The transcriptional gradient in negative-strand RNA viruses suggests a common RNA transcription mechanism
Authors: Connor R. King, Casey-Tyler Berezin, Brian Munsky, Jean Peccoud, James R Faeder, Oleg A Igoshin, James R Faeder, Eric C. Dykeman, James R Faeder, Eric C. Dykeman, James R Faeder, Eric C. Dykeman
Published: 2026

ABSTRACT
Nonsegmented negative-strand RNA viruses (NNSV) are a diverse class of medically relevant viruses which display a conserved attenuation gradient in the transcription of their genomes. This gradient has been traditionally explained by the Stop-Start model which attributes attenuation to polymerase behavior at gene junctions. In this article, we evaluate an alternative explanation where the gradient arises from polymerase dynamics during transcription. We introduce the RNA Polymerase Association Mechanism (RAM) model, a coarse-grained stochastic framework that describes transcription using two parameters related to polymerase processivity and the ability of the polymerase to backtrack. The RAM model accurately reproduces transcriptional gradients across diverse NNSVs as well as in gene-shuffled VSV variants. Additionally, the inferred polymerase processivity appears correlated to the length of the viral genomes suggesting a conserved constraint on transcription across these viruses. While the RAM model does not account for all known molecular features of NNSV transcription, it provides a parsimonious and predictive framework for relating genome architecture and transcription. These results support the view that, in tandem with the traditional junction-centric mechanisms governing transcription, nonspecific attenuation mechanisms contribute to the NNSV transcriptional gradient and warrant closer inspection in future studies which could lead to better rational genome design in viral studies and biomedical applications.

Introduction

Nonsegmented negative-strand RNA viruses (NNSVs), including Ebola (EBOV), rabies virus, and vesicular stomatitis virus (VSV), are a diverse class of viruses that are commonly used in biomedical applications such as vaccine vectors [1]. These viruses possess a single RNA strand encoding tightly packed genes (Fig 1) and their transcriptional output is linked to gene order. In the 1970s, UV inactivation studies in VSV suggested a single polymerase initiates at the 3′ end and transcribes each gene sequentially [2–4]. Although early data showed similar protein levels [5], differences in protein size implied non-equimolar expression. This difference in relative expression was characterized at both the protein and transcript level to reveal a 3’ to 5’ gradient in expression, suggesting that the control was at the transcription level [6]. This gradient led researchers to suggest two hypotheses.

Genome Architecture of NNSVs.

Graphical representation of the genomes of nonsegmented negative-strand RNA viruses (NNSVs) in 3’ to 5’ orientation with gene lengths to scale. Yellow stars indicate overlapping genes in the genome. The genes that lack labels are the regions containing the small hydrophobic (SH) gene. Common proteins present in most of these species are the N/NP = nucleocapsid, P = phosphoprotein, M = matrix protein, G/GP = glycoprotein, L = RNA-dependent RNA polymerase, F = fusion protein, and HN = hemagglutinin-neuraminidase protein.

Iverson and Rose rationalized that since the transcripts of VSV had been shown to have roughly equal degradation rates [7], the difference in relative expression was due to actions in the genome either at specific locations (gene junctions) or nonspecifically [8]. They designed probes to quantify the relative amounts of the 3’ and 5’ end of transcripts as they hypothesized that nonspecific attenuation would result in a gradient at the transcript level while specific attenuation would show equimolar levels of these fragments. This paper also established that at most gene junctions there is a 29–33% drop while the junction before the L polymerase causes a 80–90% decrease in transcription. This hypothesis was soon dubbed the “Stop-Start Hypothesis” [9] and was extended to many other NNSVs, including EBOV [10,11], Marburg virus (MV) [10,12], parainfluenza virus type 2 (PIV2), parainfluenza virus type 3 (PIV3) [13,14], parainfluenza virus type 5 (PIV5) [13,15], mumps virus (MuV) [13,16], measles virus (MeV) [17], and respiratory syncytial virus (RSV) [18,19]. While this was a hot topic for many years, the debate over this phenomenon has quieted and to this day, the Stop-Start Hypothesis remains the dominant hypothesis [20].

Much of the literature has focused on dissecting what sequences at these gene junctions control the level of attenuation observed. Particular focus has been interested in the gene junction prior to the L gene as in general, we tend to see consistent levels of attenuation for all junctions at around 30% except for the L gene which exhibits about 80–90% attenuation. The cause of this has remained elusive, as in some viruses this attenuation can be strongly attributed to gene junctions while others it cannot. Indeed, in rabies virus we can see that the level of attenuation is strongly tied to the intergenic sequence located between the stop and start signal sequences and this has been used to generate variants of rabies virus that overexpress the L polymerase [21]. Additionally, RSV also has been shown to have termination depend strongly on the gene junction present, but intriguingly there is no effect from the very structurally diverse intergenic regions [18,22].

On the other hand, some viruses like EBOV do not offer such a clean explanation. For EBOV, attenuation at the L junction is unremarkable, while we do see an effect on the NP/VP35 and VP30/VP24 junctions [23]. Although, it should be noted that the inclusion of the 5’ and 3’ UTRs might have influenced these experiments as they may impact the stability of transcripts. Another intriguing study using PIV5 looked at attenuation from different gene junctions in the context of expressing HN and L. Three different levels of attenuation were observed across the 6 genomes with the M-F, F-SH, and SH-HN junctions producing higher attenuation than the HN-L junction. Even more perplexing, these variations in attenuation rates were not observed when the genes in the minigenome were swapped for F and V/P [15].

These works suggest that there are several mechanisms that come together to orchestrate the gradient observed in literature. Indeed, some of these mechanisms that have come into focus have more to do with intrinsic polymerase properties and could be viewed as a source of nonspecific attenuation. The efficiency of 5’ capping has gained more focus in recent years where it has been shown that the formation of the 5’ cap during transcription is necessary for the proper elongation and termination of transcription [17,24–26]. When this process fails, it results in the production of truncated transcripts. Indeed such cap-less truncated transcripts had even been noted in RSV a decade earlier [27]. Something intriguing here is the observation of truncated transcripts, the absence of which was central to the initial formation of the stop-start model.

