A Machine Learning-Based Analytic Pipeline Applied to Clinical and Serum IgG Immunoproteome Data To Predict Chlamydia trachomatis Genital Tract Ascension and Incident Infection in Women.

We developed a reusable and open-source machine learning (ML) pipeline that can provide an analytical framework for rigorous biomarker discovery. We implemented the ML pipeline to determine the predictive potential of clinical and immunoproteome antibody data for outcomes associated with Chlamydia trachomatis (Ct) infection collected from 222 cis-gender females with high Ct exposure. We compared the predictive performance of 4 ML algorithms (naive Bayes, random forest, extreme gradient boosting with linear booster [xgbLinear], and k-nearest neighbors [KNN]), screened from 215 ML methods, in combination with two different feature selection strategies, Boruta and recursive feature elimination. Recursive feature elimination performed better than Boruta in this study. In prediction of Ct ascending infection, naive Bayes yielded a slightly higher median value of are under the receiver operating characteristic curve (AUROC) 0.57 (95% confidence interval [CI], 0.54 to 0.59) than other methods and provided biological interpretability. For prediction of incident infection among women uninfected at enrollment, KNN performed slightly better than other algorithms, with a median AUROC of 0.61 (95% CI, 0.49 to 0.70). In contrast, xgbLinear and random forest had higher predictive performances, with median AUROC of 0.63 (95% CI, 0.58 to 0.67) and 0.62 (95% CI, 0.58 to 0.64), respectively, for women infected at enrollment. Our findings suggest that clinical factors and serum anti-Ct protein IgGs are inadequate biomarkers for ascension or incident Ct infection. Nevertheless, our analysis highlights the utility of a pipeline that searches for biomarkers and evaluates prediction performance and interpretability. IMPORTANCE Biomarker discovery to aid early diagnosis and treatment using machine learning (ML) approaches is a rapidly developing area in host-microbe studies. However, lack of reproducibility and interpretability of ML-driven biomarker analysis hinders selection of robust biomarkers that can be applied in clinical practice. We thus developed a rigorous ML analytical framework and provide recommendations for enhancing reproducibility of biomarkers. We emphasize the importance of robustness in selection of ML methods, evaluation of performance, and interpretability of biomarkers. Our ML pipeline is reusable and open-source and can be used not only to identify host-pathogen interaction biomarkers but also in microbiome studies and ecological and environmental microbiology research.

Discovery and systematic assessment of early biomarkers that predict progression to severe COVID-19 disease.

BACKGROUND: The clinical course of COVID-19 patients ranges from asymptomatic infection, via mild and moderate illness, to severe disease and even fatal outcome. Biomarkers which enable an early prediction of the severity of COVID-19 progression, would be enormously beneficial to guide patient care and early intervention prior to hospitalization. METHODS: Here we describe the identification of plasma protein biomarkers using an antibody microarray-based approach in order to predict a severe cause of a COVID-19 disease already in an early phase of SARS-CoV-2 infection. To this end, plasma samples from two independent cohorts were analyzed by antibody microarrays targeting up to 998 different proteins. RESULTS: In total, we identified 11 promising protein biomarker candidates to predict disease severity during an early phase of COVID-19 infection coherently in both analyzed cohorts. A set of four (S100A8/A9, TSP1, FINC, IFNL1), and two sets of three proteins (S100A8/A9, TSP1, ERBB2 and S100A8/A9, TSP1, IFNL1) were selected using machine learning as multimarker panels with sufficient accuracy for the implementation in a prognostic test. CONCLUSIONS: Using these biomarkers, patients at high risk of developing a severe or critical disease may be selected for treatment with specialized therapeutic options such as neutralizing antibodies or antivirals. Early therapy through early stratification may not only have a positive impact on the outcome of individual COVID-19 patients but could additionally prevent hospitals from being overwhelmed in potential future pandemic situations.

We aimed to identify components of the blood present during the early phase of SARS-CoV-2 infection that distinguish people who are likely to develop severe symptoms of COVID-19. Blood from people who later developed a mild or moderate course of disease were compared to blood from people who later had a severe or critical course of disease. Here, we identified a combination of three proteins that were present in the blood of patients with COVID-19 who later developed a severe or critical disease. Identifying the presence of these proteins in patients at an early stage of infection could enable physicians to treat these patients early on to avoid progression of the disease.

Reduced Endometrial Ascension and Enhanced Reinfection Associated With Immunoglobulin G Antibodies to Specific Chlamydia trachomatis Proteins in Women at Risk for Chlamydia.

