key: cord-0308983-ubfs6x9u
authors: Hasan, Mohammad Rubayet; Rahman, Mahbuba; Khan, Taushif; Saeed, Amira; Sundaraju, Sathyavathi; Flores, Annaliza; Hawken, Philip; Rawat, Arun; Elkum, Naser; Hussain, Khalid; Tan, Rusung; Tang, Patrick; Marr, Nico
title: Virome-wide serological profiling reveals association of herpesviruses with obesity
date: 2020-09-03
journal: bioRxiv
DOI: 10.1101/2020.09.02.280503
sha: bfd3eee7972333d0894d199d852bd7ff9a4bf4f7
doc_id: 308983
cord_uid: ubfs6x9u

The relationship between viral infection and obesity has been known for several decades but epidemiological data related to obesity is limited to only a few viral pathogens. To identify associations between viral infections and obesity, a high-throughput virome-wide serological profiling tool, VirScan, was used to measure antibody responses to a wide range of viruses. Serum specimens from 457 Qatari adults (lean=184;obese=273) and 231 Qatari children (lean=111;obese=120) were assessed by VirScan. Pediatric specimens were simultaneously tested by conventional serology for several herpesviruses to validate VirScan results. Viral association with obesity was determined by calculation of odds ratio (OR) and p-values from Fisher test, and by multivariate regression analysis to adjust for age and gender, with Bonferroni correction for multiple testing. Comprehensive serological profiling of Qatari adult population with VirScan revealed positive and negative associations (p<0.05) of antibody responses to members of Herpesviridae and Picornaviridae families, respectively, with obesity. After adjusting p-values for multiple comparisons, only herpes simplex virus 1 (HSV-1) and Rhinovirus A were positively (OR=3.3; 95%CI 2.15-4.99; p=2.787E-08) and negatively (OR=0.4; 95%CI 0.26-0.65; p=1.175E-03) associated with obesity. At the peptide level, higher prevalence of antibodies against several peptide epitopes of HSV-1/2 was positively (OR=2.35-3.82; p≤3.981E-05) associated with obesity. No such associations were seen at the species or peptide levels in the pediatric population. By multivariate regression analysis, HSV-1 was independently associated with obesity irrespective of age and gender. These findings are in agreement with limited data on the adipogenic properties of HSV-1 observed in vitro. Importance The state of Qatar has one of the highest rates of obesity and associated morbidities in the world. Although obesity is predominantly caused by the intake of high calorie diet and reduced physical activities, other factors including infections with certain viruses have been reported. Among these viruses, human adenoviruses were widely studied but epidemiological data for other viruses in relation to human obesity are limited. Here, we studied the association of obesity in Qatari adults and children with a wide range of viral pathogens using VirScan, a virome-wide serological profiling tool. Our results indicate significant association HSV-1 with obesity in the adult population only. Furthermore, we have identified a set of HSV peptides as candidate obesogenic factors for future studies.

3 calorie diet and reduced physical activities, other factors including infections with certain 51 viruses have been reported. Among these viruses, human adenoviruses were widely studied 52 but epidemiological data for other viruses in relation to human obesity are limited. Here, we 53 studied the association of obesity in Qatari adults and children with a wide range of viral 54 pathogens using VirScan, a virome-wide serological profiling tool. Our results indicate 55 significant association HSV-1 with obesity in the adult population only. Furthermore, we 56 have identified a set of HSV peptides as candidate obesogenic factors for future studies. To determine the serological status of individuals as 'positive' or 'negative' for 174 different viruses based on VirScan data, virus-specific species score thresholds determined by 175 8 a generalized linear model (GLM) was applied as described in the materials and methods. 176

VirScan based serological data for CMV, EBV and HSV-1/2 were compared to that of 177 conventional serology. Because our conventional HSV-1/2 IgG test does not differentiate 178 between HSV-1 and -2, VirScan results for these viral species were combined. VirScan 179 results either positive for HSV-1 or -2 or for both were all considered as HSV positive. 180

