Kinship networks of seed exchange shape spatial patterns of plant virus diversity

By structuring farmers’ informal networks of seed exchange, kinship systems play a key role in the dynamics of crop genetic diversity in smallholder farming systems. However, because many crop diseases are propagated through infected germplasm, local seed systems can also facilitate the dissemination of seedborne pathogens. Here, we investigate how the interplay of kinship systems and local networks of germplasm exchange influences the metapopulation dynamics of viruses responsible for the cassava mosaic disease (CMD), a major threat to food security in Africa. Combining anthropological, genetic and plant epidemiological data, we analyzed the genetic structure of local populations of the African cassava mosaic virus (ACMV), one of the main causal agents of CMD. Results reveal contrasted patterns of viral diversity in patrilineal and matrilineal communities, consistent with local modes of seed exchange. Our results demonstrate that plant virus ecosystems have also a cultural component and that social factors that shape regional seed exchange networks influence the genetic structure of plant virus populations.

here. Some aspects related to these kinship systems are presented, but a (short) definition is required for the readers that are unfamiliar with this concept. All villages are virilocal as stated below Table 1, so do the matrilineal and patrilineal terms refer to who offers a dowry? In fact, for a naive reader, it seems obvious that the mode of transmission of cassava cuttings (documented in this study) should be the explanatory variable to address the question under study. However, after getting lost among the contradictions of Fig. 1 and Table 1, I came to conclude (despite the use of the 'matrilineal' term in Table 1 to describe maternal transfer of cassava cuttings) that kinship systems and cassava transmission are not as tangled as presented L56-61. The kinship system might be a secondary hypothesis to be tested if there are reasons to suspect it may contribute to plant and virus genetic structure beyond its effect through transgenerational cassava (and virus) transmission. To my opinion, this conceptual issue shifts the whole demonstration slightly offtarget.
-L119-120: I would think that this result showing that virus population structure matches the plant population structure is the main finding, or at least a key argument, of this study. In that respect, it deserves more than a one-line remark. A rigorous test of this association is expected here.
-L169-170: Contrary to what the authors state, the data on viruses found in MBG in 2006 and 2015 do not suggest inflow of new ACMV variants. Because sampling pressure is low (and substitution rate is high), a single virus variant has been sampled both in 2006 or in 2015. If the hypothesis of the authors is true, it implies (i) that the proportion of resampled variants is higher for traditional varieties than for new varieties and (ii) that the ACMV variants sampled in 2006 are more closely related to ACMV variants collected in 2015 on traditional varieties than on new varieties. Implication (i) is clearly not statistically significant because a single variant (MBG37) was sampled at both dates. Implication (ii) can be tested simply by comparing the distribution of the number of nucleotide differences to the closest 2006 variant for the 2015 variants from these two types of plants. In fact, I performed this Wilcoxon rank sum test, which resulted in a p value of 0.54. In the absence of a solid demonstration, this sentence should thus be removed.
- L193-195: Related to the point raised above (on L119-120), this strong assertion requires both a clarification on what exactly (virus diversity level, spatial structure of virus clades, virus phylogeny) is similar to what pattern of cassava diversity, and a clear demonstration of the new formulation of this main conclusion.
- Figure 1: There are incongruences between Fig. 1, Fig. S4, Tables 1-3 and the text. ODJ is matrilineal in the text (L118) and Fig. 1 (red dot), but patrilineal or mixed in Tables 1-3 (although  from Table 1 it is hard to understand why ODJ is not matrilineal). IMB is patrilineal in Fig. 1 (blue dot), but mixed in the text (L118) and Fig. 3, and patrilineal or mixed in Tables 1-3; in addition, in IMB the transmission of cassava cuttings by the mother is far more frequent than by the motherin-law (typical of matrilineal villages). I understand that it is hard to assign each village to a single marriage system and that choices are necessary; however, such choices must be self-consistent. This becomes a bigger issue when the decision is made subsequently to treat IMB and MOP as patrilineal in Fig. S4 and more importantly in Fig. 3. For a fair comparison of the impact of matrilineal and patrilineal systems, it is unwise to assimilate IMB and MOP to patrilineal villages; they should either be excluded from the analysis if the kinship system is really the most relevant explanatory variable, or IMB and ODJ should both be included in the group with dominant maternal transmission if the mode of transmission is considered.