BACKGROUND: Previous research revealed antibodies targeting Chlamydia trachomatis elementary bodies was not associated with reduced endometrial or incident infection in C. trachomatis-exposed women. However, data on the role of C. trachomatis protein-specific antibodies in protection are limited. METHODS: A whole-proteome C. trachomatis array screening serum pools from C. trachomatis-exposed women identified 121 immunoprevalent proteins. Individual serum samples were probed using a focused array. Immunoglobulin (Ig) G antibody frequencies and endometrial or incident infection relationships were examined using Wilcoxon rank sum test. The impact of the breadth and magnitude of protein-specific IgGs on ascension and incident infection were examined using multivariable stepwise logistic regression. Complementary RNA sequencing quantified C. trachomatis gene transcripts in cervical swab samples from infected women. RESULTS: IgG to pGP3 and CT_005 were associated with reduced endometrial infection; anti-CT_443, anti-CT_486, and anti-CT_123 were associated with increased incident infection. Increased breadth of protein recognition did not however predict protection from endometrial or incident infection. Messenger RNAs for immunoprevalent C. trachomatis proteins were highly abundant in the cervix. CONCLUSIONS: Protein-specific C. trachomatis antibodies are not sufficient to protect against ascending or incident infection. However, cervical C. trachomatis gene transcript abundance positively correlates with C. trachomatis protein immunogenicity. These abundant and broadly recognized antigens are viable vaccine candidates.

Development of Helicobacter pylori Whole-Proteome Arrays and Identification of Serologic Biomarkers for Noncardia Gastric Cancer in the MCC-Spain Study.

BACKGROUND: Helicobacter pylori (H. pylori) is a bacterial carcinogen and the leading risk factor for noncardia gastric cancer (NCGC). Detecting antibodies against specific H. pylori proteins in peripheral blood can be applied to characterize infection and determine disease associations. Most studies analyzing the association between H. pylori infection and gastric cancer have focused on previously identified antigens, predominantly the virulence factor cytotoxin-associated gene A (CagA). Selecting antigens in an unbiased approach may, however, allow the identification of novel biomarkers. METHODS: Using a combination of multiple spotting technique and cell-free, on-chip protein expression, we displayed the H. pylori genome (strain 26695) on high-density microarrays. Immunogenic proteins were identified by serum pool incubations and henceforth analyzed in individual samples. To test its applicability, we used sera from a multicase-control (MCC)-Spain study. Serologic responses between NCGC cases and controls were assessed by conditional logistic regression estimating ORs and 95% confidence intervals. RESULTS: We successfully expressed 93% of the 1,440 H. pylori open reading frames in situ. Of these, 231 (17%) were found to be immunogenic. By comparing 58 NCGC cases with 58 matched controls, we confirmed a higher seroprevalence of CagA among cases (66%) than controls (31%). We further identified a potential novel marker, the Helicobacter outer membrane protein A (HopA). CONCLUSIONS: In this study, we provide evidence that our H. pylori whole-proteome microarray offers a platform for unbiased de novo identification of serologic biomarkers. IMPACT: Given its versatile workflow, antibody responses against other H. pylori strains and possible associations with diverse H. pylori-related outcomes can be systematically analyzed.

Novel Autoantibody Signatures in Sera of Patients with Pancreatic Cancer, Chronic Pancreatitis and Autoimmune Pancreatitis: A Protein Microarray Profiling Approach.

Identification of disease-associated autoantibodies is of high importance. Their assessment could complement current diagnostic modalities and assist the clinical management of patients. We aimed at developing and validating high-throughput protein microarrays able to screen patients' sera to determine disease-specific autoantibody-signatures for pancreatic cancer (PDAC), chronic pancreatitis (CP), autoimmune pancreatitis and their subtypes (AIP-1 and AIP-2). In-house manufactured microarrays were used for autoantibody-profiling of IgG-enriched preoperative sera from PDAC-, CP-, AIP-1-, AIP-2-, other gastrointestinal disease (GID) patients and healthy controls. As a top-down strategy, three different fluorescence detection-based protein-microarrays were used: large with 6400, intermediate with 345, and small with 36 full-length human recombinant proteins. Large-scale analysis revealed 89 PDAC, 98 CP and 104 AIP immunogenic antigens. Narrowing the selection to 29 autoantigens using pooled sera first and individual sera afterwards allowed a discrimination of CP and AIP from PDAC. For validation, predictive models based on the identified antigens were generated which enabled discrimination between PDAC and AIP-1 or AIP-2 yielded high AUC values of 0.940 and 0.925, respectively. A new repertoire of autoantigens was identified and their assembly as a multiplex test will provide a fast and cost-effective tool for differential diagnosis of pancreatic diseases with high clinical relevance.

