and specificity of HPV-specific Ab responses in women with or without an abnormal Pap result. Age, race, HPV DNA status, and number of HPV types detected in patients contributing samples to the study are shown in Table 1 . The participating health clinics had a high proportion of white patients. The racial distribution of samples was random. There was no significant age difference between the four groups of women (minimum Benjamini-Hochberg FDR-adjusted p-value=0.06 among the pairwise comparisons). Of 82 women with normal cervical cytology, 41 women had no detectable HPV and 41 women had at least one of the 4 HR HPV types, HPV16, 18, 33, or 45. These three are the most common types implicated in cervical cancer (53, 54).

Among women who had been referred for colposcopy and biopsy (n=13), four had atypical squamous cells of undetermined significance (ASCUS), one had atypical glandular cells (AGC), three had low-grade squamous intraepithelial lesion (LGSIL), four had high-grade squamous intraepithelial lesion (HGSIL), and one had squamous cell carcinoma ( Table 2 ). In this group, the number of women with CIN grades I, II, and III were 4, 5, and 4, respectively ( Table 2 ).

The introduction of HPV vaccination occurred in 2006 and was approved for use in females aged 9-26. A significant proportion of women in this study (42.6%) were above the age of 30 ( Table 1 ). HPV vaccination status was known for only less than 20% of women participating in the CARE study (n=1131), preventing us from including this variable in our analyses. Nonetheless, among women with available vaccination data, less than 6% had received any doses of the HPV vaccine, indicating that the impact of vaccination on our findings was probably insignificant (33).

HPV DNA was detected in samples from cervical cytology and typed using the LA HPV Genotyping Assay which detects 37 HPV types. These include 23 high-risk (HR; 16, 18, 26, 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 64, 66, 67, 68, 69, 70, 73, 82, and IS39) and 14 low-risk (LR; 6, 11, 40, 42, 54, 55, 61, 62, 71, 72, 81, 83, 84, and 89) HPV types. Of these, 8 HPV DNA types (11, 26, 55, 64, 69, 71, 72, and IS39) were not detected in any of the samples ( Table 3 ). HPV40 and 82 DNA were each detected in only one participant. HPV6, 33, and 67 DNA were each detected in only two participants.

For all study participants, there was no co-infection with more than one of the 4 HR types – HPV16, 18, 33, and 45. The majority (63.4%) of women with normal cytology and one of the 4 HR types had infections with multiple HPV types. For women with abnormal Pap (n=54), the majority (66.7%) did not have detectable infections with multiple HPV types. The majority (63.4%) of women with normal Pap and any of 4 HR types were positive for HPV16 DNA. The second most frequently detected DNA type in these women was HPV18 (in 22% of women; Table 3 ). HPV16 DNA was also the most frequent type detected in all women with abnormal Pap (16.7%) and in almost half (46.2%) of women referred for colposcopy. The proportion of women who had any HR HPV DNA type was 53.7% among women with abnormal Pap and 51.5% among all study participants ( Table 3 ). Among women with normal Pap and any of 4 HR types, 31.7% had a co-infection with at least one LR type. The proportion of women who had any LR HPV DNA type was 29.6% among women with abnormal Pap and 21.3% among all study participants.

The frequency of HPV-specific IgG Abs in serum samples from women in the four groups under study is summarized in Table 4 . Abs to any HPV protein were detected in 47.6% (95% C.I.: 36.5-58.8%) and 40.7% (95% C.I.: 27.9-54.9%) of women with normal and abnormal cytology, respectively and in 44.9% (95% C.I.: 36.4-53.6%) of all study participants. This includes 41.5% of women who had no detectable HPV DNA in cervical cytology samples. Among women with normal Pap and any of the 4 HR HPV types 16, 18, 33, or 45, who were all HPV positive, 53.7% had HPV-specific Abs. Abs against any E protein were detected in 47.6% and 40.7% of women with normal and abnormal Pap, respectively. Abs against any L protein were detected in 39.0% and 33.3% of women with normal and abnormal Pap, respectively.

The most frequently detected Abs were against L1, detected in 30.1% (95% C.I.: 22.7-38.7%) of all study participants (in 31.7%, 95% C.I.: 22.1-43.0% and 27.8%, 95% C.I.: 16.9-41.9% of women with normal and abnormal Pap, respectively). The most frequently detected Abs against an E protein were against E1, detected in 24.3%, 95% C.I.: 17.5-32.5%) of all study participants (in 26.8% and 20.4% of women with normal and abnormal Pap, respectively). The least frequently detected Abs were against the E4 protein, detected in 13.2%, 95% C.I.: 8.2-20.4%) of all study participants (in 14.6% and 11.1% of women with normal and abnormal Pap, respectively). This was followed by Abs against E6, E7, and E2 (detected in 15.4%, 16.9%, and 16.9% of all women in the study, respectively). Two of the four women with CIN III had HPV-specific Abs. One (with HGSIL) had Abs against HPV52 E1 and the other (with squamous cell carcinoma) had Abs only to HPV31 L2 and no Abs against any of the E proteins. We found no significant difference in the Abs against the positive control proteins ( Supplementary Figure 1 ) or the negative control proteins ( Supplementary Figure 2 ) between women in the four groups in this study.

Most women who had Abs against a specific HPV type did not have type-specific HPV DNA in cervical cytology. Of all study participants, there were 6 women who showed correlation between HPV DNA in cervical cytology and Ab seropositivity. Of these, 3 had HPV16 DNA, and each of the other three had one of the three HPV DNA types 39, 45, or 58 in cervical cytology ( Table 5 ). Abs were most frequently detected against HPV51 and 52 (each in 20.6% of study participants). Even though HPV16 DNA was the most frequently detected type in cervical cytology (in 36.8%, 95% C.I.: 27.4-47.4% of study participants; Table 3 ), Abs against HPV16 were remarkably the least frequently detected in serum (in 7.4%, 95% C.I.: 3.8-13.5% of study participants; Table 5 ). Most (26/35; Table 3 ) women with HPV16 DNA in cervical cytology were in the normal Pap group, and only 9.8% of them had Abs against HPV16 in serum ( Table 5 ). Abs against the two LR types included on the arrays (HPV6 and 11) were detected in sera of 19.9% and 17.6% of study participants, respectively.

We next sought to determine the frequency of Abs against more than one homologous Ag from different HPV types (e.g. Abs against the E1 protein from more than one HPV type in serum from the same woman) among all study participants (N=136). Homologous proteins from different HPV types have sufficient DNA and amino acid sequence similarity to allow potential binding of Abs against an HPV protein from a specific type to a homologous protein from another (Ab cross-reactivity, Figure 3 ). Abs against L1 were the most cross-reactive, followed by L2. Among women with Abs against L1 and L2, 60.9% and 48.1%, respectively showed reactivity to at least one other homologous protein ( Figure 4 ). One woman had Abs against 8 different L1 proteins (from 8 different HPV types) and another against 7 different L2 proteins. The least cross-reactive were Abs against E5, with 15.4% of women with E5 Abs showing reactivity against more than one homologous E5 protein. The percentages of women with Abs that reacted to at least two homologous HPV E proteins were: 21.2% (for E1), 17.4% (E2), 27.8% (E4), 15.4% (E5), 19.0% (E6), and 26.1% (E7) ( Figure 4 ).