Does plasmid-based beta-lactam resistance increase E. coli infections: Modelling addition and replacement mechanisms.

Infections caused by antibiotic-resistant bacteria have become more prevalent during past decades. Yet, it is unknown whether such infections occur in addition to infections with antibiotic-susceptible bacteria, thereby increasing the incidence of infections, or whether they replace such infections, leaving the total incidence unaffected. Observational longitudinal studies cannot separate both mechanisms. Using plasmid-based beta-lactam resistant E. coli as example we applied mathematical modelling to investigate whether seven biological mechanisms would lead to replacement or addition of infections. We use a mathematical neutral null model of individuals colonized with susceptible and/or resistant E. coli, with two mechanisms implying a fitness cost, i.e., increased clearance and decreased growth of resistant strains, and five mechanisms benefitting resistance, i.e., 1) increased virulence, 2) increased transmission, 3) decreased clearance of resistant strains, 4) increased rate of horizontal plasmid transfer, and 5) increased clearance of susceptible E. coli due to antibiotics. Each mechanism is modelled separately to estimate addition to or replacement of antibiotic-susceptible infections. Fitness costs cause resistant strains to die out if other strain characteristics are maintained equal. Under the assumptions tested, increased virulence is the only mechanism that increases the total number of infections. Other benefits of resistance lead to replacement of susceptible infections without changing the total number of infections. As there is no biological evidence that plasmid-based beta-lactam resistance increases virulence, these findings suggest that the burden of disease is determined by attributable effects of resistance rather than by an increase in the number of infections.

Introduction

It is unknown to what extent the global increase in infections caused by antibiotic-resistant bacteria (ARB) during the last decades has changed the burden of disease. ARB infections may occur on top of infections caused by non-ARB, a scenario labelled as addition [1], but they may also replace non-ARB infections. Infections caused by ARB are more difficult to treat, resulting in more adverse health outcomes [2], which increases the healthcare burden.

In the addition scenario, there is an increase in the total number of infections together with attributable harm created by those infection caused by ARB. In the replacement scenario, the increased burden of disease due to resistance results solely from the attributable harm created by resistant compared to susceptible infections, as the total number of infections remains stable. Quantifying the relative contribution of both scenarios is of critical importance for quantifying the burden of disease created by ARB. Longitudinal observational data have been used to estimate the relative importance of addition and replacement [1,3], but the validity of these approaches suffered–inevitably–from other time-dependent changes and between study-groups differences that may influence overall incidence of infections, e.g., changes in medical procedures, population age, antimicrobial stewardship, and infection prevention measures.

As shown in Fig 1 occurrence of addition or replacement of infections caused by ARB cannot be deduced from observed time trends. Fig 1A depicts a scenario with an unobserved history and observed time trends of infections. Fig 1B–1D depict three different scenarios that can explain the same observed time trend namely, addition, replacement or a combination of the two. The scenarios depicted in Fig 1B–1D have an equal number of resistant and susceptible infections and are compatible with the observed data. Changing the unobserved prevalence suggests a different interpretation on whether addition or replacement occurs. However, as the history is unobserved it cannot be decided based on time trend data whether addition or replacement has occurred. To illustrate this with real observational data, we determined the extended-spectrum beta-lactamase (ESBL) E. coli bacteraemia prevalence in the Netherlands between 2014 and 2019.

Fig 1

Three scenarios explained by addition to and replacement of susceptible infections.

The observed prevalence of susceptible and resistant infections is depicted in panel a. In panel b, c and d, three potential scenarios for the prevalence in the unobserved period are depicted. Each scenario suggests a different interpretation regarding addition or replacement.

We, therefore, developed a mathematical model, comprising three populations, hospitalized patients, recently hospitalized patients and the general population to determine effects of various explicitly stated biological mechanisms that may change ARB carriage and subsequent infection, and thereby estimate whether these mechanisms cause replacement of or addition to susceptible infections. We start with a neutral mathematical model and in the base-case scenario we assume that susceptible and resistant strains can co-exist at any ratio if the two strains behave identical in all aspects. In this model, individuals either have a stable, high-density, colonization or a disturbed flora with low-density colonization. The neutral null model is further explained in the base-case scenario in the methods section and the concept of a neutral null model has also been discussed previously [4]. Next we move away from the neutral null model and modify the characteristics of the ARB strain to create benefits or costs of resistance. A fitness advantage for ARB will ultimately lead to dominance of ARB variants over non-ARB strains [5]. Multiple mechanisms, such as increased transmission or antibiotic use, can create a dynamical advantage of resistant over susceptible strains [6,7]. Similarly, resistance can be associated with a fitness cost, which fuelled the hope that natural selection would–in the absence of beneficial selective pressure—eventually lead to a reduction in ARB [8]. An example of a fitness cost could be faster clearance. Obviously, ARB (dis)advantages change in the context of antibiotic exposure. We focus on mechanisms that influence the prevalence of colonisation with ARB and non-ARB and keep host mechanisms affecting individuals’ risks of acquisition of carriage or infection stable.

First, we study two mechanisms with a cost of resistance (i.e., decreased growth rates of ARB and increased clearance of ARB, with clearance defined as the rate at which individuals go from high density colonization to low density colonization), and subsequently five mechanisms that benefit ARB. The first is increased virulence, defined as an increased probability to develop infection once colonisation has been established. This increased risk of infection may result from three separate mechanisms that have the same dynamical effects in our model, such as an increased bacterial load, higher intrinsic virulence of a bacterium, and failure of antibiotic prophylaxis [6,9,10]. Moreover, increased virulence may be related to other genes harbored on the ESBL-plasmid [6]. The second is increased transmission, which leads to more ARB acquisitions and, thereby, to more subsequent infections, despite a stable infection rate [11]. The third is lower clearance, which makes ARB carriage more persistent, prolonging the risk period for infection, despite a stable infection rate [12]. The fourth is within-host plasmid transfer of resistance to susceptible strains, which increases the number of ARB strains and subsequently ARB infections. The fifth is selective antibiotic pressure, which increases probabilities of ARB acquisitions through cross-transmission due to lower density of susceptible bacteria in subjects receiving antibiotics, increasing susceptibility to acquisition with resistant bacteria. Cross-transmission is defined here as the transfer of bacteria from a source, in this model, another individual, to a receiver which in mechanism five is the subject receiving antibiotics and subsequently having a lower density of susceptible bacteria.