Lesser issues: -In the abstract, 'dynamics' should be removed L30, replaced with 'modes' or 'rules' L31, and replaced with spread L33.
-I think the relationship between social organization and virus prevalences deserves being tested and discussed. At first sight, it seems that ACMV prevalence does not differ between patrilineal and matrilineal villages, which some may see as counterintuitive especially under the risk-aversion hypothesis made in Deletre et al. 2011.
-The introduction should specify where the previous work stopped, and what outstanding question is treated in this article.
-Sociological and genetic patterns of IMB and MOP are similar in Figs. 1 and 2. I did not find this information and its discussion, which I would find noteworthy, in the text.
- Fig. 2 should be split into 2 figures, maybe grouping A&C and B&D to focus the comparison on genetic patterns for the virus and the plant rather than on the two representations of each pattern.
In addition, pie charts should be replaced with the usual bootstrap proportions (or at least by black pie charts to prevent confusion with the maternal/southern color code).

Minor issues and typos:
-please define acronyms at first use (ACMV L30 not L77, NMDS L251 not 367, SSR L642) -L44, 'chronic' might be replaced with 'endemic' -L85: 'mutation' should be replaced with 'evolutionary' -L93: replace 'although' with 'however," -L120: 'four plant clusters' -L165-167: replace 'genetic relationships among viral isolates' with 'viral clades' to simplify the sentence -L181-182: remove sentence duplicating information given 2 lines before -L212: remove 'levels of' -L220-221, L237-238 and L277  : sentence on small empty circles not needed (copy-paste error from Fig. 4) -I am surprised by the absence of indication of any specific contribution by one author Reviewer #3: Remarks to the Author: This is a very interesting paper. It shows that the local structure and evolution of cassava mosaic geminivirus populations in an area of central Africa is intimately related to marriage customs. Management of the disease should take this into account. The phytopathological literature contains few studies linking disease epidemiology to sociology; in this the deeply embedded crossdisciplinarity leads to clear conclusions. Put in the right context -as here -it is logical to investigate the hypothesis tested However, it is rare to see any issues beyond an assumption of economic rationality included in discussions of disease management. I found the paper clear, wellargued and extremely stimulating and I congratulate the authorial team.
I noted a few very minor issues as I read the manuscript by line number or Figure number: 141 delete "on" eg line 174 Try to reduce abbreviation use. Abbreviations often make writing easier but reading harder. eg Misele -> MIS simply complicates things and makes confusion between methods, genotypes and locations (etc) worse, not better 205 "plays...role" Better: "prevails" or better still "is the manin method" 379 These references use a Harvard variant but the reference list is numeric. Cheng appears present 679 lack of temporal" insert "change" Fig S4. These are barely readable. As this is SI there is no reason not to adopt a more capacious presentation with the lines and markers more in proportion to the individual panels 704-707, Fig S6 Is

CORRECTIONS TO THE MANUSCRIPT NCOMMS-20-25950
Detailed answers to each remark/question are given below, as well as information on how we complied with suggestions in the revised manuscript. Each section addresses the comments of one reviewer. For ease of reference, reviewers' original comments are reproduced verbatim and in full at the start of each section. Remarks are repeated at the start of each paragraph, each addressing a specific question. Actions we took are shown in bold red.

Reviewer #1
-"This is an interesting study that proposes farmer social structure to influence cassava mosaic disease. -The fact that the matrilineal and patrilineal communities are spatially distinct is a major issue (but not the only one).
It is important to stress that although the strong cultural geographical contrast observed in Gabon may appear as a major bias, there are actually very few places in Africa where different social structures coexist and allow a comparative study of the impact of seed exchange on the genetic diversity of viral populations (and their host plants) without many of the confounding factors that would inevitably arise in a cross-cultural study over a much wider geographic region. In West Africa, societies are mostly (only) patrilineal, while in East Africa societies are predominantly matrilineal. In Central Africa, matrilineal systems predominate, forming a "matrilineal belt" that extends eastward across all of Africa through Gabon, the Congos, Zambia, Malawi, and Tanzania (Murdock 1967). In Gabon, societies with contrasted rules of descent coexist within a relatively small and otherwise relatively homogenous geographic area, which partially offsets the issue of spatial distinction.
