Estimating seroconversion rates accounting for repeated infections by approximate Bayesian computation.

This study presents a novel approach for inferring the incidence of infections by employing a quantitative model of the serum antibody response. Current methodologies often overlook the cumulative effect of an individual's infection history, making it challenging to obtain a marginal distribution for antibody concentrations. Our proposed approach leverages approximate Bayesian computation to simulate cross-sectional antibody responses and compare these to observed data, factoring in the impact of repeated infections. We then assess the empirical distribution functions of the simulated and observed antibody data utilizing Kolmogorov deviance, thereby incorporating a goodness-of-fit check. This new method not only matches the computational efficiency of preceding likelihood-based analyses but also facilitates the joint estimation of antibody noise parameters. The results affirm that the predictions generated by our within-host model closely align with the observed distributions from cross-sectional samples of a well-characterized population. Our findings mirror those of likelihood-based methodologies in scenarios of low infection pressure, such as the transmission of pertussis in Europe. However, our simulations reveal that in settings of higher infection pressure, likelihood-based approaches tend to underestimate the force of infection. Thus, our novel methodology presents significant advancements in estimating infection incidence, thereby enhancing our understanding of disease dynamics in the field of epidemiology.

Severe Acute Respiratory Syndrome Coronavirus 2 Transmission in Georgia, USA, February 1-July 13, 2020.

The serial interval and effective reproduction number for coronavirus disease (COVID-19) are heterogenous, varying by demographic characteristics, region, and period. During February 1-July 13, 2020, we identified 4,080 transmission pairs in Georgia, USA, by using contact tracing information from COVID-19 cases reported to the Georgia Department of Public Health. We examined how various transmission characteristics were affected by symptoms, demographics, and period (during shelter-in-place and after subsequent reopening) and estimated the time course of reproduction numbers for all 159 Georgia counties. Transmission varied by time and place but also by persons' sex and race. The mean serial interval decreased from 5.97 days in February-April to 4.40 days in June-July. Younger adults (20-50 years of age) were involved in most transmission events occurring during or after reopening. The shelter-in-place period was not long enough to prevent sustained virus transmission in densely populated urban areas connected by major transportation links.

Children Are Exposed to Fecal Contamination via Multiple Interconnected Pathways: A Network Model for Exposure Assessment.

In recent decades, quantitative microbial risk assessment (QMRA) has been widely used to assess exposure to fecal microbes and associated health risks. In this study, a multipathway exposure assessment model was developed to evaluate exposure to fecal microbes for children under 5 in highly contaminated urban environments. Children had contact with various environmental compartments. The contamination levels of these compartments were estimated from fecal indicator counts in the environmental samples. Structured observations of child behavior (including activities, locations, and time) were used to model behavioral sequences as a dynamic network. The exposure model combines behavior sequences with environmental contamination, using additional exposure factors when needed, to estimate the number of fecal microbes transferred from environmental sources to human oral ingestion. As fecal exposure in a highly contaminated urban environment consists of contributions from multiple pathways, it is imperative to study their relative importance. The model helps us better understand the characteristics of the exposure pathways that may be driven by variation in contamination and by variable behavior, like hygiene and high-risk activities. Importantly, the model also allows prediction of the quantitative effects of an intervention-the expected reduction in exposure due to infrastructural or behavioral changes-by means of scenario studies. Based on experience with this exposure model, we make specific recommendations for additional studies of child behavior and exposure factors in order to fill critical information gaps and improve the model structure and assumptions.

Spatial Prediction of Coxiella burnetii Outbreak Exposure via Notified Case Counts in a Dose-Response Model.

We develop a novel approach to study an outbreak of Q fever in 2009 in the Netherlands by combining a human dose-response model with geostatistics prediction to relate probability of infection and associated probability of illness to an effective dose of Coxiella burnetii. The spatial distribution of the 220 notified cases in the at-risk population are translated into a smooth spatial field of dose. Based on these symptomatic cases, the dose-response model predicts a median of 611 asymptomatic infections (95% range: 410, 1,084) for the 220 reported symptomatic cases in the at-risk population; 2.78 (95% range: 1.86, 4.93) asymptomatic infections for each reported case. The low attack rates observed during the outbreak range from (Equation is included in full-text article.)to (Equation is included in full-text article.). The estimated peak levels of exposure extend to the north-east from the point source with an increasing proportion of asymptomatic infections further from the source. Our work combines established methodology from model-based geostatistics and dose-response modeling allowing for a novel approach to study outbreaks. Unobserved infections and the spatially varying effective dose can be predicted using the flexible framework without assuming any underlying spatial structure of the outbreak process. Such predictions are important for targeting interventions during an outbreak, estimating future disease burden, and determining acceptable risk levels.