Moreover, we model two scenarios of combined mechanisms. The “mixed scenario” combines increased ARB hospital transmission and increased ARB clearance in the community, thereby maximizing ARB benefits among hospitalized patients, e.g., due to high selective antibiotic pressure, and maximizing fitness costs of ARB in the absence of selective antibiotic pressure. The “double benefit scenario” combines increased ARB virulence and increased hospital transmission of ARB (through increased selective antibiotic pressure and higher contact rate by hospital staff acting as potential vectors), thereby maximizing ABR benefits in hospitalized patients.

We focus on E. coli with plasmid-based beta-lactam resistance, since these are widely prevalent in the population and an important cause of both community-acquired and hospital-acquired infections [13]. Moreover, ESBL and carbapenem resistant E. coli have been set as a critical priority for research and development by the WHO on the global list of ARB [14].

Results

From 2014 till 2019 the ESBL E. coli bacteraemia prevalence in the Netherlands remained stable at around 5·4%, but the total number of E. coli bacteraemia increased annually with 3·4% (S1 Table and Fig 2). Using the scenarios depicted in Fig 1, the observed time trend could reflect both addition to and replacement of susceptible E. coli bacteraemia (Fig 2).

Fig 2

Number of susceptible E. coli and ESBL E. coli bacteraemia in 24 Dutch hospitals from 2014 to 2018.

In the neutral model, of 100,000 subjects, 177 were hospitalised, 694 were former patients and 99,129 were in the community. Numbers of plasmid-based beta-lactam resistant and susceptible E. coli infections were 122 and 2,320 per 100,000 subjects annually. Infections caused by ARB occurred predominantly in former (n = 67) and hospitalised patients (n = 34), and 21 occurred in the community. Infections by non-ARB also occurred predominantly in former(n = 1,275) and hospitalised patients (n = 651), and less frequently in the community (n = 395). The infection incidence caused by ARB and non-ARB together is 8% per year in the hospital, 16% in former patients and 0.5% in the community, yielding an infection incidence in the population (former patients and the community) of 2% per year.

The effects of changing rates related to costs and benefits of ARB in 10 years are depicted in Fig 3. Naturally, introducing costs of resistance without benefits of resistance reduces the carriage of resistant strains and eventually leads to elimination of resistant strains (S4 Fig and S7 Table). Whereas introducing a benefit without costs of resistance leads to replacement or addition.

Fig 3

Annual susceptible and resistant infections per 100,000 people per mechanism in 10 years of time.

Increased ARB virulence is associated with a linear increase in the incidence of infections caused by ARB: a 10% increased virulence leads to a 10% increase of infections caused by ARB with an unchanged incidence of infections caused by non-ARB. Thus, increased virulence leads to an equivalent addition to the total number of infections, as can be seen in Fig 3. The transmission rates, e.g. from S to R or R to RR, are not affected by increased virulence and the subsequent increase in the incidence of ARB infections. Because the transmission dynamics are not altered by increased virulence, resistant and susceptible strains coexist in our model, which precludes eradication of susceptible strains.

The other four mechanisms associated with a dynamical benefit of ARB, all do lead to more infections caused by ARB, but only through replacing infections caused by non-ARB. In fifty years, a 10% increase in ARB transmission rates leads to a 13% increase in infections caused by ARB and a 50% decreased clearance of ARB leads to a 72% increase in infections caused by ARB which increases the prevalence of carriage with ARB from 5% to 8·6%. The total number of infections remains stable. Introduction of plasmid acquisition also causes a linear replacement of infections caused by non-ARB. In 10 years, the number of infections caused by ARB increases with 67%.

Administering a new course of antibiotics that clears 50% of non-ARB E. coli results in replacement of non-ARB. Restricting the use of a course of antibiotics in hospitalized patients only, reduces replacement of infections caused by non-ARB (Fig 3).

Contrary to all other mechanisms, effects of increased virulence are not time dependent, as shown in the equal trends in S2 and S3 Figs. This implies that an increase in virulence leads immediately to additional infections and this increase then remains stable. Other mechanisms benefitting ARB will lead to complete replacement of non-ARB, whereas increased virulence will lead to more infections without extinction of infections caused by non-ARB (S3 Fig and S6 Table). Costs of resistance will lead to replacement of ARB.

Thereafter, we investigated two scenarios with combined mechanisms. The mixed scenario provides maximum benefits for ARB among hospitalized patients, but also maximum fitness costs of resistance in the absence of selective antibiotic pressure. In this scenario the incidence of infections caused by ARB declines with 18% (after 50 years), whereas the incidence of infections caused by non-ARB increases (with 1%), leaving the total number of infections unaffected. In the double benefit scenario (with double benefits for ARB due to increased virulence and more transmission in hospitals) the incidence of infections caused by ARB would increase with a factor 5·75 (after 50 years), at the cost of infections caused by non-ARB (-25%), and with a 5% increase in the total number of infections (S3 Table).

Finally, we investigated the effect of decreasing antibiotic use by 50% in all three populations on the number of infections under the conditions that each 1) mechanisms was changed by 50% and 2) by 75%. The largest decrease in resistant infections was observed for increased transmission, decreased growth, and increased clearance. Under all conditions, decreased antibiotics use lowered the number of resistant infections. The results can be found in S9 Table and S5 and S6 Figs.