As such, there may be value in revisiting the study of NNSV transcription from the lens of nonspecific attenuation. Indeed, it has been shown that in measles virus the slope of the transcriptional gradient can be altered based on the affinity between the X domain of the P protein (which is part of the transcription complex) and the C terminal tail of the Nucleoprotein [28]. Revisiting these proposed mechanisms is also warranted as the template dissociation-based mechanism appears to be at odds with the structure of the polymerase-template complex. The L polymerase completely encircles the genome while moving down the genome [29]. So, the polymerase would need to undergo a dramatic rearrangement that is highly unlikely in order to dissociate from the template. Moreover, junction sequences are highly diverse, even within the same virus (Figs 2A, S1 Fig.). The initiation sequence itself is poorly conserved, with the minimal required motif being 3’-UYGnnnnnnn-5’ [30]. This raises the question of how this consistent gradient that is present across all NNSVs can be the direct result of such a diverse array of sequences.

Gene Junction Diversity of NNSVs.

A. Alignments of gene junctions for 4 NNSVs. Each junction has blue and red boxes highlighting the conserved transcriptional start and stop sequences respectively. Black stars to the left of the sequence indicate that the junction is an overlapping junction and so the locations of the start and stop sequences are inverted compared to the others. B. Cartoon diagram of the types of gene junctions observed in NNSVs and overlapping gene junctions. The UTRs in this figure refer to the untranslated regions of the transcripts that are produced.

In this article, we develop a model of transcription that attributes the amount of attenuation at each junction due to the distance the polymerase must travel from the 3’ end of the genome to the gene’s termination signal. First, we demonstrate that the mechanism is consistent with a wide variety of NNSV transcriptional gradient datasets. Then we show that the resulting predictions from a model fit to VSV transcriptional data is as consistent with the observed protein expression gradients in gene shuffled VSV variants as the Stop-Start model despite having fewer parameters (1 instead of 4). Additionally, there is even some weak evidence that the RAM model produces slightly better predictions than the Stop-Start model when considering the expression of genes in gene-shuffled VSV variants. Given that it is unlikely that the polymerase dissociates from the genome template [29], it is unclear why moving down the genome results in fewer transcripts being produced. It is possible that “losing” the nascent transcript results in a polymerase state that is unable to continue transcription until it restarts all over again at the 3’ entry site. Indeed, there are extensive regulatory interactions that have been characterized between the nascent RNA and the RNA polymerase [31]. Since we attribute the amount of attenuation at each junction to the polymerase’s ability to remain associated with the nascent transcript during transcription, we name this model the RNA Polymerase Association Mechanism (RAM) model.

Results

Modeling NNSV transcription as a walk

The RAM model consists of a single polymerase performing a walk beginning at the 3’ end of the genome (Fig 3). In each step, the polymerase takes one of two actions. The polymerase either takes a single base step forward and continues productive transcription or it dissociates from the transcript and enters a nonproductive state where it can no longer initiate transcription. The probabilities associated with these actions will be referred to as pmaintain and pdrop, respectively. Since pdrop is the complement of pmaintain, these represent a single parameter in this model. The relative level of expression of most genes (pexpress) can be calculated from the distance of the stop signal for the gene from the 3’ end of the genome (n) using the equation:

Diagram of Stop-Start and RAM Models.

(A.) Stop-Start Model and (B.) RAM Model. In both models transcription begins at the 3’ most end of the genome (state 1) where the polymerase recognizes the start signal and initiates transcription. The polymerase then proceeds down the genome to the termination signal (state 2) and releases the synthesized transcript. For standard gene structures, the polymerase reinitiates at the next downstream start signal (state 1). If there is an overlap, the polymerase must move backwards to state 4 before proceeding to state 1. The models differ in where the attenuation we observe comes from. In the RAM model, the polymerase slips and releases the nascent transcript, entering a state that can no longer recognize transcription start signals (state 3). In the Stop-Start model, the polymerase dissociates from the genome at the junction (state 3). Although this too may instead be due to some sort of conformational change into an inactive polymerase conformation at these junctions. These two models are not necessarily mutually exclusive. Both processes are likely to be occurring. Created in BioRender. King, C. (2026) https://BioRender.com/z9eob9v.

Where n is the number of bases that the termination sequence is from the 3’ end of the genome. It is important to note that some NNSVs have overlapping genes, where the transcriptional start signal of one gene is located within the upstream gene (Fig 2). In cases where overlapping occurs and the transcription start site of a gene (Gene 2) is located within an upstream gene (Gene 1), then the transcription of Gene 2 only occurs if the polymerase backtracks back to the start signal of Gene 2 after finishing transcription of Gene 1 [20,32]. We characterize this using one additional parameter called pbacktrack. In these cases pexpress is calculated using this formula:

The parameters pmaintain/pdrop and pbacktrack were determined by the fitting the model to maximize the likelihood of the data given the model under the assumption that their values were between 0 and 1, and posterior uncertainties in parameters were quantified using Markov Chain Monte Carlo (MCMC) algorithm (S2 Table) [33,34]. The data used in this article are very diverse, including sequencing data from Illumina sequencing and direct RNA sequencing as well as blotting-based methodologies. The code for fitting the models can be found in Notebook1 on the GitHub associated with this article.

The RAM model fits the transcriptional gradient of diverse NNSVs

The RAM model was fit to the observed gradients from 9 NNSVs of the Pneumoviridae, Paramyxoviridae, Filoviridae, and Rhabdoviridae families (Fig 4, S2 Table). For Rhabdoviridae, the model was fit to the original dataset used to describe the ratio of VSV expression in 1981 [8]. For Filoviridae, EBOV and MV were used [35]. For Pneumoviridae, RSV was used [36]. For Paramyxoviridae, data from MeV, MuV, PIV2, PIV3, and PIV5 were used [13,37]. It is important to note that the published data associated with the transcriptional behavior of RSV are inconsistent. Some articles propose an entirely different gradient behavior of RSV with several inconsistencies when compared to the characteristic simple gradient [32,38]. In this article, the RSV gradient that most closely resembled the characteristic NNSV gradient was selected, as this is the one that is most consistent with the current understanding of NNSV transcription in literature.