-These villages may differ in a myriad of factors including climate, vector abundance and movement, the abundance and diversity of alternative hosts (many have been identified for ACMV), farm spatial connectivity etc etc.
Virus spread and whitefly populations are dependent on climatic factors that influence cassava growth, including temperature and rainfall (Fargette and Thresh 1994), but also topography (elevation) and vegetation cover. Below we present additional background information about the distribution of study sites relative to several environmental factors that could influence the distribution of ACMV. We also discuss briefly alternate hosts to cassava geminiviruses. Most of the discussion and novel findings presented below have been incorporated into the manuscript.  861-877 (2004).

(A) Vegetation cover
As we wrote above, Gabon is relatively homogenous, with no major differences in vegetation cover. Nearly 85% of the territory (267,667 km 2 ) is covered with equatorial forest, the rest by small areas of savannah and mangroves, while cultivated areas represent <1% of the territory (FAO 2008) ( Fig. 1-A).

(B) Elevation
In Madagascar, Harimalala et al. (2015) showed that whitefly abundance and CMD prevalence are sensitive to elevation and decrease as altitude increases (see also Thresh et al. 1994). In particular, they found that the distribution of ACMV varied significantly with altitude compared to other cassava mosaic geminiviruses, with higher prevalence at higher altitude (> 800 m a.s.l.). We did not observe such relationship for ACMV prevalence in Gabon (Fig. 2), possibly because of the limited altitudinal range in Gabon (the highest point is at 1,000 m.a.s.l.) and small number of study sites ( Fig. 1-B). In our sample, elevation ranged from 0 to 600 m.a.s.l.   459-478 (1994).
(C) Temperature seasonality (variable BIO4) The relative importance of temperature and rainfall on CMD epidemiology varies depending on how seasonality affects cassava growth. In humid environments, the influence of temperature is more important than in drier areas where rainfall is the limiting factor for crop growth (Fargette andThresh 1994, Thresh et al. 1994).
Temperature seasonality measures the amount of temperature variation over the year based on the standard deviation of monthly temperature averages. Temperatures vary more in the southern part of Gabon than in the north ( Fig. 1-C). However, we found no correlation between temperature seasonality and viral diversity ( 1 D, the effective number of phylotypes detected in the community; Pearson's r(10) = 0.54, p value = 0.068).  459-478 (1994).

(D) Annual precipitation (variable BIO12)
With the exception of CCB where rainfall is the highest in Gabon, total annual precipitation does not vary significantly across the study sites ( Fig. 1-D) and we found no correlation between rainfall and the prevalence of ACMV across the area surveyed (Fig. 3). Seasonal variations of temperature and rainfall greatly influence cassava growth and have been associated with changes in abundance of whiteflies and in the incidence of ACMV. Precipitation seasonality is the only factor in Gabon that shows a clear -although moderategeographic contrast, with overall greater variability in the southern part of the country than in the north. Precipitation seasonality (coefficient of variation) measures the variation in monthly precipitation over the course of the year. It is calculated as the ratio of the standard deviation of monthly precipitation to the annual mean. It is expressed as a percentage of precipitation variability. While we found no correlation between the prevalence of ACMV and either BIO4 (temperature seasonality; Pearson's r(10), p value = 0.469) or BIO15 (precipitation seasonality; Pearson's r, p value = 0.301), we found a strong correlation between precipitation seasonality and viral diversity (Pearson's r = 0.62, p value = 0.032). Whether seasonal variations could have an influence on the spatial distribution of ACMV lineages through their influence on cassava growth cannot be ruled out and would need to be investigated, but we consider that it is very unlikely that they would explain differences in viral diversity that we observed between matrilineal and patrilineal societies, in particular greater genetic relatedness among viral haplotypes in the latter. At the village level, cassava varietal diversity was on average twice higher in matrilineal villages than in patrilineal villages. Viral diversity was highly positively correlated with host diversity, while host diversity was not correlated with environmental factors. In the revised manuscript, we discuss the potential influence of climatic factors on the spatial distribution of the two main viral clades (L. 309-321). Maps presented here have been added as Supplementary file S7. We also corrected a mistake in  -Also, the framing of the study is a bit off in that is presented as an epidemiological study while the data is in fact viral diversity, not measuring a change in virus prevalence in space-time.