Discussion

The modelling of different biological mechanisms affecting E. coli with and without plasmid-based beta-lactam resistance revealed that increased virulence is the only mechanism that would lead to a higher burden of infections due to emerging antibiotic resistance. Such an increase in the incidence of infections would occur instantaneously and linearly, without affecting the number of infections caused by non-ARB. All other biological mechanisms investigated, such as increased transmission, decreased clearance, plasmid transfer, and selection through antibiotic use resulted in scenarios in which infections caused by ARB would increase at the cost of infections caused by non-ARB, with a stable overall incidence of infections. This can be explained by increased virulence being the only mechanism which increases the infection rate. The other mechanisms influence parameters affecting colonisation, and subsequently affect the ratio at which resistant and susceptible infections occur. A difference in the ratio of colonisation with susceptible and resistant strains will lead to an equal number of total infections if the infection rate for resistant and susceptible strains is assumed to be equal, because we start with a neutral null model. Moreover, if resistance has both costs and benefits, emergence may lead to a stable co-existence of susceptible and resistant strains, for instance if resistance is beneficial in one niche, e.g., the hospital, but disadvantageous in the community, as reflected in the mixed scenario.

The starting point of this work was the recognition of difficulties in interpreting observational longitudinal data on antibiotic resistance prevalence. In the Netherlands the incidence of E. coli bacteraemia increased over time towards a stable prevalence of ESBL-producing E. coli bacteraemia (Fig 2). Yet, where the prevalence of ESBL-producing strains among 173 E. coli bacteraemia isolates was zero around 2000 in the Netherlands [15], it was around 5% in 2014. With a stable population size and unchanged hospital policies related to infection diagnosis the 5% increase in resistance in-between both time points could have resulted from both addition and replacement mechanisms.

The biological plausibility of some of the biological mechanisms that we studied is not obvious. For instance, there is no biological prove that resistance in itself increases virulence [8]. If ARB would not have increased virulence, the possibility of the addition scenario would be excluded as we found that only increased virulence results in additional infections. Furthermore, the dogma that antibiotic resistance, such as ESBL-production in E. coli, is associated with fitness costs has been questioned [6,8,16]. Indeed, if resistance comes without fitness costs through, for instance, increased clearance or decreased growth, ESBL-producing E. coli will not to be completely replaced by susceptible E. coli and will persist, even in the absence of antibiotic selective pressure.

Two previous studies attempted to disentangle whether a longitudinal increase in the incidence density of infections caused by ARB resulted from replacement or addition. Firstly, de Kraker et al. [17] investigated bacteraemia trends of five pathogens; Staphylococcus aureus, E. coli, Streptococcus pneumoniae, Enterococcus faecalis and Enterococcus faecium, using the European Antimicrobial Resistance Surveillance System (EARSS) database from 2002 to 2008. They report an increasing incidence in bacteraemia and suggested that this was mainly driven by resistant strains, implying that resistant clones add to rather than replace infections caused by susceptible bacteria. Secondly, Ammerlaan et al. [1] investigated temporal trends in annual incidence densities (events per 100,000 patient-days) of nosocomial bloodstream infections caused by methicillin-resistant Staphylococcus aureus (MRSA), ARB other than MRSA and non-ARB in 14 hospitals between 1998 and 2007. Seven hospitals had high incidence of MRSA infection in 1998 and no specific program to control the spread of MRSA. The other seven hospitals had low incidence of nosocomial infections with MRSA and infection control programs to maintain low incidence levels of MRSA over this period. During the 10-year period the increase in the incidence density of non-ARB infections was similar in both hospital groups. Yet, the incidence of infections caused by ARB increased from 3.0 to 4.7 per 100,000 patient-days between 1998 and 2007 in hospitals that effectively controlled ARB infections, and from 4.6 to 29.1 per 100,000 patient-days between 1998 and 2007 in hospitals with pre-existing higher MRSA and ARB infection rates. From this the authors concluded that ARB infections were additive to non-ARB infections. The observed increase in incidence density of ARB was 10-fold higher than in non-ARB. Based on the findings in the current study, for such an increase to be fully explained by addition would require a similar (i.e., linear) increase in bacterial virulence (S5 Text and S8 Table). It, therefore, seems more plausible that also in these hospitals replacement had occurred rather than addition.

A limitation of our study is that colonisation density is not modelled quantitatively, precluding modelling of within-host competition between strains. Yet, dynamics of within-host competition have not been accurately determined and it is, therefore, unknown, whether addition of this component improves modelling of the population level colonisation prevalence and infection incidence. Further, we limit acquisition routes of plasmid-based beta-lactam resistant E. coli and E. coli to humans, neglecting other reservoirs, such as livestock. Recent findings reported by Mughini-Gras et al. [18] imply that–at least in the Netherlands—the human reservoir is largely responsible for transmission among humans, with relatively little contribution of the animal reservoir.

In our study we explicitly investigate the effects of multiple mechanisms on the incidence of infections caused by ARB using a single model, which can also be used for other pathogens and in which other mechanisms can be included. In the current analyses the assumption that all individuals carry E. coli is vital. Investigating dynamics of pathogens for which this assumption does not hold, such as Staphylococcus aureus, would require an additional compartment for uncolonised subjects. Moreover, as we assume that everyone is colonised with E. coli, a mere increase in prevalence of the resistant strain, assuming that characteristics leading to infection are equal to the susceptible strain, does increase the total number of infections. An increase in prevalence of the resistant strain may cause an increase in the total number of infections if a bacterium is assumed not the be carried by everyone, as those uncolonised are not at risk for infection and become at risk once colonised.

Increased virulence of ARB, compared to non-ARB, is the only mechanism studied that would increase the total number of infections due to emergence of ARB. Other mechanisms, such as increased transmission, decreased clearance, plasmid transfer, and antibiotic use result in replacement of non-ARB infections by ARB infections, with a stable incidence of the total number of infections.