The RAM Model Accurately Predicts Transcriptional Gradients Across NNSV Species.

Model predictions of the transcriptional gradient of NNSVs using the RAM model (red bars) compared to published data (blue bars). These predictions are across predominant families of NNSVs: Rhabdoviridae (A), Filoviridae (B and C), and Paramyxoviridae (D-F, H, I), and Pneumoviridae (G). Black error bars are 4 standard deviations in length, or were excluded if standard deviations were not available for the data. The log likelihood is included for every model.

The RAM model was able to achieve high-quality fits for all these viruses (Fig 4). The confidence in parameter pmaintain associated with each virus are well-determined (S2 Table). In general, the pbacktrack parameter is well-determined in the viruses that have overlaps in their genomes while it appears underdetermined in those that do not (S2 Table). This is not surprising as models of viruses without overlaps never actually utilize this parameter. As such, in these scenarios, this model can be reduced to a single parameter. In viruses that have overlapping genes, in order for the downstream gene of an overlapping junction to be transcribed the polymerase must backtrack to the transcriptional start signal upon completion of transcribing the upstream gene. This should lead to additional attenuation that comes from the parameter pbacktrack in these viruses. However, the junctions in each virus appears to behave substantially different from one another with RSV demonstrating the largest amount of additional attenuation (94.15-94.33% more) attributed to the overlap while MV and EBOV had very little if any attenuation (1.6-20.05% more and 0-.065% more respectively. In RSV, it has been shown that part of this additional attenuation is due to premature stoppage of transcription at the internal stop signal [39]. It is possible that since the overlap in MV and EBOV are less dramatic than in RSV (Figs 2A, S1 Fig.) this effect does not occur in these viruses. Alternatively, it could be due to the larger distance the polymerase must travel.

Genome size is correlated to polymerase processivity

Since pdrop is the probability of the polymerase releasing the nascent transcript at any point, it can be viewed as a type of polymerase processivity. Intriguingly, there appears to be a negative correlation between the value of pdrop and the length of the genome (Fig 5). This correlation suggests that for every base increase in the genome, the probability of the polymerase releasing the transcript at every step down the genome decreases by about 1.79 x 10-8 ± 1.72 x 10-8 (p-value = .044, t-test statistic = 2.45). This observation is reasonable as all polymerases must enter the 3’ end of the genome and so the polymerase must stay active for longer stretches in order for genes at the 5’ end of longer viral genomes to be expressed.

The Processivity of Viral Polymerases is Correlated to Genome Length.

This figure shows a linear model showing how the probability of the polymerase releasing the nascent transcript at each step (Pdrop) decreases as the length of the genome increases. The estimated slope (-1.81 x 10-8 ± 1.70 x 10-8) and intercept (5.25 x 10-4 ± 2.72 x 10-4) are both statistically significant (t(7) = -2.50, p = .041 and t(7) = 4.58, p = .003 respectively). The dots and error bars are the estimates and 95% confidence intervals respectively for the value of pdrop for each virus.

The RAM model accurately predicts gene expression in reorganized genomes

Historically, VSV has been studied using “gene-shuffled” variants where the position of genes is shifted to leverage the transcriptional gradient to predictably alter transcription rates [40,41]. Here, the RAM model and the Stop-Start model were both evaluated on their ability to predict how shuffling the position of VSV’s genes alters the observed expression ratio of the genes [41,42]. The Stop-Start model used here is model with 4 parameters, one for each junction. The amount of attenuation expected at each junction is essentially the amount of attenuation observed in the training dataset. The available data is all at the protein level, however, it is generally understood that the gradient in the expression of VSV genes at the transcript level also occurs at the protein level, with some potential to deviate due to other mechanisms that may be controlling protein expression [40–42].

Both models were first fit to the Iverson and Rose VSV transcriptional gradient data [8] using MCMC. Then, to evaluate the models parameters were drawn from the posterior distributions. Each of these draws was used to simulate the models and the average of these predictions was used to calculate the Mean Absolute Error between the predictions and observed expression levels for gene-shuffled variants from two datasets [41,42]. Both models do well predicting the relative expression levels (Fig 6A and B) and show a statistically significant correlation between the predicted expression levels. This is consistent with the previous statement that the expression gradient is carried on to the protein level and may occur at a nearly 1:1 level. However, there do appear to be some deviations between predicted and observed expression levels. It is possible that this is due to differing environmental factors, such as the degradation rate [43,44].

The RAM Model Accurately Predicts Gradients Across Gene-Shuffled VSV Variants.

This figure shows the linear fits for the predicted relative mRNA transcript levels in different VSV gene-shuffled variants for the (A.) RAM Model and (B.) Stop-Start Model (y-axis) compared to the observed protein expression levels of experimentally constructed gene-shuffled VSV variants from 2 previous publications, Ball et al. 10.1128/jvi.73.6.4705-4712.1999 and Flanagan et al. 10.1128/jvi.74.17.7895-7902.2000 (x-axis). The slopes of both the RAM model (0.848 ± 0.152) and the Stop-Start model (0.872 ± 0.161) are statistically significant (t(48) = 11.25, p = 4.67 x 10-15 and t(48) = 10.89, p = 1.46 x 10-14). C. Shows the distribution of absolute errors between predicted and observed values between both models along with the Mean Absolute Error. The differences between these errors is not statistically significant (t(49) = -1.74, p = 0.087).

When looking at the error in these two models, while it is not statistically significant there is a very weak benefit from using the RAM model which has an MAE of 0.0501 as opposed to the Stop-Start model which has an MAE of 0.0545 (Fig 6C). It is likely that this is since the Stop-Start model is much more susceptible to overfitting since the number of variables is equal to the number of gene junctions (4) while the RAM model relies on only 1 parameter.