Estimating seroprevalence of human papillomavirus type 16 using a mixture model with smoothed age-dependent mixing proportions.

BACKGROUND: The presence in serum of antibodies to viral antigens is generally considered a well-defined marker of past infection or vaccination. However, analyses of serological data that use a cut-off value to classify individuals as seropositive are prone to misclassification bias, in particular when studying infections with a weak serological response, such as the sexually transmitted human papillomavirus (HPV). METHODS: We analyzed the serological concentrations of HPV type 16 (HPV16) antibodies in the general Dutch population in 2006-2007, before the introduction of mass vaccination against HPV. We used a 2-component mixture model to represent persons who were seronegative or seropositive for HPV16. Component densities were assumed to be log-normally distributed, with parameters possibly dependent on sex. The age-dependent mixing proportions were smoothed using penalized splines to obtain a flexible seroprevalence profile. RESULTS: Our results suggest that HPV16 seropositivity is associated with higher antibody concentrations in women as compared with men. Seroprevalence shows an increase starting from adolescence in men and women alike, coinciding with the age of sexual debut. Seroprevalence stabilizes in men around age 40, whereas it has a decreasing trend from age 50 onwards in women. Analyses that rely on a cut-off value to classify persons as seropositive yield substantially different seroprevalence profiles, leading to a qualitatively different interpretation of HPV16 infection dynamics. CONCLUSIONS: Our results provide a benchmark for examining the effect of HPV16 vaccination in future serological surveys. Our method may prove useful for estimating seroprevalence of other infections with a weak serological response.

Kinetics of the long-term antibody response after meningococcal C vaccination in patients with juvenile idiopathic arthritis: a retrospective cohort study.

OBJECTIVES: The kinetics of the antibody response induced by meningococcal serogroup C (MenC) conjugate vaccination was analysed in patients with juvenile idiopathic arthritis (JIA) to assess their long-term protection against MenC disease. METHODS: In The Netherlands, a nationwide catch-up campaign was performed in 2002 during which children aged 1-19 years, including JIA patients, received the MenC conjugate vaccination. From 127 JIA patients, IgG antibody concentrations against MenC-polysaccharide were determined by a fluorescent-bead-based immunoassay in 402 serum samples collected between 2002 and 2010. Using a hierarchical linear regression model, the 8 years course of MenC-specific antibodies was analysed in four age groups (13-19, 9-12.9, 5-8.9 and 1-4.9 years), and in patients starting with methotrexate or biologicals. In 65 randomly selected samples, the correlation of MenC-specific IgG concentrations with serum bactericidal assay (SBA) titres was assessed. MenC-specific IgG concentrations at 4.2 years after vaccination were compared with those of 1527 age-matched healthy controls. RESULTS: MenC-specific IgG concentrations postvaccination were highest in patients aged 13-19 years at time of vaccination. Antibodies gradually waned over time in patients, but their estimated concentrations at 4.2 years postvaccination were similar to those measured in controls. MenC-specific IgG concentrations correlated well with SBA titres (r=0.72, p<0.001). By contrast with methotrexate, starting treatment with biologicals induced a trend towards accelerated decline of MenC-specific antibodies. CONCLUSIONS: Persistence of MenC-specific IgG antibodies in JIA patients is similar to healthy controls, but treatment with biologicals may induce accelerated antibody waning, resulting in unprotected patients who may need revaccination.

When case reporting becomes untenable: Can sewer networks tell us where COVID-19 transmission occurs?