Methods

Time trend data

As an illustration of the use of time trend data to estimate addition and replacement, we estimated changes in the prevalence of ESBL in community-acquired E. coli bacteraemia in the Netherlands from 2014 to 2018. We used data from the national surveillance system of antimicrobial resistance (ISIS-AR) based on routinely collected data from medical microbiological laboratories [19]. We selected blood-isolates containing E. coli (one isolate per patient per year) with a sample moment maximum two days after hospital admission, or sampled at the emergency department or outpatient clinic, to define community-acquired bacteraemia. ESBL E. coli were determined based on ESBL-confirmatory tests or resistance to cefotaxime/ceftriaxone and/or ceftazidime, according to local practice. For the analysis, 24 hospitals with complete data available for the total study period were used.

Theoretical framework

To study the effects of seven mechanisms on the number of E. coli and plasmid-based beta-lactam resistant E.coli, we use a continuous time deterministic model, similar to Cooper et al. [20], consisting of a hospital population, a population of recently discharged patients, and the community (Fig 4A). S2 Table contains the parameters and definitions per mechanisms and S4 Table provides the parameters of transition pathways between the compartments per population. The total population size is constant and we ignore death and birth rates as these are substantially smaller than transmission and decolonisation rates in most parts of the population. Hospitalized patients are discharged to the former patient population at a constant rate. Subjects are hospitalized at constant rates, with a higher rate in the former patient population. Former patients transit to the community at constant rates, after which their hospital admission rate becomes lower.

Fig 4

Compartmental model of ESBL E. coli and E. coli colonisation states.

We assume homogeneous mixing in the hospital, allowing transmission between hospitalized patients at a rate four times higher than in the community due to frequent contacts with healthcare workers and medical devices that may act as vectors for transmission. Among former patients the transmission rate is assumed to be higher compared to the community, also because of interactions with healthcare workers [21]. Further, transmission occurs between and within the former patient and community, which is one homogeneously mixed population, but we assume no transmission between the community and the hospital. We use frequency-dependent transmission based on the total fraction of colonised individuals. Mathematically, it means that the rate at which an individual acquires an E. coli strain is proportional to the fraction of individuals colonised with that strain. Moreover, the three populations have equal ARB decolonisation rates and transition rates between populations are independent of colonisation status. S4 Table presents the model parameters.

Base-case scenario

We developed a neutral null base-case model for resistant and susceptible bacteria which are indistinguishable except for their antibiotic susceptibility. In a neutral null model neither strain can have a selective advantage and the prevalence ratio of the two strains should remain constant over time [4]. In our base-case, we assume no antibiotic use and, therefore, the base-case model lacks an intrinsic mechanism to promote stable coexistence between strains. When we consider other scenarios, we explicitly state the selective (dis)advantages of a strain [4]. We choose a starting ARB colonisation prevalence of 5% in all populations [22].

E. coli is a commensal bacterial species but pathogenic variants also exist. Due to, amongst other ways, point mutations and plasmid transfers E. coli strains can vary in their characteristics. In our model, we assume only two strains, a susceptible and a resistant strain. Carriage with one or two of the strains is possible, but we assume everyone always carries at least one E. coli strain. Hence there is no compartment of uncolonized subjects and individuals can be colonized with resistant strains (R-compartment), susceptible strains (S-compartment) or both (SR-compartment). Individuals in the S-compartment may acquire ARB and then move to the SR-compartment. Similarly, individuals in the R-compartment move to the SR-compartment after acquisition of susceptible strains. To maintain neutrality in absence of selective advantages, we added SS and RR compartments [4] which we interpret as states with a high density of colonization. Without these compartments, introduction of ARB into a population in which everyone could acquire ARB would lead to an equilibrium prevalence of 50%, even in absence of selective advantages [23]. Therefore, subjects colonised with susceptible (S) or resistant strains ® remain susceptible to other strains as long as they are “low density colonised”. Once progressed to high density colonised (labelled SS, RR or SR) they are non-susceptible to acquisition. We thus interpret the high-density RR, SR and SS-compartments as stable intestinal flora, preventing colonisation with other bacteria. We stress that high density colonisation is not the same as having an infection.

Infection rates are equal in high- and low-density compartments and proportional to the number of subjects colonised with resistant and susceptible strains. The infection rate is defined as the number of infections per day per person. Infections in the SR-compartment occur at an infection rate which is the average of the infection rate of the S- and R-compartment. The proportion of the infections in the SR-compartment, which are caused by susceptible bacteria and by resistant bacteria is proportional to the infection rate in the S- and the R-compartment, as can be seen in formula 16 and 17 in S1 Text. Differential equations specifying these models are given in S1 Text. E. coli colonisation states are depicted in Fig 4B.The parameters used in the model can be found in S4 Table.

ARB benefits and costs

To investigate the dynamical effects of benefits and costs of resistance, we distinguish seven mechanisms, and determine the effects by gradually increasing or decreasing the applicable rates with 10% to 100%, compared to the rate in the neutral model. Mechanisms benefitting resistant strains increase the ARB colonisation prevalence. As we assume a maximum colonisation capacity, meaning that new strains cannot successfully colonize individuals in state RR, SR or SS, an increased ARB prevalence renders less people susceptible for acquiring new colonisation with susceptible strains. Details of these biological mechanisms are described in S2 Table. The influence of each mechanism on the paths in the compartmental model is shown visually in Fig 5.

Fig 5

The biological mechanisms shown in the compartmental model.

The mechanism of plasmid transfer was restricted to within-host and within-species transfer, thereby increasing the transition from SR to RR carriage status. Additionally, a scenario with plasmid transfer from other pathogens to E. coli was analysed, see S4 Text and S1 Fig.