Discussion

The model proposed here provides a simplified, quantitative approximation of the transcriptional gradient in NNSVs without explicitly incorporating junction-specific regulation. The RAM model can recapitulate the transcriptional gradient in a diverse array of viruses using only one (and at most two) parameters (Fig 4) providing a more parsimonious model as opposed to the Stop-Start model which has one parameter for each gene junction. This simplicity lends the model to be less susceptible to overfitting, the effects of which can be seen in the minor advantage the model has over the Stop-Start model when predicting the relative expression of genes in gene-shuffled VSV variants (Fig 6C).

The RAM model’s robustness across diverse datasets ranging from traditional methods that use 3H-uridine to quantify viral RNAs [8] to next-generation sequencing data [35] further supports its broad applicability. The model presented in this article is effectively a successor to the non-specific attenuation model that had been previously discarded in the early 1980s [8]. By showing that a simple polymerase-based stochastic process can reproduce the transcriptional gradient across multiple NNSVs, our results suggest that these observed polymerase dynamics may represent a conserved underlying constraint in viral transcription. A result supported by the correlation between polymerase processivity and genome length (Fig 5).

Despite the benefits of the RAM model, there are limitations. The RAM model inherently does not capture all aspects of NNSV transcription. There are many cases where the gene expression gradient is strongly impacted by the gene junctions, such as rabies virus [21] and RSV [18,22]. These observations are more consistent with the Stop-Start model and highlight why the Stop-Start model is still a central model in NNSV research. Rather than replacing the Stop-Start model, the RAM model represents a complementary explanation that captures baseline attenuation and encourages the investigation of nonspecific mechanisms also known to be at work, such as premature termination due to failed 5’ capping [17,24–26].

However, open questions remain. One of the major implications of the RAM model is the generation of truncated transcripts. Such transcripts have a complicated history in the NNSV transcriptions space. On one hand, Iverson and Rose used the lack of observed truncated transcripts as evidence for refuting a nonspecific attenuation mechanism [8]. On the other, truncated transcript have been observed in RSV [27] and when studying the role of 5’ capping in [17,24–26]. It is unclear if these discrepancies are due to biological differences in the viruses or methodological differences in these studies. It also should be noted that in most studies the transcript levels observed are the steady-state levels. So, in some cases the absence of truncated transcripts could be due to reduced stability from a lack of 5’ caps and/or polyadenylated tails [13,45]. Another alternative interpretation is that this is an artifact of the assumption that there is a uniform probability of attenuation across the genome. Rather, the dissociation of the polymerase from the transcript may be limited to specific stages in transcription. For example, it may be that the phenomenon associated with this is in fact the aforementioned 5’ capping-dependent transcriptional termination mechanism. However, why the amount of attenuation would be correlated with the length of the gene is unclear.

To resolve these discrepancies, we advocate moving beyond minigenome assays toward large-scale variant libraries. While this was extremely challenging during these initial studies, it is now possible to leverage homology-based methods to generate large panels of variants [46,47]. Gene-shuffled and junction-swapped constructs, analyzed by qPCR or RNA-seq (in the absence of poly-A pulldowns), could reveal the individual contributions of gene order, junction sequences, and polymerase behavior. Such data would allow for a more rigorous understanding of NNSV transcription and help discern what mechanisms are at play. However, we should not neglect the fact that it is highly likely that both the specific and nonspecific mechanisms discussed in this article participate together in the regulation of transcription and therefore the emergence of a transcriptional gradient in NNSVs.

Overall, this work brings attention back to the potential for a per-base attenuation mechanism that may also be at work in the NNSV transcriptional gradient. At minimum, this model provides a useful null control for the assumed gradient for any NNSV with an uncharacterized gradient. At best, this brings back the idea that there should be some revisitation to a per-base nonspecific attenuation mechanism that was discarded very early on in the study of NNSV transcription. Such per-base attenuation could be exceedingly useful in the biomedical community as it may allow for improved calibration of viral transcriptional profiles in therapies. For example, it has been a major point of focus to understand how the alteration of transcription rates influences the fitness of the VSV, primarily to attenuate it for medical applications [41,44,48]. Our model suggests new design strategies beyond gene shuffling (namely, gene overlaps and gene length modulation). Considering overlapping genes as binary variables yields 1,920 possible VSV variants. Incorporating gene-length changes, which might influence transcription rates and are feasible due to VSV’s tolerance for large genome insertions [49–51], dramatically expands the design space. Together, gene order, overlap, and length could provide unprecedented control over viral transcription. This would enable not just understanding of viral biology but also the rational design of attenuated or customized NNSV variants with finely tuned expression profiles.

Materials and methods

Identification of gene junctions

Gene junctions were identified in reference genomes by using literature reported gene junctions as a guide (S1 Table) [15,22,23,37,52–56]. Junctions were then aligned in SnapGene using MUSCLE. Reference genomes, junctions, and alignments can be found in Folder4 on GitHub.

Parameter estimation

To estimate parameters, the log likelihood for observing the data based on the model-predicted ratios was maximized. The minimize function from SciPy was utilized. Each model was fit 10,000 times starting at a value between 0 and 1. The parameters chosen for downstream analyses were the ones associated with the largest log likelihood. Confidence in parameters was assessed using MCMC algorithm [33,34]. The code for fitting models and performing MCMC can be found in Notebook1 on GitHub. The code for then making predictions for gene shuffled variants can be found in Notebook2.

Linear regression

Linear regression was performed in R using the base lm function. The performance package was used to test model assumptions of normality, homogeneity of variance, and linearity. No violations were observed. The code for fitting the linear regression model connecting polymerase processivity and genome length can be found in the Markdown1 file on GitHub. The code for fitting the linear regression model connecting RAM model predicted expression and observed expression for gene shuffled variants can be found in the Markdown2 file on GitHub.

Supporting information

Alignments of Gene Junctions for 9 NNSVs.

Each junction has blue and red boxes highlighting the conserved transcriptional start and stop sequences respectively. Black stars to the left of the sequence indicate that the junction is an overlapping junction and so the locations of the start and stop sequences are inverted compared to the others.

(TIFF)

Literature Reported Gene Junction Consensus Sequences.

(PDF)

Best Fit Parameter Along with Confidence Statistics from MCMC for RAM Model.

(PDF)

GenBank Accession Numbers for Nonsegmented Negative-Stranded RNA Virus Genomes.