Monitoring SARS-CoV-2 in wastewater is a valuable approach to track COVID-19 transmission. Designing wastewater surveillance (WWS) with representative sampling sites and quantifiable results requires knowledge of the sewerage system and virus fate and transport. We developed a multi-level WWS system to track COVID-19 in Atlanta using an adaptive nested sampling strategy. From March 2021 to April 2022, 868 wastewater samples were collected from influent lines to wastewater treatment facilities and upstream community manholes. Variations in SARS-CoV-2 concentrations in influent line samples preceded similar variations in numbers of reported COVID-19 cases in the corresponding catchment areas. Community sites under nested sampling represented mutually-exclusive catchment areas. Community sites with high SARS-CoV-2 detection rates in wastewater covered high COVID-19 incidence areas, and adaptive sampling enabled identification and tracing of COVID-19 hotspots. This study demonstrates how a well-designed WWS provides actionable information including early warning of surges in cases and identification of disease hotspots.

Quantifying transmission of norovirus during an outbreak.

BACKGROUND: Healthcare workers are thought to play a role in nosocomial transmission of norovirus, but the level and direction of norovirus transmission between patients and healthcare workers in sustaining transmission during an outbreak have not been quantified. METHODS: We developed a method for finding plausible transmission trees of who acquired their infection from whom. We applied the method to data from an outbreak of norovirus in 4 wards of a psychiatric institution in the Netherlands in 2008. The simulated transmission trees were based on serial intervals for time between symptom onsets, weighted for the number of days that healthcare workers were present. The obtained transmission trees were linked to the Barthel Index, a measure of patient reliance on healthcare in their basic daily activities. RESULTS: The dominant recognized transmission route was from patient to patient (64%), followed by patient to healthcare worker (29%). The overall estimated reproduction number for healthcare workers was low compared with patients (0.25 vs. 1.20; mean difference = 0.95 [95% confidence interval (CI) = 0.60 to 1.30]). The average number of all subsequent cases attributable to the downstream branch of one single infected healthcare worker in the transmission tree was 4.4 compared with 6.5 for cases attributable to one single infected patient (mean difference = 2.1 [95% CI = -4.7 to 8.9]). In the ward with patients requiring the highest level of care from healthcare workers, the attack rate among healthcare workers was highest. CONCLUSION: This approach provides a framework to quantify the magnitude and direction of transmission between healthcare workers and patients during a norovirus outbreak. The utility of this method in outbreaks of other infections and in different settings should be explored.

Dose response for Salmonella Typhimurium and Enteritidis and other nontyphoid enteric salmonellae.

This dose response assessment combines data from 6 human challenge studies and 44 outbreaks to determine infectivity and pathogenicity of several serotypes of nontyphoid Salmonella. Outcomes focus on the major serotypes Salmonella Enteritidis and Typhimurium, showing that Typhimurium is less infectious and has a lower probability of causing acute illness in infected subjects. The dose response relation of Salmonella Enteritidis is less steep than that of Typhimurium, indicating greater heterogeneity in infectivity and pathogenicity. This study revisits an older study with less flexible methods that could not combine the widely different outcomes of challenge studies and outbreaks, and had limited capability for dealing with missing information. Reported outcomes are in a format that allows use in calculations of uncertainty for quantitative risk assessment.

Dose-response relationship of Ralstonia solanacearum and potato in greenhouse and in vitro experiments.

Ralstonia solanacearum is the causative agent of bacterial wilt of potato and other vegetable crops. Contaminated irrigation water contributes to the dissemination of this pathogen but the exact concentration or biological threshold to cause an infection is unknown. In two greenhouse experiments, potted potato plants (Solanum tuberosum) were exposed to a single irrigation with 50 mL water (non-invasive soil-soak inoculation) containing no or 102 - 108 CFU/mL R. solanacearum. The disease response of two cultivars, Kondor and HB, were compared. Disease development was monitored over a three-month period after which stems, roots and tubers of asymptomatic plants were analyzed for latent infections. First wilting symptoms were observed 15 days post inoculation in a plant inoculated with 5x109 CFU and a mean disease index was used to monitor disease development over time. An inoculum of 5x105 CFU per pot (1.3x102 CFU/g soil) was the minimum dose required to cause wilting symptoms, while one latent infection was detected at the lowest dose of 5x102 CFU per pot (0.13 CFU/g). In a second set of experiments, stem-inoculated potato plants grown in vitro were used to investigate the dose-response relationship under optimal conditions for pathogen growth and disease development. Plants were inoculated with doses between 0.5 and 5x105 CFU/plant which resulted in visible symptoms at all doses. The results led to a dose-response model describing the relationship between R. solanacearum exposure and probability of infection or illness of potato plants. Cultivar Kondor was more susceptible to brown-rot infections than HB in greenhouse experiments while there was no significant difference between the dose-response models of both cultivars in in vitro experiments. The ED50 for infection of cv Kondor was 1.1x107 CFU. Results can be used in management strategies aimed to reduce or eliminate the risk of bacterial wilt infection when using treated water in irrigation.