Antibiotic use occurs in all populations—independent of infection. In the Dutch community antibiotic courses are prescribed at a rate of 0.374 per year per person [24], and we assumed that 50% of persons receiving a course of antibiotics clear susceptible E. coli. Conceptually, this moves a subject from SS and SR to S and R compartments, respectively, after which the likelihood of moving to SS, RR and SR increases (Fig 4B). For hospitalized patients we assumed that 33.8% received antibiotics on any given day, based on published data [24,25]. We did not consider repeated prescriptions in the community or during hospitalization, as the intestinal flora will already be influenced by the first course of antibiotics. For changes in antibiotic use we also investigate a scenario with daily antibiotic use of 33.8% of all Dutch hospitalized subjects [25] and without antibiotic use in former patients and the community. Furthermore, we modelled two scenarios of combined mechanisms. The “mixed scenario” combines 20% increased ARB hospital transmission with 20% increased ARB clearance in the community, thereby maximizing the benefits of ARB among hospitalized patients, e.g., due to high selective antibiotic pressure, and maximizing fitness costs of ARB in the absence of selective antibiotic pressure. The “double benefit scenario” combines 20% increased ARB virulence (through increased selective antibiotic pressure) and 20% increased hospital transmission of ARB (through both increased selective antibiotic pressure and higher contact rate by hospital staff acting as potential vectors), thereby maximizing the benefits of ARB in hospitalized patients. Scenarios and mechanisms are studied during time periods of 10 years, 50 years and infinite time. The outcome is the annual number of infections after the subsequent time period (S5, S6, and S7 Tables and S2, S3, and S4 Figs).

Differential equations of the mathematical model.

(DOCX)

Click here for additional data file.

Infection rate calculations.

(DOCX)

Click here for additional data file.

Calculations admission and readmission rates.

(DOCX)

Click here for additional data file.

External plasmid transfer.

(DOCX)

Click here for additional data file.

Calculations of increased virulence.

(DOCX)

Click here for additional data file.

Compartmental model of E. coli with plasmid transfer from other sources.

(DOCX)

Click here for additional data file.

Annual ESBL E. coli infections per 100,000 people per mechanism in 10 years of time.

(DOCX)

Click here for additional data file.

Annual ESBL E. coli infections per 100,000 people per mechanism in in 50 years of time.

(DOCX)

Click here for additional data file.

Annual ESBL E. coli infections per 100,000 people per mechanism when letting time run to infinity.

(DOCX)

Click here for additional data file.

Annual ESBL E. coli infections per 100,000 people with 50% change in mechanisms and decreased antibiotics use.

(DOCX)

Click here for additional data file.

Annual ESBL E. coli infections per 100,000 people with 75% change in mechanisms and decreased antibiotics use.

(DOCX)

Click here for additional data file.

Annual prevalence of ESBL in E. coli bacteraemia in 24 Dutch hospitals.

(DOCX)

Click here for additional data file.

Scenarios and mechanisms studied in this paper.

(DOCX)

Click here for additional data file.

In 50 years, the number of resistant and susceptible infections per 100,000 inhabitants under different scenarios.

(DOCX)

Click here for additional data file.

Parameters used in the model.

(DOCX)

Click here for additional data file.

In 10 years, the number of infections per 100,000 people per year for each mechanism.

(DOCX)

Click here for additional data file.

In 50 years, the number of infections per 100,000 people per year for each mechanism.

(DOCX)

Click here for additional data file.

Letting time run to infinity, the number of infections per 100,000 people per year for each mechanism.

(DOCX)

Click here for additional data file.

Observed growth reported in Ammerlaan et al. [4] and expected growth.

(DOCX)

Click here for additional data file.

In 10 years, the number of infections for 50% and 75% change per mechanism when using antibiotics and when decrease overall antibiotic use with 50%.

(DOCX)

Click here for additional data file.

Transfer Alert

This paper was transferred from another journal. As a result, its full editorial history (including decision letters, peer reviews and author responses) may not be present.

12 Aug 2021

Dear Mrs. Godijk,

Thank you very much for submitting your manuscript "Does plasmid-based beta-lactam resistance increase E. coli infections: Modelling addition and replacement mechanisms" for consideration at PLOS Computational Biology.

As with all papers reviewed by the journal, your manuscript was reviewed by members of the editorial board and by several independent reviewers. In light of the reviews (below this email), we would like to invite the resubmission of a significantly-revised version that takes into account the reviewers' comments.

We cannot make any decision about publication until we have seen the revised manuscript and your response to the reviewers' comments. Your revised manuscript is also likely to be sent to reviewers for further evaluation.

When you are ready to resubmit, please upload the following:

[1] A letter containing a detailed list of your responses to the review comments and a description of the changes you have made in the manuscript. Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

[2] Two versions of the revised manuscript: one with either highlights or tracked changes denoting where the text has been changed; the other a clean version (uploaded as the manuscript file).

Important additional instructions are given below your reviewer comments.

Please prepare and submit your revised manuscript within 60 days. If you anticipate any delay, please let us know the expected resubmission date by replying to this email. Please note that revised manuscripts received after the 60-day due date may require evaluation and peer review similar to newly submitted manuscripts.

Thank you again for your submission. We hope that our editorial process has been constructive so far, and we welcome your feedback at any time. Please don't hesitate to contact us if you have any questions or comments.

Sincerely,

Roger Dimitri Kouyos

Associate Editor

PLOS Computational Biology

Virginia Pitzer

Deputy Editor-in-Chief

PLOS Computational Biology

***********************

Reviewer's Responses to Questions

Comments to the Authors:

Please note here if the review is uploaded as an attachment.

Reviewer #1: The authors are investigating with the help of mathematical modelling whether the presence of an antibiotic resistant strain causes an increase in the total number of infections within a population in comparison to a population with only a sensitive strain present or whether the total number of infections is stable and sensitive infections are just replaced by resistant infections. For this investigation different characteristics of resistant strains are considered and their effect. As an exemplary pathogen E.coli is examined with and without plasmid-mediated antibiotic resistance and the model is parameterized according to Dutch health care data.