Antibodies against human papillomavirus proteins in Barrett's dysplasia and intramucosal esophageal adenocarcinoma.

High-risk human papillomavirus (HPV) types 16/18 have been associated with Barrett's dysplasia (BD)/esophageal adenocarcinoma (EAC). Nevertheless, no data exist in relation to serological analysis for HPV antibodies in BD/EAC with site-specific viral DNA status. We prospectively examined antibodies to multiple HPV types in 438 patients representing hospital/reflux controls and Barrett's metaplasia (BM)/BD/intramucosal EAC. Antibody responses to HPV6/11/16/18/31/33/45/52/58 were analyzed using multiplex serology, including antibodies to E6/E7/E1/E2 and L1 antigens. Seropositivity for individual HPV proteins was infrequent in both cases and controls and was ≤10.2%. There was no difference in the seroprevalence of antibodies to any HPV antigen/antibody combination between reclassified cases (BD/EAC) and controls (hospital/reflux/BM) or between HPV16 or HPV18 DNA cases and controls, respectively. Among HPV16 DNA-positive BD/EAC cases, antibodies to HPV16 E7 were significantly more prevalent (3/26, 11.5%) than in hospital and reflux controls plus BM (5/328, 1.5%) (adjusted OR = 10.12, 95% CI: 1.61-63.73, P = 0.014). Among HPV18 DNA-positive cases, antibodies to HPV18 E1 were present in 3/6 (50%) cases versus 5/328 (1.5%) controls (adjusted OR = 44.28, 95% CI: 6.10-321.47, P = 0.0002). Although antibodies against HPV were generally uncommon in cases and controls, immune responses against two early proteins of HPV16/18 were significantly more frequent in viral DNA-positive BD/intramucosal EAC.

The transcription factor FLI1 promotes cancer progression by affecting cell cycle regulation.

Binding of transcription factors to mutated DNA sequences is a likely regulator of cancer progression. Noncoding regulatory mutations such as those on the core promoter of the gene encoding human telomerase reverse transcriptase have been shown to affect gene expression in cancer. Using a protein microarray of 667 transcription factor DNA-binding domains and subsequent functional assays, we looked for transcription factors that preferentially bind the mutant hTERT promoter and characterized their downstream effects. One of them, friend leukemia integration 1 (FLI1), which belongs to the E26 transforming-specific family of transcription factors, exhibited particularly strong effects with respect to regulating hTERT expression, while the even better binding ELK3 did not. Depletion of FLI1 decreased expression of the genes for cyclin D1 (CCND1) and E2F transcription factor 2 (E2F2) resulting in a G1/S cell cycle arrest and in consequence a reduction of cell proliferation. FLI1 also affected CMTM7, another gene involved in G1/S transition, although by another process that suggests a balanced regulation of the tumor suppressor gene's activity via opposing regulation processes. FLI1 expression was found upregulated and correlated with an increase in CCND1 expression in pancreatic cancer and brain tumors. In non-neoplastic lung cells, however, FLI1 depletion led to rapid progression through the cell cycle. This coincides with the fact that FLI1 is downregulated in lung tumors. Taken together, our data indicate a cell cycle regulatory hub involving FLI1, hTERT, CCND1 and E2F2 in a tissue- and context-dependent manner.

Sexually transmitted infections and risk of epithelial ovarian cancer: results from the Nurses' Health Studies.

BACKGROUND: Sexually transmitted infections (STIs) are associated with pelvic inflammatory disease and tubal pathologies. Given the tubal origin of a proportion of ovarian cancers, STIs may be relevant in their aetiology. METHODS: Antibodies indicating past infection with Chlamydia trachomatis, Mycoplasma genitalium, herpes simplex virus type 2, and against human papillomavirus oncogenes (L1 and E6+E7 oncoproteins of types 16, 18, 45) were measured in prediagnosis plasma samples in a nested case-control study in the Nurses' Health Studies (n = 337 cases 1:1 matched to controls). Logistic regression was used to estimate multivariable-adjusted relative risks (RRs) and 95% confidence intervals [CIs] comparing women seropositive vs. seronegative among all cases (invasive and borderline), invasive (n = 257), and invasive serous ovarian cancers; n = 170), and borderline ovarian tumours (n = 80). RESULTS: C. trachomatis seropositivity was associated with higher risk of ovarian cancer overall (RR = 2.07 [1.25-3.43]); results were similar for invasive, invasive serous, and borderline tumours. We observed no associations for the other STIs. Relative to women seronegative to all infections, strongest associations were observed for seropositivity to C. trachomatis plus another STI (2.74 [1.20-6.27]; C. trachomatis alone, 1.88 [1.03-3.42]; all cases); however, the RRs were not significantly different. CONCLUSIONS: C. trachomatis infection may increase ovarian cancer risk; additional studies are required.