(PDF)

19 Sep 2025

Submitted filename: PLOS_Responses_v1.pdf

18 Nov 2025

-->PCOMPBIOL-D-25-01912

The Transcriptional Gradient in Negative-Strand RNA Viruses Suggests a Common RNA Transcription Mechanism

PLOS Computational Biology

Dear Dr. Peccoud,

Thank you for submitting your manuscript to PLOS Computational Biology. After careful consideration, we feel that it has merit but does not fully meet PLOS Computational Biology's publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Jan 18 2026 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at ploscompbiol@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pcompbiol/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

* A rebuttal letter that responds to each point raised by the editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. This file does not need to include responses to formatting updates and technical items listed in the 'Journal Requirements' section below.

* A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

* An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, competing interests statement, or data availability statement, please make these updates within the submission form at the time of resubmission. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter

We look forward to receiving your revised manuscript.

Kind regards,

Oleg A Igoshin

Academic Editor

PLOS Computational Biology

James Faeder

Section Editor

PLOS Computational Biology

Additional Editor Comments :

In order to give the manuscript a proper consideration, we have decided to invite two reviewers who were not associated with the original decision but nevertheless top experts in the field. There are still concerns voiced about authors understanding or misinterpretations of the literature, taking the previously published data into consideration in developing thir model. However, in comparision to the previous version, the manuscript is significantly improved and we are happy to allow the authors to resubmit after major revision.

Journal Requirements:

1) Please ensure that the CRediT author contributions listed for every co-author are completed accurately and in full.

At this stage, the following Authors/Authors require contributions: Jean Peccoud. Please ensure that the full contributions of each author are acknowledged in the "Add/Edit/Remove Authors" section of our submission form.

The list of CRediT author contributions may be found here: https://journals.plos.org/ploscompbiol/s/authorship#loc-author-contributions

2) We ask that a manuscript source file is provided at Revision. Please upload your manuscript file as a .doc, .docx, .rtf or .tex. If you are providing a .tex file, please upload it under the item type u2018LaTeX Source Fileu2019 and leave your .pdf version as the item type u2018Manuscriptu2019.

3) Please upload all main figures as separate Figure files in .tif or .eps format. For more information about how to convert and format your figure files please see our guidelines:

https://journals.plos.org/ploscompbiol/s/figures

4) Some material included in your submission may be copyrighted. According to PLOSu2019s copyright policy, authors who use figures or other material (e.g., graphics, clipart, maps) from another author or copyright holder must demonstrate or obtain permission to publish this material under the Creative Commons Attribution 4.0 International (CC BY 4.0) License used by PLOS journals. Please closely review the details of PLOSu2019s copyright requirements here: PLOS Licenses and Copyright. If you need to request permissions from a copyright holder, you may use PLOS's Copyright Content Permission form.

Please respond directly to this email and provide any known details concerning your material's license terms and permissions required for reuse, even if you have not yet obtained copyright permissions or are unsure of your material's copyright compatibility. Once you have responded and addressed all other outstanding technical requirements, you may resubmit your manuscript within Editorial Manager.

Potential Copyright Issues:

i) Figure 3. Please confirm whether you drew the images / clip-art within the figure panels by hand. If you did not draw the images, please provide (a) a link to the source of the images or icons and their license / terms of use; or (b) written permission from the copyright holder to publish the images or icons under our CC BY 4.0 license. Alternatively, you may replace the images with open source alternatives. See these open source resources you may use to replace images / clip-art:

- https://commons.wikimedia.org

- https://openclipart.org/.

5) Please amend your detailed Financial Disclosure statement. This is published with the article. It must therefore be completed in full sentences and contain the exact wording you wish to be published.

1) State what role the funders took in the study. If the funders had no role in your study, please state: "The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript."

6) Please revise your current Competing Interest statement to the standard "The authors have declared that no competing interests exist."

7) Please cite and label the supplementary tables in numerical order as “S1 Table” and “S2 Table,” and so forth.

Note: If the reviewer comments include a recommendation to cite specific previously published works, please review and evaluate these publications to determine whether they are relevant and should be cited. There is no requirement to cite these works unless the editor has indicated otherwise.

Reviewers' comments:

Reviewer's Responses to Questions

Reviewer #1: The authors took several of the reviewers’ comments into consideration and added enough caveats about their model to significantly improve the manuscript.

However, this reviewer is still concerned by some overinterpretations or misinterpretations of the literature. For instance, the authors state that “In RSV, similar assays yield conflicting results on whether intergenic regions drive attenuation (17, 20)”. But this is misleading since these two studies did not compare the same thing. In ref 17, the authors did switch the intergenic region, i.e., the non-transcribed part of the gene junction, but in ref 20, the authors compared the whole gene junction, i.e., the gene end (GE) + intergenic region (IR) + gene start (GS). Therefore, ref 17 and ref 20 do not show “conflicting” results but rather complementary results.

Another example is the interpretation of the results of ref 24. The data in Figure 6b are normalized by the ratio found of the N-P junction. Hence, the fact that the ratio is higher for the P-M junction does not mean that there is a “higher downstream-than-upstream expression at the P-M junction” but rather that the attenuation is lower between P and M than between N and P. Also, the “only ~60% attenuation before L” is not absolute but corresponds to 60 % stronger attenuation than the one at the N-P junction.

In the next sentence, the authors state that “The presence of overlapping genes in RSV, EBOV, and MV further challenges the Stop-Start model, which assumes unidirectional transcription”. Why would the Stop-Start model assume unidirectional transcription? In this model, once the polymerase has recognized a GE sequence, generated a polyA tail, and released the mRNA, it scans the genome to find another GS sequence. But why would the scanning be unidirectional? The review cited as a reference for the Start-Stop model (ref 19) clearly mentions that a polymerase can “scan backward to access a gs signal”.