The work covers interesting aspects and shows relevance for a topic of high interest: estimating the burden of antibiotic resistant bacteria. However, I had difficulties with the interpretation of the model and comprehending certain model choices. I think the paper would benefit in some parts with a more detailed explanation and justification of model choices. Generally, it is not an easy task to explain structural neutral epidemiological models, in the following I attach comments, which may be of aid to the authors.

1) As the colonization of the intestinal flora is considered, one patch, I would find the model structure described in equation (14) in the Lipsitch paper easier to interpret biologically as it is hard to see a biological correspondence of the single double colonized compartments if the intestine is not considered as multiple patches (co-infection through infection of multiple patches of one host). Furthermore, in a structure like (14) there would be an interpretable transmission route from the co-colonised compartment to the single colonised compartment.

I am a bit confused by the terminology of “low density colonised” and “high density colonised”, for me this suggests for example that the SR (or SS) compartment has a higher total population of bacteria than the S / R compartment, which should not be the case in a neutral model considering one patch that can be colonized by indistinguishable strains. In L. 285 a maximum colonisation capacity is assumed, which seems to be higher for the (SS, RR, SR) compartments than for S and R. In the simplest case the intestine can be interpreted as one environment, the carrying capacity for E.coli should therefore be the same for the double colonised compartments and the singly colonised compartments, except if we assume that the intestine can be seen as two patches. If two patches are assumed in this paper, then further explanation would be helpful.

2) If we have two indistinguishable strains (neutral model) shouldn’t the clearance rate be the same for resistant and non-resistant bacteria? (Referring to table S4, where it is described how rates should look like in order to get a neutral model)

3) It took me quite a while to understand what is actually assumed by clearance, as not the clearance of infection or colonization is referred to as clearance, but the transition of a double colonised compartments to single colonised compartments.

Minor comments:

1) L.73 the short form ESBL E. coli is used without introducing it before

2) It is the first time where the neutral mathematical model is mentioned (besides the abstract) I think here a reference to a part of the text where a neutral model is explained in more detail would be helpful, as I expect some readers not to be familiar with the concept

3) L. 82/83 would be nice to name an example for the multiple mechanisms

4) L. 98 definition of cross-transmission?

5) L. 92 reference (6) regarding beta-lactamase-plasmids increasing virulence is cited as a reference on how virulence can be increased, but in the abstract there it is stated that there is no biological evidence that plasmid-based beta-lactam resistance increases virulence. From what the authors describe in the text I understood that the resistance genes itself do not cause an increased virulence. If plasmid-based resistance is considered, can the resistance gene be looked at as a separate unit? Should not the other additional genes which are on the plasmid that cause higher virulence be associated to the resistance?

6) When introducing the ODES, a reference to table S4 would be helpful to see the definitions of the parameters and the nomenclature/description of compartments

7) Supplement end of the ODE: yr_rro typo?

8) Maybe I have overseen it, but a mathematical description of the number of people that get infected (infection + cure of infection) and to see where the rate Omega was used would be nice for understanding

Reviewer #2: Thank you for the opportunity to read this paper on understanding whether resistant infections in high carriage prevalent commensal bacteria such as E coli emerge in addition to or instead of susceptible infections. The paper offers one of the first (if not the first?) modelling approach to a really difficult question, and I am interested in what it finds.

For me, the modelling presentation could be tightened up that would improve readability and interpretation.

I also have a few queries about the approach and the findings on which I would appreciate the author responses.

1. The testing of the hypotheses is highly conditional on the set up of the model. However, I found the communication of the different models to be very confusing, with a lot of text description and poor connection with the equations given in the Appendix that were not well formatted or particularly well defined. As the model structures are so important to the interpretation of the conclusion of the paper, I would strongly advise the authors to do all their hard work justice by putting together a main text multi-panel compartment diagram that intuitively and quantitatively describes their models (see under Specific comments below for some suggestions).

2. The research question concerns the incidence of infections, rather than carriage prevalence. It may be a simple point, but it took me a while to understand the set up of the model because everyone in the population is assumed to be colonised, either singly or dually. I would like the authors to be a little more explanatory in their motivation within the main text. That is, the typical resistance data we have characterise the resistance frequencies of infections rather than that of carriage. To observe increased numbers of resistant infections we could imagine them either replacing sensitive infections or adding to the total infections (as per Figs 2-3). Alternatively, we could imagine an increase in the number of resistant infections because of an increase in the carriage prevalence of resistant strains (again, either instead of or in addition to sensitive strains). Is the second option not supported by data? The authors will know this better than me, but I think that this alternative should be spelled out and reasons given for why this is not consistent with the data we see, or why the model doesn't account for this possibility (as the model structure assumes a 100% carriage prevalence it can't formulate this possibility).

3. Following from 2. I am confused about the way the different models / hypotheses are tested. The outcome is the number of infections per strain type across the total population as a function of the parameter governing the mechanism. Is this the best way to test the different models? As far as i understand, for each hypothesis, the size of the parameter is varied between 0 and 50% increase away from the null model (i.e. increasing the virulence from 0 to 50% within the "Virulence" model leads to associated increasing the total numbers of all infections). However, is the dialling up and down of the different parameters what would happen when we observe changing numbers of infections? I don't think so. In reality, there will be a fixed parameter value governing a particular mechanism (e.g. a relative virulence of res vs sens strains with a parameter of 10% will give rise to the number of infections we see). The pertinent question is then - under a mechanism - if we implement interventions that, say, reduce antibiotic use, would any resulting reductions in the number of resistant infections be replaced by a compensatory number of sensitive infections. I think to properly address the issue, Fig 3 would show the number of infections of different types as, e.g., antibiotic use decreases, rather than changing a parameter (e.g. pi) that would in reality be fixed.