We fully agree that the focus of the study is on differences between matrilineal and patrilineal societies in terms of viral diversity. Reviewer 2 made a similar point, suggesting changing the title of the paper. We have endeavored to address this confusion in the revised manuscript by shifting the focus on the population structure and genetic diversity of the virus and not the epidemiology of the disease. We also changed the title of the paper to "Kinship networks of seed exchange shape spatial patterns of plant virus diversity".
-The available data and information should be incorporated into a predictive (epidemiological) modelling framework that would allow estimating the relative importance of different potential factors that contribute to virus diversity to truly aim to validate the importance of farmer social structure on Cassava Mosaic Disease.
We agree that there is a need for further research, supported by epidemiological models, to account for all the parameters that could influence the dynamics of CMD before any decisive conclusion can be drawn on the relative importance of sociological factors in the dynamics of crop plant diseases. This is very ambitious, however, and exceeds the scope of our current paper, but we do hope that our study will stimulate further research in this direction. As we stressed in the manuscript, and as all three reviewers agree, the strength of our paper is that it bridges across natural and social sciences, combining social anthropology and plant virology in a novel way, while offering fresh insights on the problematic of resilience of smallholder farming systems to emerging plant pathogens. The study is built upon a unique, crossdisciplinary dataset that offers a rare opportunity to explore how social factors influence the spatial distribution of plant viruses.
In the ten years that followed our first paper on seed transmission systems in which we highlighted the importance of social structures in promoting exchanges of germplasm between communities of smallholder cassava farmers (Delêtre et al. 2011), there has been a fastexpanding body of literature demonstrating the importance of social factors, in particular kinship systems and matrimonial networks, for the spatial dynamics of crop diversity (e.g., Westengen et al. 2014;Labeyrie et al. 2016). We believe that our manuscript makes an important novel contribution to this larger body of research by investigating the corollary impact of seed exchanges on movements of seedborne pathogens (in our case, infected cassava cuttings). Our study addresses particularly topical questions in little-explored domains of plant epidemiology, food security and sustainable development. This study is intended as a proof of concept and we hope it will pave the way for further research on the interplay of social structures and local/regional networks of seed exchange and their role in the dynamics of crop plant diseases. We acknowledged in the discussion the need for including our findings within a robust modelling framework that would allow disentangling the effects of social factors from that of other environmental parameters (L. 448-451).   Figure 1: There are incongruences between Fig. 1, Fig. S4, Tables 1-3 and the text. ODJ is matrilineal in the text (L118) and Fig. 1 (red dot), but patrilineal or mixed in Tables 1-3 (although from Table 1 it is hard to understand why ODJ is not matrilineal). IMB is patrilineal in Fig. 1 (blue dot), but mixed in the text (L118) and Fig. 3, and patrilineal or mixed in Tables 1-3; in addition, in IMB the transmission of cassava cuttings by the mother is far more frequent than by the mother-in-law (typical of matrilineal villages). I understand that it is hard to assign each village to a single marriage system and that choices are necessary; however, such choices must be self-consistent. This becomes a bigger issue when the decision is made subsequently to treat IMB and MOP as patrilineal in Fig. S4 and more importantly in Fig. 3  In matrilineal societies, the main source of the varieties is the farmer's mother, whereas in patrilineal societies it is exclusively the mother-in-law (with some exceptions). Farmers also acquire varieties through other networks e.g., from neighbors, but such horizontal exchanges are much less frequent in patrilineal villages compared to matrilineal villages. In pluriethnic communities (e.g., Mopia), vertical transmission seems to replace affinal transmission where intermarriage is more frequent. However, it is worth noting that affinal transmission never appears among matrilineal societies or in mixed societies of matrilineal descent but is systematically observed among patrilineal societies.