Finally, to complete the discussion, it might be worth mentioning that the “gradient” measured in infected cells is also affected by the intrinsic half-life of the individual mRNAs. Indeed, since in infected cells we only measure the accumulated mRNAs at a given time point and not the transcription activity itself, if a downstream mRNA is less expressed but more stable than its upstream one, it may seem expressed at the same level, or even at a higher level. Inversely, if an mRNA is well expressed but relatively unstable, the upstream gene junction will look like a strong attenuation step. Thus, the gradient measured by sequencing or radiolabeling of accumulated mRNAs is also affected by the differences in mRNA half-lives. This parameter may explain some of the discrepancies observed between minigenome systems and viruses. Indeed, in the bicistronic minigenome assays where the gene junctions are switched, the two mRNA stay mostly identical (except for a few mutations in the GS and GE), while in viruses, each mRNA is vastly different from the others. Hence, the differences in mRNA half-lives likely impact the mRNA ratios in infected cells much more than the ratios in minigenome assays.

Reviewer #2: The non-segmented negative strand RNA viruses have a transcription gradient in which genes from the 3´ end of the genome are more abundant than those from the 5´ end. Early experiments by Iverson and Rose, 1981, using both VSV infected cells and an in vitro transcription assay showed that attenuation primarily occurred at the gene junctions (stop-start model). This early work has been largely supported in subsequent studies of the gene junction regions of non-segmented negative strand RNA viruses. In this resubmitted manuscript, King and coworkers present an alternative hypothesis, in which attenuation occurs at every nucleotide along the genome. They test this hypothesis by comparing their mathematical model with published data and it appears to fit.

It is entirely possible that there is some attenuation at each nucleotide and that the authors may be correct in their assertion that both their RAM model and the stop-start model models come into play. However, the authors minimize experimental data that demonstrates the key role that the gene junctions play and demonstrate misconceptions regarding published data. Without careful assessment of prior experimental data, and integration of these data into their model, it is difficult to assess how valid the RAM model is.

1. Lines 39-40, 57-59 and throughout manuscript: Throughout the manuscript, the authors refer to the polymerase “falling off” or dissociating from the template. This is the terminology used in earlier literature in the field and is a phenomenon that can occur with polymerases of positive strand RNA viruses. However, with recent breakthroughs in understanding of non-segmented negative strand RNA virus polymerase structure, it seems highly unlikely that the polymerase can dissociate from the template. Unlike the positive strand RNA virus polymerases, which have an open “right-hand” structure, the non-segmented negative strand RNA virus polymerases completely encircle the template, i.e., the template channel threads through the polymerase (e.g., Sala et al., 2025; PMID 40050611). The polymerase would need to undergo highly dramatic rearrangement, far greater than any structural variation that has been observed in the numerous polymerase structures available to date, to be able to dissociate from the template, except for unthreading when it reaches the template’s 5’ end. On the other hand, the polymerase could dissociate from the newly synthesized transcript, and there is ample evidence to indicate that this happens at the gene end signals, and during abortive RNA synthesis (e.g., in the event of capping failure). The available data indicates that polymerase that dissociates prematurely from the transcript cannot reinitiate at a gene start signal, likely because it needs to undergo the process of polyadenylation and transcript release to be able to recognize a gene start signal (likely because these processes cause necessary conformational changes in the polymerase). This aspect of polymerase structure and its relationship with the template doesn’t necessarily change the conclusions of this manuscript, but the manuscript text should reflect the biological details of the transcriptional machinery.

2. Lines 264-265 “Our model predicts abundant truncated transcripts, which are largely unobserved….it is possible that they ….are degraded much more rapidly than fully mature mRNAs”. The original paper providing experimental evidence for the stop-start model (Iverson and Rose, 1981; PMID6258804) examined RNA made both in cells and in vitro (in which RNA degradation would be minimal). In addition, Iverson and Rose specifically looked for prematurely terminated RNA and could not detect it, whereas they could detect prematurely terminated RNA generated from a UV-exposed genome RNA. On the other hand, a study of RSV transcripts generated in vitro did identify prematurely terminated mRNAs (Liuzzi et al., 2005; PMID 16189012). It is not clear if this is due to differences in the approach or differences between the viruses. Regardless this point is not adequately addressed.

3. Measurement of mRNA transcript levels in a cellular environment is confounded by transcript stability. This is an issue with all experiments measuring steady state levels of different viral RNAs (e.g., RNA-seq data) and with any experiments in which gene junction substitutions affected the mRNA sequences. The situation in which intracellular mRNA stability is not a confounding factor is in experiments in which only the non-transcribed intergenic sequence was substituted and its effect on transcription examined. In this regard, reviewer 3 recommended that the authors examine a paper by Finke et al., 2000 which showed that changing the non-transcribed intergenic region in rabies virus impacted transcription of the downstream gene. This paper shows that for this virus, the intergenic region had a significant impact. The authors did not include rabies virus in their model fitting (Figure 4) and have not adequately addressed this key comment.

4. The authors argue that the presence of overlapping genes is not consistent with the stop start model. Line 92-94: “The presence of overlapping genes in RSV, EBOV, and MV further challenges the stop-start model, which assumes unidirectional transcription”. This sentence conveys a lack of understanding of current models of mononegavirus polymerase behavior and the stop-start model of transcription. In considering the stop-start model it is important to distinguish between the polymerase in a transcribing mode in which the polymerase has the 3’ end of the transcript in its active site and is actively synthesizing RNA versus a scanning mode in which the polymerase does not have a transcript in its active site but remains associated with the template. In the transcribing mode, the polymerase can only move in a 3’ to 5’ direction relative to the template (it is “forced” to have a direction because of nucleotide addition onto the 3’ end of the transcript RNA), but in the scanning mode, there is nothing to determine in which direction the polymerase can move. The current model for how the polymerase accesses overlapping genes is that a transcribing polymerase releases the mRNA of the upstream gene at a gene end signal; having released the mRNA it becomes a scanning polymerase; it then can scan the template in both directions to find the next gene start signal and reinitiate. This model is supported by experimental evidence. E.g., examination of the mechanism by which the RSV L gene start signal is accessed showed that both the gene start and gene end signal of the upstream gene had to be intact for the polymerase to recognize the overlapping L gene start signal (Fearns and Collins, 1999; PMID 9847343). Similar results were also found in a study of Ebola virus overlapping gene junctions (Brauberger et al., 2015; PMID 26656691). The manuscript should be modified to acknowledge the experimentally derived data regarding gene overlaps.