Chlamydia trachomatis Whole-Proteome Microarray Analysis of The Netherlands Chlamydia Cohort Study.

Chlamydia trachomatis (Ct) whole-proteome microarrays were utilized to identify antibody patterns associated with infection; pelvic inflammatory disease (PID), tubal factor infertility, chronic pelvic pain (CPP) and ectopic pregnancy in a subsample of the Netherlands Chlamydia cohort study. Serum pools were analyzed on whole-proteome arrays. The 121 most reactive antigens identified during whole-proteome arrays were selected for further analysis with minimized microarrays that allowed for single sera analysis. From the 232 single sera; 145 (62.5%) serum samples were reactive for at least one antigen. To discriminate between positive and negative serum samples; we created a panel of in total 18 antigens which identified 96% of all microarray positive samples. Antigens CT_858; CT_813 and CT_142 were most reactive. Comparison of antibody reactivity's among women with and without Ct related sequelae revealed that the reactivity of CT_813 and CT_142 was less common among women with PID compared to women without (29.0% versus 58.6%, p = 0.005 and 25.8% versus 50.6%, p = 0.017 respectively). CT_858 was less common among CPP cases compared to controls (33.3% versus 58.6; p = 0.028). Using a whole-proteome array to select antigens for minimized arrays allows for the identification of novel informative antigens as general infection markers or disease associated antigens.

In situ, Cell-free Protein Expression on Microarrays and Their Use for the Detection of Immune Responses.

Until recently, whole-proteome microarrays for comprehensive studies of protein interactions were mostly produced by individual cloning and cellular expression of very many open reading frames, followed by protein isolation and purification as well as array production. To overcome this cumbersome process, we have developed a method to generate microarrays representing entire proteomes by a combination of multiple spotting and on-chip, cell-free protein expression. Here, we describe the protocol for the production of bacterial protein microarrays. With slight adaptations, however, the procedure can be applied to the proteome of any organism. Expression constructs of each gene are generated by PCR on bacterial genomic DNA followed by a common secondary amplification that is adding relevant regulative elements to either end of the constructs. The unpurified PCR-products are spotted onto the microarray surface. Full-length proteins are directly expressed in situ in a cell-free manner and stay attached to the surface without further action. As an example of a typical application, we describe here the proteome-wide analysis of the immune response to a bacterial infectious agent by characterizing the binding profiles of the antibodies in patient sera.

Antibodies Against Chlamydia trachomatis and Ovarian Cancer Risk in Two Independent Populations.

BACKGROUND: Pelvic inflammatory disease (PID) has been associated with ovarian cancer risk. To clarify the role of Chlamydia trachomatis and other infectious agents in the development of ovarian cancer, we evaluated the association of serologic markers with incident ovarian cancer using a staged approach in two independent populations. METHODS: Studies included: 1) a case-control study in Poland (244 ovarian cancers/556 control subjects) and 2) a prospective nested case-control study in the PLCO Cancer Screening Trial (160 ovarian cancers/159 control subjects). Associations of serologic marker levels with ovarian cancer risk at diagnostic as well as higher thresholds, identified in Poland and independently evaluated in PLCO, were estimated using multivariable adjusted logistic regression. RESULTS: In the Polish study, antibodies (based on laboratory cut-point) against the chlamydia plasmid-encoded Pgp3 protein (serological gold standard) were associated with increased ovarian cancer risk (adjusted odds ratio [OR] = 1.63, 95% confidence interval [CI] = 1.20 to 2.22); when a positive result was redefined at higher levels, ovarian cancer risk was increased (cut-point 2: OR = 2.00, 95% CI = 1.38 to 2.89; cut-point 3 [max OR]: OR = 2.19, 95% CI = 1.29 to 3.73). In the prospective PLCO study, Pgp3 antibodies were associated with elevated risk at the laboratory cut-point (OR = 1.43, 95% CI = 0.78 to 2.63) and more stringent cut-points (cut-point 2: OR = 2.25, 95% CI = 1.07 to 4.71); cut-point 3: OR = 2.53, 95% CI = 0.63 to 10.08). In both studies, antibodies against other infectious agents measured were not associated with risk. CONCLUSIONS: In two independent populations, antibodies against prior/current C. trachomatis (Pgp3) were associated with a doubling in ovarian cancer risk, whereas markers of other infectious agents were unrelated. These findings lend support for an association between PID and ovarian cancer.