4. Finally, following on from 3. I found the results unsurprising given the set up of the model. As far as I understand, the total number of infections is determined exclusively by the virulence rate of each of the demographic groups. None of the models alter the number of people in each demographic compartment, and none alter the virulence rate directly (with the exception of the "Virulence" model). Therefore, there is no mechanism through which any of the models (except for Virulence) could change the total number of infections. I found this result a bit limiting. If the authors had comprehensive conceptual diagrams of how the different models (perhaps as suggested below in the specific comments) I think this main limitation would be clearer.

Some specific suggestions on these points and other comments are given below:

1. The introduction contains many paragraphs of description and justification of the approach. My suggestion would be to give a brief summary of the overall goal of the paper and the method to tackle the goal. The rest can be moved to the Results section by way of a motivation for the Results as the Methods are presented later.

2. 'Open population' - might be better described consistently as the 'community' or 'community population'? (At least for the differential equations it would be nice to see dc/dt rather than do/dt.)

3. I appreciated the description of the models as adjustments away from the neutral 'null' model. I found this a particularly helpful exposition.

4. I found the description of the model particularly confusing as a result of the equations and a repeated but less detailed compartment diagram being in the Supp Mat, with only a text description in the main ms and a simple compartment diagram. My suggestion would be two-fold: first, to add a much more detailed model compartment diagram to the main text; for example, in Colijn et al. (JRSI 2010) the authors use different line types to differentiate between (super)colonisation, clearance, and treatment; further, to create full diagrams that show the different demographic groups and infection compartments, as well as the different model structures (trying to understand Table S3 was a bit of a headache for me). Second, to reformat the equations so they are easier to read. Can anyone in the group perhaps write them in latex? Or at least, not use all the '*' notations and terms like 'sizeo', which makes the terms hard to distinguish and a headache to read.

5. Although parameters are defined in Table S4, there seem to be lots of notation that is not defined. For example, do the equations track the fraction of number of infections? What is 'y'. I couldn't properly assess the model because it was all rather difficult to piece together.

6. Some of the model description I find a bit confusing. For example, I don't know what is meant by the following sentences:

a) "In the SR-compartment, the infection rate is the average of the S- and R-compartment, with infections with resistant or susceptible strains being proportional to the two infection rates."

b) "Without these compartments, introduction of ARB into a population in which everyone could acquire ARB would lead to an equilibrium prevalence of 50%, even in absence of selective advantages." Would this be the single dual infection model (ie X, S, R, SR) as described in Colijn et al. JRSI 2010?

c) "Mechanisms benefitting resistant strains and increasing ARB colonisation prevalence render less people susceptible for acquiring new colonisation with susceptible strains, since we assume a maximum colonisation capacity (RR, SR or SS)"

7. Table S4: Should 'infection rate hospital' just be 'infection rate' ? There are no units given for this - or any of the other parameters.

8. The differential equations don't seem to be complete. For example, I can't find where any equation where the rate of infection, \\pi, which depends on the group (hosp, inpatient or recently hosp), is used. It would be helpful to show compartments called 'infections' (Sens / Res) that show how the number of infections are tracked, especially as this is the main outcome of the paper.

9. "Naturally, introducing costs of resistance without benefits of resistance reduces the incidence of infections caused by ARB and eventually leads to extinction". If you are talking about extinction here, I would suggest rephrasing to carriage (rather than infection), and rename as elimination of resistant strains.

Reviewer #3: Please see the attachment.

**********

Have the authors made all data and (if applicable) computational code underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data and code underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data and code should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data or code —e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: Yes: Katherine Atkins

Reviewer #3: No

Figure Files:

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, . PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at .

Data Requirements:

Please note that, as a condition of publication, PLOS' data policy requires that you make available all data used to draw the conclusions outlined in your manuscript. Data must be deposited in an appropriate repository, included within the body of the manuscript, or uploaded as supporting information. This includes all numerical values that were used to generate graphs, histograms etc.. For an example in PLOS Biology see here: http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001908#s5.

Reproducibility:

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Submitted filename: Reviewer Comments on Manuscript titled.pdf

Click here for additional data file.

18 Oct 2021

Submitted filename: Response Godijk et al.docx

Click here for additional data file.

15 Dec 2021

Dear Mrs. Godijk,

Thank you very much for submitting your manuscript "Does plasmid-based beta-lactam resistance increase E. coli infections: Modelling addition and replacement mechanisms" for consideration at PLOS Computational Biology. As with all papers reviewed by the journal, your manuscript was reviewed by members of the editorial board and by several independent reviewers. The reviewers appreciated the attention to an important topic. Based on the reviews, we are likely to accept this manuscript for publication, providing that you modify the manuscript according to the review recommendations.

Please prepare and submit your revised manuscript within 30 days. If you anticipate any delay, please let us know the expected resubmission date by replying to this email.

When you are ready to resubmit, please upload the following:

[1] A letter containing a detailed list of your responses to all review comments, and a description of the changes you have made in the manuscript. Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out

[2] Two versions of the revised manuscript: one with either highlights or tracked changes denoting where the text has been changed; the other a clean version (uploaded as the manuscript file).

Important additional instructions are given below your reviewer comments.

Thank you again for your submission to our journal. We hope that our editorial process has been constructive so far, and we welcome your feedback at any time. Please don't hesitate to contact us if you have any questions or comments.

Sincerely,

Roger Dimitri Kouyos

Associate Editor

PLOS Computational Biology

Virginia Pitzer

Deputy Editor-in-Chief

PLOS Computational Biology

***********************

A link appears below if there are any accompanying review attachments. If you believe any reviews to be missing, please contact ploscompbiol@plos.org immediately:

[LINK]

Reviewer's Responses to Questions

Comments to the Authors:

Please note here if the review is uploaded as an attachment.