5. Lines 82-99: In this paragraph, the authors point to apparent inconsistencies in the literature as justification for exploring an alternative model. However, this text contains numerous misconceptions and flaws. For example, this section cites Kuo et al., 1996 (PMID: 8709239) and Hardy et al.,1999 (PMID 9847319) as an argument that the literature on RSV is inconsistent. However, these papers examined different parts of the gene junction region. The manuscript by Kuo et al. examined the effect of substituting the intergenic region (i.e., the non-transcribed region that lies between the gene end and the gene start signals) whereas Hardy et al. examined the effect of switching out the gene junction (i.e., the gene end signal, the intergenic region, and the gene start signal). Kuo et al did not see an effect on the efficiency of upstream gene termination, whereas Hardy et al. did see an effect. The logical conclusion from comparing these papers is that in RSV the gene end signal is a key determinant of how efficiently the polymerase terminates mRNA synthesis at a gene junction, a conclusion which is borne out in another paper, also from Kuo et al. (Kuo et al., 1997, PMID: 9188557). Also note that the different groups used different experimental conditions that could have had additional effects. Hardy and coworkers were inadvertently using a mutant L for these studies that prevented efficient recognition of gene end signals – see Cartee et al., 2003; PMID 12805433 - and Kuo et al., were using virus superinfection to drive their minigenome, which meant that there was probably relatively limited M2-1 available; both these factors would affect levels of readthrough mRNA. Other evidence cited in this section also lacks attention to detail. For example, the authors state “For EBOV, attenuation at the L junction is unremarkable, while the NP/VP35 and VP30/VP24 junctions paradoxically suppress upstream gene expression”. However, in this paper, the region that was substituted was not only the gene junction, but also the non-translated regions of the upstream and downstream genes. As these regions could have a significant impact on the stability of the mRNAs, it is difficult to draw conclusions regarding efficiency of transcription across the gene junction from comparing the levels of the two mRNAs relative to each other (and this is not the point that the authors of the original paper were trying to make). Throughout the manuscript, the authors should pay more careful consideration of the prior literature, examining in detail exactly which sequences were substituted, which experimental system was used to interrogate them, and the fact that mRNA stability also needs to be considered. They should then integrate the findings regarding the effect of gene junction sequences into their model.

6. The authors acknowledge that co-transcriptional cap addition is a key process that could have a significant effect on mRNA synthesis. However, the significance of capping is not fully acknowledged in the manuscript. Multiple groups have shown that failure to cap causes abortive RNA synthesis. This could be a key factor causing attenuation at gene junctions.

Reviewer #3: I am a late comer to this review process of this manuscript, not being involved in the previous round of reviewing. This manuscript is interesting and provocative, and I can see from previous reviewer comments that there was consensus in that the conclusions of a previous version were a little overstated, with most of these claims being softened in a revised version.

In my opinion, the biggest fail of the previous stop/start model is in explaining transcription of the L gene, always in the promoter distal position of NSNS RNA viruses. To the best of my knowledge, transcription of this gene across all NSNS viruses is always lower than predicted by the stop/start model, inconsistent with the simple explanation of attenuation localized to the gene junction alone. The results shown in figure 4 are interesting as this newer model appears to account for this low L gene transcription quite well.

It is a failing of the field that no-one has really made any serious attempts to unpick this L gene expression discrepancy, even though it would not be such a tricky thing to do. However, we do have the results from gene rearranged viruses and for these viruses, I feel the new model has not done such a good job in explaining their transcription characteristics.

Major comments:

1. There are many examples, particularly in gene rearranged viruses where alterations to the L gene junction have affected its transcription levels. For example, doi: 10.1128/jvi.73.6.4705-4712.1999 and doi: 10.1128/JVI.73.8.6228-6234.1999. In the latter example, the MV H/L gene junction was altered by swapping out the upstream cistron resulting in a 4 fold difference in H/L gene expression ratios. Thus, L expression cannot be just a matter of how many nucleotides are being transcribed, This paper is mentioned, but not in any detail and it should be.

2. The model has a single value for P-off, but there are many examples where sequences are known to lead to altered attenuation. The two gene rearrangement papers, mentioned above for example, and there are many more. This should be stated, as surely it points to the new model being highly oversimplified.

3. I think the addition of a figure similar to the existing figure 4, showing how the stop/start model predicts mRNA levels and map onto the transcript data would be very useful.

4. For figure 6, I am curious as to how the RNA quantities were acquired, as the levels are not stated in these papers.

5. Line 95 – The statement, “Measles virus also deviates from the model, with higher downstream-than-upstream expression at the P-M junction and only ~60% attenuation before L” is not reflected in the corresponding figure 4

**********

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: None

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: No

Reviewer #3: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

Figure resubmission:

While revising your submission, we strongly recommend that you use PLOS’s NAAS tool (https://ngplosjournals.pagemajik.ai/artanalysis) to test your figure files. NAAS can convert your figure files to the TIFF file type and meet basic requirements (such as print size, resolution), or provide you with a report on issues that do not meet our requirements and that NAAS cannot fix.

-->

After uploading your figures to PLOS’s NAAS tool - https://ngplosjournals.pagemajik.ai/artanalysis, NAAS will process the files provided and display the results in the "Uploaded Files" section of the page as the processing is complete. If the uploaded figures meet our requirements (or NAAS is able to fix the files to meet our requirements), the figure will be marked as "fixed" above. If NAAS is unable to fix the files, a red "failed" label will appear above. When NAAS has confirmed that the figure files meet our requirements, please download the file via the download option, and include these NAAS processed figure files when submitting your revised manuscript.

Reproducibility:

To enhance the reproducibility of your results, we recommend that authors of applicable studies deposit 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

11 Apr 2026

Submitted filename: PLOS_Responses.pdf

5 Jun 2026

PCOMPBIOL-D-25-01912R1

The Transcriptional Gradient in Negative-Strand RNA Viruses Suggests a Common RNA Transcription Mechanism

PLOS Computational Biology

Dear Dr. Peccoud,

Thank you for submitting your manuscript to PLOS Computational Biology. After careful consideration, we feel that it has merit but does not fully meet PLOS Computational Biology's publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Aug 05 2026 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at ploscompbiol@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pcompbiol/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

* A letter that responds to each point raised by the editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. This file does not need to include responses to formatting updates and technical items listed in the 'Journal Requirements' section below.