It is important to stress that seed inheritance rules apply to married farmers, while widows and divorced farmers often derogate to the affinal rule (in patrilineal communities), which might give a misleading impression that affinal transmission is not that important. Single women still reside in their maternal village and grow cassava cuttings that they received from their mother. Divorced farmers, on the other hand, usually return to their maternal village and revert to growing varieties present in the village, leaving behind those they grew in their husband's household. In our paper we did not specify the marital status of farmers we interviewed, but we realize that this information is important. Information about the typology of farmers (village of origin, status and source of their varieties) is now provided as a supplementary file (Table  S3). These data were used to produce Fig. 1 and Table 1 and help clarifying some of the apparent contradictions noted by Reviewer 2.

-I found the title slightly misleading because the article provides little information on dynamic aspects and, although cassava mosaic disease (CMD) threatens a staple food in Africa, the processes under study (transgenerational germplasm transmission) corresponds more to endemic disease spread than to pandemic spread. In addition, 'social organization' is a bit vague. As an alternative, I would suggest something like: 'Marriage systems influence the spread of a plant virus'.
Following remarks from Reviewer 1, who made a similar remark, we changed the title of the paper to "Kinship networks of seed exchange shape spatial patterns of plant virus diversity". We also avoided using the term 'social organization' and replaced it by 'kinship systems' in the title and through the manuscript to reflect the focus of the paper on social rules that structure marriage networks between communities. Finally, we clarified in the manuscript that we are focusing on endemic circulation of viral lineages and not implying a role of seed exchange networks in the wider pandemics of CMD in Africa.
-L54-61: The kinship (matrilineal and patrilineal) system is the main explanatory variable studied here. Some aspects related to these kinship systems are presented, but a (short) definition is required for the readers that are unfamiliar with this concept.

A definition of kinship system was added in the introduction (L. 63-66): "Kinship systems are cultural representations of relationships between individuals based on the notion of clan membership. By defining rules of descent and incest prohibitions, kinship systems structure matrimonial networks between communities and normalize social interactions between kin (related by descent) and affine (related by marriage)."
-All villages are virilocal as stated below All villages in Gabon are virilocal, meaning that the bride leaves her maternal village to move in with her husband. 'Matrilineal' and 'patrilineal' refer to the mode of transmission of clan membership. In matrilineal societies, children born after marriage belong to the mother's clan.
In patrilineal societies, they belong to the father's clan. As we wrote in Delêtre et al. (2011), "in most patrilineal societies, the family-in-law must pay a "bride-price" in return for the transfer of jural authority over the bride. Although seeds do not enter into the composition of this bridewealth, divergences in rules of seed transmission between patrilineal and matrilineal systems reflect a difference in the investment of grandparents in the offspring of the marriage. In matrilineal virilocal societies, marriage payments tend to be rare or absent. Because filiation is handed down through the female line, daughters remain attached to their maternal kin after marriage, and children belong to their maternal clan. In patrilineal virilocal societies, in contrast, the parents of the groom must repay the parents of the bride for the acquisition of rights in genetricem-that is, the right to filiate the children to be born to their daughter-inlaw".
The main argument developed by Delêtre et al. (2011) is that seed transmission is an element of the economic system of traditional societies (sensu Meillassoux 1960), which determines rules of inheritance by asking the question "in whom is it better to invest?". In patrilineal societies, affinal transmission (mother-in-law to daughter-in-law) makes more sense than vertical transmission, socially but also economically. Seeds are part of the patrimony that must be transmitted. Investing in a daughter makes little sense for a mother because her children will be "lost" for the clan, while a daughter-in-law is more "valuable" because her children will belong to the clan. Similar rules of seed inheritance apply to other crops and intricate links between marriage exchanges and seed exchanges have been described for cassava in Amazonia ( To address this point, we applied Syrjala's distributional test (Syrjala 1966) to test for spatial congruence between host and virus spatial clusters. Syrlaja test is a nonparametric statistical procedure based on a generalization of the Cramér-von-Mises test, which can be used to test the null hypothesis that there is no difference in the spatial distributions of two populations. As the test in nonparametric, no assumption is made about the distributions of the populations. The test is sensitive to differences in spatial distributions and density gradients between two populations but not to differences in abundance between the two populations. The test uses georeferenced density measures (here the prevalence of each of the three viral clades identified by BAPS) to test the null hypothesis that across the study area the normalized distributions of the two populations are the same. The test statistic (Ψ) is calculated as the squared difference between two cumulative distribution functions summed over all sampling locations. Significance is tested through data permutations (here 10,000) by repeatedly and randomly recalculating the original data after permutation, each time recomputing the test statistic to generate an empirical distribution of permuted values. The P-value is then determined by evaluating where the original test statistic falls within this empirical distribution.