Age-dependent patterns of infection and severity explaining the low impact of 2009 influenza A (H1N1): evidence from serial serologic surveys in the Netherlands.

Despite considerable research efforts in specific subpopulations, reliable estimates of the infection attack rates and severity of 2009 influenza A (H1N1) in the general population remain scarce. Such estimates are essential to the tailoring of future control strategies. Therefore, 2 serial population-based serologic surveys were conducted, before and after the 2009 influenza A (H1N1) epidemic, in the Netherlands. Random age-stratified samples were obtained using a 2-stage cluster design. Participants donated blood and completed a questionnaire. Data on sentinel general practitioner-attended influenza-like illness and nationwide hospitalization and mortality were used to assess the severity of infection. The estimated infection attack rates were low in the general population (7.6%, 95% confidence interval: 3.6, 11) but high in children aged 5-19 years (35%, 95% confidence interval: 25, 45). The estimated hospitalization and mortality rates per infection increased significantly with age (5-19 years: 0.042% and 0.00094%, respectively; 20-39 years: 0.12% and 0.0025%; 40-59 years: 0.68% and 0.032%; 60-75 years: >0.81% and >0.068%). The high infection attack rate in children and the very low attack rate in older adults, together with the low severity of illness per infection in children but substantial severity in older adults, produced an epidemic with a low overall impact.

Managing risks from virus intrusion into water distribution systems due to pressure transients.

Low or negative pressure transients in water distribution systems, caused by unexpected events (e.g. power outages) or routine operation/maintenance activities, are usually brief and thus are rarely monitored or alarmed. Previous studies have shown connections between negative pressure events in water distribution systems and potential public health consequences. Using a quantitative microbial risk assessment (QMRA) model previously developed, various factors driving the risk of viral infection from intrusion were evaluated, including virus concentrations external to the distribution system, maintenance of a disinfectant residual, leak orifice sizes, the duration and the number of nodes drawing negative pressures. The most sensitive factors were the duration and the number of nodes drawing negative pressures, indicating that mitigation practices should be targeted to alleviate the severity of low/negative pressure transients. Maintaining a free chlorine residual of 0.2 mg/L or above is the last defense against the risk of viral infection due to negative pressure transients. Maintaining a chloramine residual did not appear to significantly reduce the risk. The effectiveness of ensuring separation distances from sewer mains to reduce the risk of infection may be system-specific. Leak detection/repair and cross-connection control should be prioritized in areas vulnerable to negative pressure transients.

Estimation of the likelihood of fecal-oral HEV transmission among pigs.

Sources for human hepatitis E virus (HEV) infections of genotype 3 are largely unknown. Pigs are potential animal reservoirs for HEV. Intervention at pig farms may be desired when pigs are confirmed as a source for human infections, requiring knowledge about transmission routes. These routes are currently understudied. The current study aims to quantify the likelihood of pig feces in causing new HEV infections in pigs due to oral ingestion. We estimated the daily infection risk for pigs by modeling the fate of HEV in the fecal-oral (F-O) pathway. Using parameter values deemed most plausible by the authors based on current knowledge the daily risk of infection was 0.85 (95% interval: 0.03-1). The associated expected number of new infections per day was ∼4 (2.5% limit 0.1, the 97% limit tending to infinity) compared to 0.7 observed in a transmission experiment with pigs, and the likelihood of feces causing the transmission approached 1. In alternative scenarios, F-O transmission of HEV was also very likely to cause new infections. By reducing the total value of all explanatory variables by 2 orders of magnitude, the expected numbers of newly infected pigs approached the observed number. The likelihood of F-O transmission decreased by decreasing parameter values, allowing for at most 94% of infections being caused by additional transmission routes. Nevertheless, in all scenarios F-O transmission was estimated to contribute to HEV transmission. Thus, despite the difficulty in infecting pigs with HEV via oral inoculation, the F-O route is likely to cause HEV transmission among pigs.