* A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

* An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, competing interests statement, or data availability statement, please make these updates within the submission form at the time of resubmission. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

As the corresponding author, your ORCID iD is verified in the submission system and will appear in the published article. PLOS supports the use of ORCID, and we encourage all coauthors to register for an ORCID iD and use it as well. Please encourage your coauthors to verify their ORCID iD within the submission system before final acceptance, as unverified ORCID iDs will not appear in the published article. Only the individual author can complete the verification step; PLOS staff cannot verify ORCID iDs on behalf of authors.

We look forward to receiving your revised manuscript.

Kind regards,

Eric C. Dykeman, Ph.D.

Academic Editor

PLOS Computational Biology

James Faeder

Section Editor

PLOS Computational Biology

Additional Editor Comments:

Dear authors

I realise that this manuscript has been under review for a while. My apologies for the delay.

However there are just minor typos / corrections that one of the reviewers pointed out that I believe are important to be addressed for accuracy of wording that is used in the virus field.

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Authors:

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

Reviewer #1: I thank the authors for taking my comments into consideration.

Reviewer #2: This is a revised manuscript that provides a model to explain the transcription gradient that is characteristic of non-segmented negative strand RNA viruses. The authors have been responsive to prior critiques and have modified the manuscript to take into consideration details of previously published experimental data on this topic. The resulting manuscript is more nuanced and is a more accurate representation of the field and their findings. While the manuscript is largely sound overall, there are some minor editing errors that should be corrected.

Please note that line numbers refer to the marked-up version of the manuscript rather than the “clean” version.

1. Line 138-129: “..formation of 5’ cap during transcription is necessary for proper initiation and termination of transcription”. Here, “initiation” should be “elongation”.

2. Line 252: “…overlapping occurs and the promoter of a gene (gene 2)…”. Here “promoter” should be referred to as “transcription start site”.

3. Line 271: …”down the genome the termination signal (state 2)…”. Here, there should be “to” inserted between “genome “ and “the”.

4. Line 276 (Figure 3 legend): Here it should be stated that the two models are not mutually exclusive. It would be helpful if this was more clearly stated in the abstract and discussion sections too.

5. Throughout the manuscript, virus names do not follow convention as set out by the ICTV (https://ictv.global/sites/default/files/web-files/General_Information/How-to-write-virus%2C-species%2C-and-other-taxa-names.2.pdf). For example, on lines 107-110, “Virus” should not be capitalized, and the only virus names in this list with capitalization should be Marburg and Ebola (these are place names, i.e., proper nouns whereas the other viruses in the list are not named based on proper nouns), on line 289, Pneumoviridae and other virus family names listed here should retain the capitalization but also be italicized (these are formal taxonomic names rather than virus names), line 434 “Rabies Virus” should be “rabies virus”.

6. As noted above, the two models proposed are not mutually exclusive and it is highly likely that both come into play during non-segmented negative strand RNA virus transcription. The discussion hints at this but it could be stated more clearly. To add weight to their argument, the authors might want to consider citing a review on the extensive work that has been done in other polymerase systems (e.g., Zhang and Landick, 2016, PMID: 26822487).

Reviewer #4: King et al. present a model to describe the gradient in transcription of genes of nonsegmented negative-strand RNA viruses (NNSV). Their model uses one (or at most two) parameters to describe the process of attenuated expression, which has historically (since 1981) been described by four or more parameters based on sites of intergenic attenuation. The model is simple, elegant, and it performs well across multiple datasets of different NNSVs. Limitations of the model are clearly described.

**********

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: None

Reviewer #2: Yes

Reviewer #4: 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: No

Reviewer #4: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

Figure resubmission:

-->While revising your submission, we strongly recommend that you use PLOS’s NAAS tool (https://ngplosjournals.pagemajik.ai/artanalysis) to test your figure files. NAAS can convert your figure files to the TIFF file type and meet basic requirements (such as print size, resolution), or provide you with a report on issues that do not meet our requirements and that NAAS cannot fix.-->-->

After uploading your figures to PLOS’s NAAS tool - https://ngplosjournals.pagemajik.ai/artanalysis, NAAS will process the files provided and display the results in the "Uploaded Files" section of the page as the processing is complete. If the uploaded figures meet our requirements (or NAAS is able to fix the files to meet our requirements), the figure will be marked as "fixed" above. If NAAS is unable to fix the files, a red "failed" label will appear above. When NAAS has confirmed that the figure files meet our requirements, please download the file via the download option, and include these NAAS processed figure files when submitting your revised manuscript.-->

Reproducibility:

To enhance the reproducibility of your results, we recommend that authors of applicable studies deposit 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

11 Jun 2026

Submitted filename: PLOS_Response_to_Reviewers.pdf

12 Jun 2026

Dear King,

We are pleased to inform you that your manuscript 'The Transcriptional Gradient in Negative-Strand RNA Viruses Suggests a Common RNA Transcription Mechanism' 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,

Eric C. Dykeman, Ph.D.

Academic Editor

PLOS Computational Biology

James Faeder

Section Editor

PLOS Computational Biology

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

Reviewer's Responses to Questions

Comments to the Authors:

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

Reviewer #2: All the suggested revisions have been made.

**********

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 #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 #2: No

PCOMPBIOL-D-25-01912R2

The Transcriptional Gradient in Negative-Strand RNA Viruses Suggests a Common RNA Transcription Mechanism

Dear Dr King,

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.

For Research, Software, and Methods articles, you will receive an invoice from PLOS for your publication fee after your manuscript has reached the completed accept phase. If you receive an email requesting payment before acceptance or for any other service, this may be a phishing scheme. Learn how to identify phishing emails and protect your accounts at https://explore.plos.org/phishing.

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

With kind regards,

Janani Seenivasan

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