Viral populations fall essentially within two main clades: north-eastern [blue] and southwestern [red]. A third, smaller eastern clade [yellow] was also observed, mainly in ODJ but also in MOP and IMB. Overall, the spatial distribution of these three lineages matched the four geographic clusters observed among host plants, with a clear southwestern genetic cluster encompassing all matrilineal villages in the south, a northern cluster restricted to patrilineal villages in the north, and an eastern cluster comprising ODJ, MOP and IMB. A fourth cluster mainly associated with CCB was also detected. The spatial distribution of the southwestern genetic cluster of the host plant matched that of the south-western viral clade, whereas the spatial distribution of the northeastern clade matched the northern and eastern clusters detected among the host plant.
The test did not reveal any significant difference between the spatial distribution of the two main viral lineages (north-eastern and south-western) and that of the host main genetic clusters, whereas differences were significant between host NE-virus SW and host SW-virus NE (Table 1).  We rewrote the corresponding section to give a more accurate description of the spatial patterns observed and include results of the test (L. 141-153). We have also added a description of Syrjala's distributional test in the Methods (L. 565-566) and updated the References section accordingly.
- The sentence has been amended and now reads "The striking similarity between the spatial distribution of viral clades of ACMV and the distribution of cassava genetic clusters in Gabon suggests that the spread of the virus is constrained by factors that shape cassava diversity at the landscape level." (L. 276) - Figure 1: There are incongruences between Fig. 1, Fig. S4, Tables 1-3 and the text. ODJ is matrilineal in the text (L118) and Fig. 1 (red dot), but patrilineal or mixed in Tables 1-3 (although from Table 1 it is hard to understand why ODJ is not matrilineal).

L169-170: Contrary to what the authors state, the data on viruses found in MBG in 2006 and 2015 do not suggest inflow of new ACMV variants. Because sampling pressure is low (and substitution rate is high), a single virus variant has been sampled both in 2006 or in 2015. If the hypothesis of the authors is true, it implies (i) that the proportion of resampled variants is higher for traditional varieties than for new varieties and (ii) that the ACMV variants sampled in
Odjouma (ODJ) is a small Teke community located at the border of Gabon with Congo. Patterns of seed exchanges observed in ODJ are typical of a matrilineal community (i.e., mother ® daughter). We chose to treat ODJ as a matrilineal community to be consistent with our -IMB is patrilineal in Fig. 1 (blue dot), but mixed in the text (L118) and Fig. 3, and patrilineal or mixed in Tables 1-3; in addition, in IMB the transmission of cassava cuttings by the mother is far more frequent than by the mother-in-law (typical of matrilineal villages). I understand that it is hard to assign each village to a single marriage system and that choices are necessary; however, such choices must be self-consistent. This becomes a bigger issue when the decision is made subsequently to treat IMB and MOP as patrilineal in Fig. S4 and more importantly in Fig. 3

. For a fair comparison of the impact of matrilineal and patrilineal systems, it is unwise to assimilate IMB and MOP to patrilineal villages; they should either be excluded from the analysis if the kinship system is really the most relevant explanatory variable, or IMB and ODJ should both be included in the group with dominant maternal transmission if the mode of transmission is considered.
IMB is a mixed community of patrilineal Bantu and sedentarized Pygmy farmers. As explained in more details in the footnotes of Dataset S1, rules of descent and post-marital residence are more ambivalent among hunter-gatherer groups as they often adopt the social organization of their Bantu neighbors (see Knight 2003, Matsuura 2006, and they are sometimes qualified as 'mixed' (Le Bomin and Mbot 2012). Mixed marriages between Bantu and Pygmy appear to be mostly limited to unions between Pygmy women and Bantu men (Batini et al. 2011 and references therein). Unusually, in IMB we noticed that among Pygmy farmers cassava cuttings were often inherited from the maternal grandmother, which can be assimilated to matrilineal inheritance. Because Pygmy still face discrimination, offspring of mixed marriages are sometimes considered Pygmy (see Knight 2003), and Pygmy women have to rely on their maternal kin to source cassava cuttings. Agriculture is still a relatively recent activity among Pygmies (in IMB, it started in the 1960s; Soengas 2009) and Pygmy farmers would have originally borrowed their varieties from their Bantu neighbors. Horizontal exchanges, usually from Bantu neighbors at the request of Pygmy farmers, are also frequent (Soengas 2009).