Reviewer #1: The authors thoroughly addressed the points suggested by the reviewers, nice work. I will just mention one point that I still stumbled upon while reading:

(LL.290-305): In lines 297-299 it is described, that the compartments SS and RR are added to ensure a neutral model, but it is stated that those compartments do not have an obvious meaning. In lines 310-311 it is stated that those compartments are interpreted as high-density compartments. Isn't therefore a biological meaning assigned to the compartments?( The implicit assumptions of within-host state of disturbed and stable flora).I think this paragraph might benefit from a clearer formulation to avoid confusion. Furthermore I think it could be made clearer that the double compartments (SS,SR,RR) are seen as the default colonisation of a "healthy" human, which only changes to S and R due to disturbances. (Maybe starting out with explaining the double compartments and then the single colonised ones? But this is probably just a matter of taste)

(Typo in L 198?: If ARB would not have increaseD?)

Reviewer #2: The authors have responded to my comments very well - thank you for the attention to detail. I especially compliment them on rewriting the models equations and drawing a model diagram. This has made the paper much more user-friendly.

My only remaining comment is really the way that the authors responded to my second point. That is, I was suspicious about the method that tested the hypothesis of replacement of resistant infections. I was arguing that the test of whether resistant strains are in addition or instead of sensitive strains should really be in the context of when the system is perturbed - most notably in the case of introducing an intervention (rather than on a change of parameters that in reality, remain fixed). The authors have included a Table (S9) that assess the impact of these interventions on the number of resistant and sensitive infections with increasing antibiotic control.

Perhaps I am missing something here, but the ‘base case’ scenario - i.e. the first column with ‘full antibiotic use’ indicates different numbers of infections for each mechanism. I don’t believe this is correct way to compare the impact of antibiotic control in the presence of each of the mechanisms. All mechanisms should start with the same base case number of infections and any intervention should therefore be easily compared to each other on the basis of deviation away from this base case.

I would also suggest that this is a figure for ease of understanding - but this is up to the authors’ discretion.

Finally, there is a Typo on line 173.

Reviewer #3: The authors were able to clarify most of the concerns raised before. I still have reservations about how the infection rate being estimated in the manuscript, but I can understand that estimating the infection rate is not an easy task and could make the revision impossible. So I do not have further questions about this manuscript.

**********

Have the authors made all data and (if applicable) computational code underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data and code underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data and code should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data or code —e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: Yes: Katherine Atkins

Reviewer #3: No

Figure Files:

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org.

Data Requirements:

Please note that, as a condition of publication, PLOS' data policy requires that you make available all data used to draw the conclusions outlined in your manuscript. Data must be deposited in an appropriate repository, included within the body of the manuscript, or uploaded as supporting information. This includes all numerical values that were used to generate graphs, histograms etc.. For an example in PLOS Biology see here: http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001908#s5.

Reproducibility:

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

References:

Review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript.

If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

10 Jan 2022

Submitted filename: Reply Godijk et al 10 01 2022.docx

Click here for additional data file.

27 Jan 2022

Dear Mrs. Godijk,

We are pleased to inform you that your manuscript 'Does plasmid-based beta-lactam resistance increase E. coli infections: Modelling addition and replacement mechanisms' has been provisionally accepted for publication in PLOS Computational Biology.

Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. A member of our team will be in touch with a set of requests.

Please note that your manuscript will not be scheduled for publication until you have made the required changes, so a swift response is appreciated.

IMPORTANT: The editorial review process is now complete. PLOS will only permit corrections to spelling, formatting or significant scientific errors from this point onwards. Requests for major changes, or any which affect the scientific understanding of your work, will cause delays to the publication date of your manuscript.

Should you, your institution's press office or the journal office choose to press release your paper, you will automatically be opted out of early publication. We ask that you notify us now if you or your institution is planning to press release the article. All press must be co-ordinated with PLOS.

Thank you again for supporting Open Access publishing; we are looking forward to publishing your work in PLOS Computational Biology.

Best regards,

Roger Dimitri Kouyos

Associate Editor

PLOS Computational Biology

Virginia Pitzer

Deputy Editor-in-Chief

PLOS Computational Biology

***********************************************************

Reviewer's Responses to Questions

Comments to the Authors:

Please note here if the review is uploaded as an attachment.

Reviewer #1: The revised version made the model interpretation more accessible for the reader in my opinion. Thanks for clarifying. I have no further comments.

Reviewer #2: I have no more comments. Thank you for the interesting paper.

**********

Have the authors made all data and (if applicable) computational code underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data and code underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data and code should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data or code —e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: Yes: Katherine Atkins

21 Feb 2022

PCOMPBIOL-D-21-00975R2

Does plasmid-based beta-lactam resistance increase E. coli infections: Modelling addition and replacement mechanisms

Dear Dr Godijk,

I am pleased to inform you that your manuscript has been formally accepted for publication in PLOS Computational Biology. Your manuscript is now with our production department and you will be notified of the publication date in due course.

The corresponding author will soon be receiving a typeset proof for review, to ensure errors have not been introduced during production. Please review the PDF proof of your manuscript carefully, as this is the last chance to correct any errors. Please note that major changes, or those which affect the scientific understanding of the work, will likely cause delays to the publication date of your manuscript.

Soon after your final files are uploaded, unless you have opted out, the early version of your manuscript will be published online. The date of the early version will be your article's publication date. The final article will be published to the same URL, and all versions of the paper will be accessible to readers.

Thank you again for supporting PLOS Computational Biology and open-access publishing. We are looking forward to publishing your work!

With kind regards,

Katalin Szabo

PLOS Computational Biology | Carlyle House, Carlyle Road, Cambridge CB4 3DN | United Kingdom ploscompbiol@plos.org | Phone +44 (0) 1223-442824 | ploscompbiol.org | @PLOSCompBiol