Incidence and reproduction numbers of pertussis: estimates from serological and social contact data in five European countries.

BACKGROUND: Despite large-scale vaccination programmes, pertussis has remained endemic in all European countries and has been on the rise in many countries in the last decade. One of the reasons that have been discussed for the failure of vaccination to eliminate the disease is continued circulation of the pathogen Bordetella pertussis by mostly asymptomatic and mild infections in adolescents and adults. To understand the impact of asymptomatic and undiagnosed infection on the transmission dynamics of pertussis we analysed serological data from five European countries in combination with information about social contact patterns from five of those countries to estimate incidence and reproduction numbers. METHODS AND FINDINGS: We compared two different methods for estimating incidence from individual data on IgG pertussis toxin (PT) titres. One method combines the cross-sectional surveys of titres with longitudinal information about the distribution of amplitude and decay rate of titres in a back-calculation approach. The second method uses age-dependent contact matrices and cross-sectional surveys of IgG PT titres to estimate a next generation matrix for pertussis transmission among age groups. The next generation approach allows for computation of basic reproduction numbers for five European countries. Our main findings are that the seroincidence of infections as estimated with the first method in all countries lies between 1% and 6% per annum with a peak in the adolescent age groups and a second lower peak in young adults. The incidence of infections as estimated by the second method lies slightly lower with ranges between 1% and 4% per annum. There is a remarkably good agreement of the results obtained with the two methods. The basic reproduction numbers are similar across countries at around 5.5. CONCLUSIONS: Vaccination with currently used vaccines cannot prevent continued circulation and reinfection with pertussis, but has shifted the bulk of infections to adolescents and adults. If a vaccine conferring lifelong protection against clinical and subclinical infection were available pertussis could be eliminated. Currently, continuing circulation of the pathogen at a subclinical level provides a refuge for the pathogen in which it can evolve and adjust to infect vaccinated populations. Please see later in the article for the Editors' Summary.

The effect of mask use on the spread of influenza during a pandemic.

Face masks have traditionally been used in general infection control, but their efficacy at the population level in preventing transmission of influenza viruses has not been studied in detail. Data from published clinical studies indicate that the infectivity of influenza A virus is probably very high, so that transmission of infection may involve low doses of virus. At low doses, the relation between dose and the probability of infection is approximately linear, so that the reduction in infection risk is proportional to the reduction in exposure due to particle retention of the mask. A population transmission model was set up to explore the impact of population-wide mask use, allowing estimation of the effects of mask efficacy and coverage (fraction of the population wearing masks) on the basic reproduction number and the infection attack rate. We conclude that population-wide use of face masks could make an important contribution in delaying an influenza pandemic. Mask use also reduces the reproduction number, possibly even to levels sufficient for containing an influenza outbreak.

Noroviruses are highly infectious but there is strong variation in host susceptibility and virus pathogenicity.

Noroviruses are a major public health concern: their high infectivity and environmental persistence have been documented in several studies. Genetic sequencing shows that noroviruses are highly variable, and exhibit rapid evolution. A few human challenge studies have been performed with norovirus, leading to estimates of their infectivity. However, such incidental estimates do not provide insight into the biological variation of the virus and the interaction with its human host. To study the variation in infectivity and pathogenicity of norovirus, multiple challenge studies must be analysed jointly, to compare their differences and describe how virus infectivity and host susceptibility vary. Since challenge studies can only provide a small sample of the diversity in the natural norovirus population, outbreaks should be exploited as an additional source of information. The present study shows how challenge studies and 'natural experiments' can be combined in a multilevel dose response framework. Infectivity and pathogenicity are analysed by secretor status as a host factor, and genogroup as a pathogen factor. Infectivity, characterized as the estimated mean infection risk when exposed to 1 genomic copy (qPCR unit)is 0.28 for GI norovirus, and 0.076 for GII virus, both in Se+ subjects. The corresponding risks of acute enteric illness are somewhat lower, about 0.2 (GI) and 0.035 (GII), in outbreaks. Se- subjects are protected, with substantially lower risks of infection (0.00007 and 0.015 at a dose of 1 GC of GI and GII virus, respectively). The present study shows there is considerable variability in risk of infection and especially risk of acute symptoms following infection with norovirus. These challenge and outbreak data consistently indicate high infectivity among secretor positives and protection in secretor negatives.