In comparing levels of viral diversity between matrilineal and patrilineal villages, we grouped villages based on the rule of descent that predominates in the community (the explanatory variable tested is the kinship system). ODJ was treated as 'matrilineal' whereas IMB and MOP were both treated as 'patrilineal'. In MOP, 17/21 of farmers interviewed were patrilineal. In IMB farmers were in majority Pygmy (14/21), but as cassava farming is heavily influenced by Bantu neighbors (all patrilineal, with affinal seed transmission rules) we felt justified to treat all farmers in IMB as patrilineal. Removing IMB and MOP from the analysis does not actually change results. Tables have been modified so that IMB and MOP are listed as patrilineal and ODJ is now included with matrilineal villages.
The abstract has been modified accordingly.
-I think the relationship between social organization and virus prevalences deserves being tested and discussed. At first sight, it seems that ACMV prevalence does not differ between patrilineal and matrilineal villages, which some may see as counterintuitive especially under the risk-aversion hypothesis made in Deletre et al. 2011.
In Delêtre et al. 2011, we indeed mentioned indeed risk aversion strategy as one possible reason for farmers to rely mostly on their local varieties and not to seek new varieties outside their community. The cost of choosing affinal transmission over vertical transmission may appear counter-adaptive as it deprives farmers of an important source of diversity. Similarly, there is a higher risk in allowing germplasm to move freely between communities. However, as we develop later on in that paper, seed inheritance systems respond above all to a need for social coherence and participate in strengthening social cohesion. By analysing the social and economic role of seed transmission one can understand differences between patrilineal and matrilineal societies (Delêtre et al. 2011).
Field interviews have revealed that farmers were generally unconcerned by or unaware of cassava diseases (in patrilineal and matrilineal societies alike) and did not identify CMD as a disease (Delêtre 2010 andDelêtre 2004, unpublished). In simple infection, the impact of ACMV on the yield was insignificant and went unnoticed by farmers, who therefore did not rogue infected plants. In fact, as Fresco (1986) reported in RD Congo and as we did too in Gabon (M. Delêtre, pers. obs.), cassava leaves showing symptomatic chlorosis and foliar deformation are sometimes preferred for culinary preparations as 'they taste better'.
As ACMV is ubiquitous wherever cassava is grown -one of the reasons we chose to focus on this particular virus -it is not surprising that we do not observe any difference in prevalence between patrilineal and matrilineal villages. With the arrival of the Ugandan strain of the East African cassava virus (EACMV-UG) in Northern Gabon, however, farmers' attitude towards CMD has changed and it would be interesting to compare the prevalence of EACMV-UG in patrilineal and matrilineal villages now that the variant is firmly established all over the territory. We do not have any recent data for the matrilineal/southern part of the country, but this would certainly be an interesting research question for a follow-up project. -The introduction should specify where the previous work stopped, and what outstanding question is treated in this article.
The previous study on seed exchange networks (Delêtre et al. 2011) is based on the first author's doctoral thesis (Delêtre 2010). The data presented in the 2011 paper represent only one aspect that the thesis covered. Other factors that could influence regional patterns of the host plant diversity (social, cultural, economic, historical, and environmental) were also examined in detail. In each village, the history of cassava introduction was documented, and the organization of the farming systems was explored through semi-directive interviews, structured into five major themes: (i) land management (rules of appropriation of land and fallow), (ii) agricultural calendar (timing of clearing, burning, planting and harvesting), (iii) intercropping (distribution of crops in time and space), (iv) weeding and (v) pest management (identification of local pests and diseases). Farmers were also questioned about their folk ecological knowledge, in particular about cassava reproductive biology (flowers, fruits, and seeds) and criteria of selection of cuttings. Folk nomenclature systems of cassava landraces were also examined. Although attention was given to farmers' knowledge about and behaviors towards cassava diseases, the material collected was not screened for pathogens at the time, but factors impinging on spatial patterns of cassava diversity are now well understood (see Delêtre 2010, Delêtre et al. 2011, Delêtre et al. 2016, and McKey and Delêtre 2017.