Specimens from the pediatric cohort were simultaneously tested by VirScan and conventional 181 serology. The sensitivity, specificity and accuracy of VirScan results compared to 182 conventional methods are all 98% for CMV, 100%, 83% and 97% for EBV, and 89%, 91% 183 and 90% for HSV, respectively (Table 2) . p=0.0001) of rhinovirus A seropositivity. With the same adjusted p-value cut-off, no such 198 association is observed in the pediatric population. Apart from the viral species that belong to 199 the Herpesviridae and Picornaviridae families, the seroprevalence of rotavirus A is higher in 200 the obese groups and is nominally associated (p<0.05) with obesity in both cohorts. In 201 addition, obesity is nominally associated with higher seroprevalences of human coronavirus 202 HKU1, human adenovirus D, influenza C virus, human parainfluenza virus 1 and human 203 parvovirus B19 in the adult cohort and higher seroprevalences of influenza A and B viruses 204 in the pediatric cohort. Consistent with VirScan data, seroprevalences of CMV, EBV and 205 HSV in the pediatric cohort determined by conventional serology are not significantly 206 different in obese versus lean groups (Fig. S2) . 207

The seroprevalence of HSV-1 among Qatari general population is unknown but in a 208 previous study HSV-1 seroprevalence among Qatari male blood donors was estimated to be 209 82. 3% (19) . Based on our VirScan data, the overall prevalence of HSV-1 in Qatari adult 210 population is 74.1%, with 70.4% among the males and 76.1% among the females. The 211 seroprevalence of HSV-1 in Qatari adult obese and lean population is 81.3% and 57.1%, 212 respectively. Seroprevalence of HSV-1 is the pediatric population is 47%, which increased to 213 55% by the age of 30 years, and to 90% by the age >55 years. To assess whether our results 214 are affected by gender and age of the study participants, we used a multivariate regression 215 model and examined the coefficients of association with adjusted beta values. The 216 associations of age, gender and BMI status were studied with respect to adjusted virus scores. 217

Tests for a total of 190 associations were performed with adjusted viral scores for 38 viral 218 species (>10% prevalence) in 457 adult and 231 pediatric samples with five features (age, 219 male, female, lean and obese). An association was considered to be significant if absolute 220 coefficient of association (|beta|) was 0.678 and p-value was 0.00013 (-log10(pval) > 221 3.88). A total of 25 and 4 associations were determined to be significant for the adult and 222 pediatric cohort, respectively ( cohorts. Differentially enriched antigenic peptides were primarily derived from viral 253 structural proteins, including glycoprotein G (gG) and tegument proteins of HSV (Table S2) . 254

Currently there is compelling epidemiological data on the association between 256

Adenovirus-36 and obesity as well as animal data on several other viruses (7, 20-22), but a 257 broad screen for these and other infectious causes of obesity cannot be performed without a 258 well-established, high throughput platform for comprehensive serological profiling. The 259 establishment of the VirScan PhIP-Seq serological profiling platform (16) Unlike previous VirScan-based studies that applied empirically determined virus-285 specific species scores to determine seropositivity (16, 18), we established a generalized 286 linear model using control serum, tittered for antibodies against different viruses, and taking 287 into the accounts of number of peptides available in the library for each viral species. We 288 then tested the specimens from the pediatric cohort for CMV, EBV and HSV with standard 289 serological assays that are used for patient testing. Serological data on HSV, EBV and CMV 290 were then used to validate the VirScan-based serological data demonstrating 90% to 98% 291 accuracy compared to the standard methods. Based on VirScan based serology, after 292 adjustment of p-values for multiple testing, HSV-1 seropositivity is significantly associated 293 with obesity in the adult population only (Figs. 2 A and B) . Consistently, HSV-1 is associated 294 with obesity in the adult population independent of age and gender by multivariate regression 295 analysis (Fig. 3A) . Other herpesviruses such as HSV-2, CMV and EBV are also nominally 296 associated with obesity, and by multivariate analysis, they are associated with obesity in a 297 gender specific manner. 298

To obtain further insights on the relationship of herpes viruses with obesity, we 299 analyzed the association of virus specific peptides with obesity. Consistent with our findings 300 13 at the species level, it is mostly the HSV-1 associated peptides that show the strongest 301 associations with obesity ( Fig. 4 A and B) . Peptides from other herpes viruses are also 302 nominally associated with obesity in the Qatari adult population. Seroprevalence of 303 herpesviruses are not significantly different between obese and lean population in the 304 pediatric cohort but differentially enriched, HSV-1/2 and EBV peptides are nominally 305 associated with obesity (data not shown). We also found some HSV-1/2 peptides that are 306 independently associated with obesity in both populations (Table S2 ). These results suggest 307 that acquiring HSV-1 infection as people gets older may increase the odds of becoming 308

obese. An early sign of this is observed by the nominal association of HSV peptides with 309 obesity in the pediatric population. Positive association of these HSV-1/2 peptides with 310 obesity observed in our population also correlates with earlier evidence of association 311 between HSV and excessive adiposity in different age and gender groups. However, in our adult (>18 years) population, seropositivity to CMV was associated with 318 obesity in the male population only. 319