Norwalk virus: how infectious is it?

Noroviruses are major agents of viral gastroenteritis worldwide. The infectivity of Norwalk virus, the prototype norovirus, has been studied in susceptible human volunteers. A new variant of the hit theory model of microbial infection was developed to estimate the variation in Norwalk virus infectivity, as well as the degree of virus aggregation, consistent with independent (electron microscopic) observations. Explicit modeling of viral aggregation allows us to express virus infectivity per single infectious unit (particle). Comparison of a primary and a secondary inoculum showed that passage through a human host does not change Norwalk virus infectivity. We estimate the average probability of infection for a single Norwalk virus particle to be close to 0.5, exceeding that reported for any other virus studied to date. Infected subjects had a dose-dependent probability of becoming ill, ranging from 0.1 (at a dose of 10(3) NV genomes) to 0.7 (at 10(8) virus genomes). A norovirus dose response model is important for understanding its transmission and essential for development of a quantitative risk model. Norwalk virus is a valuable model system to study virulence because genetic factors are known for both complete and partial protection; the latter can be quantitatively described as heterogeneity in dose response models.

Acute illness from Campylobacter jejuni may require high doses while infection occurs at low doses.

Data from a set of different studies on the infectivity and pathogenicity of Campylobacter jejuni were analyzed with a multilevel model, allowing for effects of host species (nonhuman primates and humans) and different strains of the pathogen. All challenge studies involved high doses of the pathogen, resulting in all exposed subjects to become infected. In only one study a dose response effect (increasing trend with dose) for infection was observed. High susceptibility to infection with C. jejuni was found in a joint analysis of outbreaks and challenge studies. For that reason four outbreaks, associated with raw milk consumption, were also included in the present study. The high doses used for inoculation did not cause all infected subjects to develop acute enteric symptoms. The observed outcomes are consistent with a dose response effect for acute symptoms among infected subjects: a conditional illness dose response relation. Nonhuman primates and human volunteers did not appear to have different susceptibilities for developing enteric symptoms, but exposure in outbreaks (raw milk) did lead to a higher probability of symptomatic campylobacteriosis.

Time to acquire and lose carriership of ESBL/pAmpC producing E. coli in humans in the Netherlands.

A subset of the study population from a cross-sectional study of carriership of ESBL/pAmpC-producing E. coli (ESBL-E) in the general population was followed up by five successive samples over an approximate half year period, leading to six samples in 333 persons. Fecal samples were cultured and analyzed for the presence of E. coli types as characterized by MLST, and ESBL/pAmpC genes were analysed by PCR and sequencing. The study included 255 persons who had a negative first sample, to allow observations of acquiring carriership of ESBL-E. Any individual record thus consisted of a series of snapshots of episodes of presence and absence of ESBL-E carriage. A survival model was built to estimate times to acquire or lose carriership, allowing for any combination of ESBL/pAmpC gene and E. coli MLST type. In carriers, the mean time to lose carriership was 1.1 (95% range 0.8-1.6) years. The estimated mean time to acquire carriership was 3.0 (95% range 1.6-6.3) years. Analysis of these times by ESBL/pAmpC gene found substantial variation among resistance genes both in persistence of carriership and in rates of acquiring carriership: blaCTX-M-1, blaCTX-M-14, blaCTX-M-15, blaCTX-M-27 and blaSHV-12 were easily acquired, but blaCTX-M-1 and blaSHV-12 were also easily lost, while blaCTX-M-15, blaCTX-M-27 and blaCMY-2 were more likely to persist. When in addition bacterial host types were included, some combinations appeared more persistent than others (blaCTX-M-1 in ST10 and ST58; blaCTX-M-14, blaCMY-2, and blaSHV-12 in ST69), or were acquired with higher frequency (blaCTX-M-14 in ST38, ST69, and ST131; blaCTX-M-15 and blaCTX-M-27 in ST131; blaSHV-12 in ST69). The relatively short duration of carriership means that when an intervention drastically reduces the exposure of humans to ESBL-E, the prevalence will be halved in 0.66 years. The observed differences between carriage rates of ESBL/pAmpC genes and E. coli strains need further investigation.

Correction: Time to acquire and lose carriership of ESBL/pAmpC producing E. coli in humans in the Netherlands.

[This corrects the article DOI: 10.1371/journal.pone.0193834.].