The dataset on which we build the present manuscript on the African cassava mosaic virus is therefore unique in scope, depth of study, and cross-disciplinarity, and particularly suited for exploring the impact of social networks on the spread of crop plant viruses. Although much work has been done in recent years on the importance of seed exchange networks for agrobiodiversity ( -Sociological and genetic patterns of IMB and MOP are similar in Figs. 1 and 2. I did not find this information and its discussion, which I wouldfind noteworthy, in the text. Imbong (IMB) and Mopia (MOP) are both mixed communities of (predominantly) patrilineal descent that belong to the eastern genetic cluster identified in the host plant. In both villages, the lesser common eastern clade of the virus (which is predominant in Odjouma), was also detected. However, the sociological settings in IMB and MOP are quite distinct.
Mopia (MOP) is a large pluriethnic settlement comprising eight different linguistic groups, some patrilineal and other matrilineal. Mixed marriages are frequent in the village. As developed in Delêtre et al. (2011), in heterogeneous communities patterns of seed transmission do not reflect a dominance of vertical or affinal transmission but instead a mixture of both. Vertical transmission seems to replace affinal transmission where intermarriages are more frequent, reflecting a need for alternative strategies of social reproduction in multicultural contexts. This is particularly true in the case of mixed marriages between Pygmy and Bantu communities, as in IMB.
-add spaces between numbers and their units Spaces have been added.
- Fig. S5: sentence on small empty circles not needed (copy-paste error from Fig. 4) The duplicated sentence has been removed (L. 900).
-I am surprised by the absence of indication of any specific contribution by one author" This has been rectified.
-"This is a very interesting paper. It shows that the local structure and evolution of cassava mosaic geminivirus populations in an area of central Africa is intimately related to marriage customs. Management of the disease should take this into account. The phytopathological literature contains few studies linking disease epidemiology to sociology; in this the deeply embedded cross-disciplinarity leads to clear conclusions. Put in the right context -as here -it is logical to investigate the hypothesis tested However, it is rare to see any issues beyond an assumption of economic rationality included in discussions of disease management. I found the paper clear, well-argued and extremely stimulating and I congratulate the authorial team.
I noted a few very minor issues as I read the manuscript by line number or Figure number: -141 delete "on" -eg line 174 Try to reduce abbreviation use. Abbreviations often make writing easier but reading harder. eg Misele -> MIS simply complicates things and makes confusion between methods, genotypes and locations (etc) worse, not better -205 "plays...role" Better: "prevails" or better still "is the manin method" -379 These references use a Harvard variant but the reference list is numeric. Cheng appears present -679 lack of temporal" insert "change" - Fig S4. These are barely readable. As this is SI there is no reason not to adopt a more capacious presentation with the lines and markers more in proportion to the individual panels -704-707, Fig S6 Is  -141 delete "on" • This sentence has been changed (L. 184).
-eg line 174 Try to reduce abbreviation use. Abbreviations often make writing easier but reading harder. eg Misele -> MIS simply complicates things and makes confusion between methods, genotypes and locations (etc) worse, not better • We have amended the text to limit the use of abbreviations for locality names in the main text.
-205 "plays...role" Better: "prevails" or better still "is the manin method" • The original sentence "Dramatic increase in whitefly population density has been shown to play a central role in driving the epidemic front of severe CMD pandemics" reads now: "Dramatic increase in whitefly population density has been shown to drive the epidemic front of severe CMD pandemics" (L. 296).
-379 These references use a Harvard variant but the reference list is numeric. Cheng appears present • To facilitate the reviewing process, the reference section was re-ordered alphabetically and will later be formatted as per the journal guidelines.
- Fig S4. These are barely readable. As this is SI there is no reason not to adopt a more capacious presentation with the lines and markers more in proportion to the individual panels • Fig. S4 has been edited to increase readability. -710- Fig S7. This should be in table form as well, since otherwise anyone looking at it has to reconstruct the numbers with a (software or hardware) ruler." • All datasets supporting the figures presented in the paper are now provided along with the manuscript (see "Source data").