In our study, obesity was not associated with human adenovirus or any of its 320 serotypes (Adv-5, Adv-9, Adv-31, Adv-36 and Adv-37) that were previously reported to be 321 associated with obesity, epidemiologically, or linked to enhanced lipogenesis in vitro, 322 suggesting that adenoviruses may have a limited role in the incidence of high rates of obesity 323 among Qataris. In a recent study, Lessan et al. demonstrated that the seroprevalence of Adv-324 36 in United Arab Emirates (UAE) is much higher in their population although there is no 325 14 significant difference in prevalence of this virus in their obese and lean populations (24). Our 326 adenovirus data provides additional support to this finding in a highly similar population. 327 328 An incidental but interesting finding in our study is the negative association of the 329 family of picornaviruses with obesity. This phenomenon was only observed in the adult 330 population and may potentially be related to the waning immunity to these viruses with age. 331

Obesity associated impaired immunity to various infections such as influenza, pneumonia, 332

Helicobacter pyrolii and nosocomial infections or poor response to vaccines such as hepatitis 333 B, tetanus and influenza vaccines have been described in the literature (21). However, there 334 are no reports on impaired immunity or higher susceptibility to infections with picornaviruses 335 such as enteroviruses and rhinoviruses in relation to obesity. Interestingly, no such 336 association is seen in the pediatric population. This may be related to higher rates of infection 337 with these viruses in children, and their immune systems being continuously challenged by 338 these viruses. These findings are novel and warrant further investigations in a separate study. 339

Using a comprehensive virome-wide analysis, we identified viral species previously 340 associated with obesity, but were unable to find novel viral associations with obesity. 341 However, the association of different herpesviruses or their specific epitopes with obesity 342 confirms the association of these viruses described for other populations. Furthermore, 343 despite very high rates of obesity, no viral association studies have ever been described for 344 the Qatari population. One limitation of our study is that VirScan results are based on 345 antibody binding of linear epitopes only. Therefore, potential antibody interactions that relies 346 on tertiary structure of epitopes may have been missed. Also, our clinical validation of 347

VirScan serology data was limited to CMV, EBV and HSV only. Our study failed to provide 348 any mechanistic insights on the role of HSV-1 infection in adipogenesis. But, for the first 349 time, we have described high resolution, peptide-epitope level data for HSV-1 in association 350 15 with obesity. Our analysis of differentially enriched peptides from a large collection of 351 peptides covering the entire human virome reveals a number of viral epitopes that could be 352 utilized for further mechanistic studies. We have listed candidate HSV-1 and -2 peptides that 353

were strongly correlated to obesity in the adult population and are highly prevalent in both 354 adult and pediatric obese populations (Table S 1 and 2) . Interestingly, the majority of these 355 peptides belongs to the HSV glycoprotein family. Both HSV and CMV are known as 356 lipogenic viruses and are known to affect cellular metabolism in different ways or are known 357 to cause expansion of adipose tissues by their effects on inflammatory pathways (7, 10). 358

Further studies on the role of HSV-1 candidate peptides identified in this study in cell culture 359 or animal models may reveal novel pathways for virus induced adipogenesis. 360

In conclusion, we have conducted a virome-wide seroprevalence study to detect 361 associations between past viral exposures and obesity in a population that is highly endemic 362 for obesity. Our analysis revealed a strong positive association of obesity with HSV-1 and 363 nominal associations with other herpesviruses. Our finding may have implications for 364 understanding the underlying causes of higher prevalence of obesity not only among the 365 Qataris but also among the citizens of other countries in the Arabian Peninsula. 366 367

For the adult cohort, 800 randomly selected serum specimens were obtained from Qatar 370 Biobank (QBB). This cohort represents Qatari nationals and long-term residents (lived in 371

Qatar for >15 years) aged ≥18 years. Data on age, gender, ethnicity and BMI were collected. 372

Specimens from non-Qatari participants and specimens for which no BMI data were 373 available were excluded from analysis. Also, specimens from subjects that fall into the 374 overweight category (BMI >25 to <30) were excluded from the analysis. 