Differences in Chlamydia trachomatis seroprevalence between ethnic groups cannot be fully explained by socioeconomic status, sexual healthcare seeking behavior or sexual risk behavior: a cross-sectional analysis in the HEalthy LIfe in an Urban Setting (HELIUS) study.

BACKGROUND: In the Netherlands, there are strong disparities in Chlamydia trachomatis (CT) prevalence between ethnic groups. The current study aims to identify whether socioeconomic status, sexual risk behavior and sexual healthcare seeking behavior may explain differences in CT seroprevalence between ethnic groups. METHODS: We used 2011-2014 baseline data of the HELIUS (HEalthy LIfe in an Urban Setting) study, a multi-ethnic population-based cohort study in Amsterdam, the Netherlands, including participants from Dutch, African Surinamese, South-Asian Surinamese, Ghanaian, Moroccan and Turkish origin. For this analysis, we selected sexually active, heterosexual participants aged 18-34 years old. CT seroprevalence was determined using a multiplex serology assay. The CT seroprevalence ratios between different ethnicities are calculated and adjusted for potential indicators of socioeconomic status, sexual risk behavior and sexual healthcare seeking behavior. RESULTS: The study population consisted of 2001 individuals (52.8% female) with a median age of 28 years (IQR 24-31). CT seropositivity differed by ethnicities and ranged from 71.6% (African Surinamese), and 67.9% (Ghanaian) to 31.1% (Turkish). The CT seroprevalence ratio of African Surinamese was 1.72 (95% CI 1.43-2.06) and 1.52 (95% CI 1.16-1.99) of Ghanaian as compared to the Dutch reference group, after adjustment for socioeconomic status, sexual risk behavior and sexual healthcare seeking behavior. CONCLUSIONS: Indicators of socioeconomic status, sexual risk behavior, and sexual health seeking behavior could not explain the higher CT seroprevalence among African Surinamese and Ghanaian residents of Amsterdam.

Advancing the public health applications of Chlamydia trachomatis serology.

Genital Chlamydia trachomatis infection is the most commonly diagnosed sexually transmitted infection. Trachoma is caused by ocular infection with C trachomatis and is the leading infectious cause of blindness worldwide. New serological assays for C trachomatis could facilitate improved understanding of C trachomatis epidemiology and prevention. C trachomatis serology offers a means of investigating the incidence of chlamydia infection and might be developed as a biomarker of scarring sequelae, such as pelvic inflammatory disease. Therefore, serological assays have potential as epidemiological tools to quantify unmet need, inform service planning, evaluate interventions including screening and treatment, and to assess new vaccine candidates. However, questions about the performance characteristics and interpretation of C trachomatis serological assays remain, which must be addressed to advance development within this field. In this Personal View, we explore the available information about C trachomatis serology and propose several priority actions. These actions involve development of target product profiles to guide assay selection and assessment across multiple applications and populations, establishment of a serum bank to facilitate assay development and evaluation, and development of technical and statistical methods for assay evaluation and analysis of serological findings. The field of C trachomatis serology will benefit from collaboration across the public health community to align technological developments with their potential applications.

Immunoprofiling of Chlamydia trachomatis using whole-proteome microarrays generated by on-chip in situ expression.

Using Chlamydia trachomatis (Ct) as a complex model organism, we describe a method to generate bacterial whole-proteome microarrays using cell-free, on-chip protein expression. Expression constructs were generated by two successive PCRs directly from bacterial genomic DNA. Bacterial proteins expressed on microarrays display antigenic epitopes, thereby providing an efficient method for immunoprofiling of patients and allowing de novo identification of disease-related serum antibodies. Through comparison of antibody reactivity patterns, we newly identified antigens recognized by known Ct-seropositive samples, and antigens reacting only with samples from cervical cancer (CxCa) patients. Large-scale validation experiments using high-throughput suspension bead array serology confirmed their significance as markers for either general Ct infection or CxCa, supporting an association of Ct infection with CxCa. In conclusion, we introduce a method for generation of fast and efficient proteome immunoassays which can be easily adapted for other microorganisms in all areas of infection research.