Obesity and overweight : Fact sheet

Global, regional, and national prevalence of 468 overweight and obesity in children and adults during 1980-2013: a systematic analysis 469 for the Global Burden of Disease Study

Health Effects of Overweight and Obesity in 478 195 Countries over 25 Years

Qatar Biobank: background and methods

Viral Infections and Obesity

Infectobesity: obesity of infectious origin

Role of adenoviruses in obesity

Viral activation of cellular metabolism

Cytomegalovirus infection causes an increase of arterial blood pressure

Cytomegalovirus Infection in the General Population

Characterising the aetiology 500 of severe acute gastroenteritis among patients visiting a hospital in Qatar using real-501 time polymerase chain reaction

Viral 504 etiology, seasonality and severity of hospitalized patients with severe acute 505 respiratory infections in the Eastern Mediterranean Region

Adenovirus 36 and Obesity: An Overview

Viral 511 immunology. Comprehensive serological profiling of human populations using a 512 synthetic human virome

Measles virus infection diminishes preexisting antibodies that offer protection from 516 other pathogens

Temporal virus serological profiling of kidney graft recipients using 519

Estimating 521 seroprevalence of herpes simplex virus type 1 among different Middle East and North 522

African male populations residing in Qatar

Microbes and obesity--interrelationship between 524 infection, adipose tissue and the immune system

Obesity and infection: reciprocal causality

Adenovirus 36 Infection Increased the Risk of Obesity: A Meta-Analysis Update

Associations between excessive 531 adiposity and seroprevalence of herpes simplex virus type 1 and type 2 among US 532 adults: a population-based age-period-cohort analysis

Atkinson 534 RL. 2020. Obesity and Diabetes in an Arab population: Role of Adenovirus 36

Inherited human IFN-gamma 541 deficiency underlies mycobacterial disease

Sensitivity, %(95% CI) 98%(94-100%) 100%(98-100%) 89%

Specificity, %(95% CI) 98%(92-100%) 83%(68-93%) 91%

Accuracy, %(95% CI) 98%(95-99%) 97%(93-99%) 90%

cohort, 231 serum samples were collected from Qatari obese and lean children admitted to 376 Sidra Medicine, which is a 400-bed tertiary care children's and women's hospital in Qatar, 377 during the period October 2018 to November 2019. Residual specimens from comprehensive 378 metabolic panel (CMP) and basic metabolic panel (BMP) tests were collected. Specimens 379 from subjects that satisfies inclusion criteria were identified through laboratory information 380 management systems (LIMS) query on a weekly basis using Discern Analytics 2.0 (Cerner). 381The inclusion criteria were: i) age 7-15 years ii) Qatari nationality and, iii) BMI centile 5% to 382 85% as lean individuals or 95% as obese. VirScan PhIP-Seq analysis of serum specimens from the adult and pediatric cohort 392 was carried out by the methods described previously except that an expanded bacteriophage 393 library (containing 2x10 10 plaque-forming units) displaying 115,753 peptides was used, and 394 sequencing was performed using an Illumina NextSeq500 platform and NextSeq 500/550 395 High Output Kit v2.5 (75 Cycles) kits (Illumina). Each specimen was tested in two technical 396replicates. Each sequencing batch consisted of specimens representing obese and lean 397 subjects, mock-IP controls, positive control and input library in a random manner with an aim 398 to minimize batch effects on the output data. PhIP-Seq analysis of a total of 688 specimens 399 was completed in 12 NextSeq500 sequencing runs. VirScan data were filtered for enriched peptides as described previously (16). Briefly, we 414 first imputed p-values (-log10 transformed) by fitting a zero-inflated generalized Poisson 415 model to the distribution of output counts followed by regressing the parameters for each 416 peptide sequence based on the input read counts. Peptides with a reproducibility threshold (-417 log10(P-value)  2.3) in both technical repeats were further filtered for sporadic hits by 418 removing peptides which were also significantly enriched in at-least two mock-IP controls 419 (beads only). We also considered a peptide hit to be significant if the same peptide enriched 420 in more than one sample. Bonferroni correction for multiple testing. We considered a virus to be differentially enriched 435 in one group if the prevalence is 10% in at least one group, |log(OR)|  log10 (2)