Comment on Rohrscheib et al. 2016 "Intensity of mutualism breakdown is determined by temperature not amplification of Wolbachia genes"

Unlike other intracellular bacteria, Wolbachia genomes are highly labile largely due to the presence of repetitive sequences such as transposons, active phage and high rates of recombination [1–3]. Within the wMelPop genome a locus of considerable instability, referred to as the Octomom locus, has been described by several studies [4–8]. As this locus is one of the few variable loci between wMel (single Octomom locus, low bacterial density and non-pathogenic) and wMelPop (3–12 Octomom loci, high bacterial density and pathogenic) genomes, and no genetic transformation tools are available, there is considerable interest in understanding the link between genotype and phenotype. Chrostek and Teixeira hypothesized “that Octomom region amplification underlies wMelPop virulence” and went on to correlate increased Octomom copy number to increased bacterial density within the host and strength of pathology [5]. Absent from this study was a systematic assessment of temperature, which previous studies had shown to affect wMelPop pathology, and its effect on wMelPop density and Octomom copy number. We hypothesized that temperature would affect Octomom copy number and/or gene expression, and that such changes would affect the strength of pathology [8]. Our study showed that while Octomom copy number did vary over developmental time, no consistent trend was observed among Octomom copy number, bacterial density nor pathology [8]. We determined that temperature, not Octomom copy number or bacterial density, had the greatest effect on wMelPop infected host lifespan. The link between Octomom copy number and pathology was further challenged by our discovery of a pathogenic Wolbachia strain (wMel3562) that maintained a single copy of the Octomom locus, and supports an earlier study of a wMelPop variant that lacks the Octomom locus but which is still pathogenic [7–9]. These observations suggest that at the very least, genetic elements beyond the Octomom locus are responsible for pathology...

particular, single flies from a non-controlled wMelPop Drosophila stock have Octomom copy numbers ranging from two to ten copies (Fig. 1A of Chrostek and Teixeira [4]). This variability seems to decrease under selection for Octomom copy number ( Fig. 2 and Fig. S2 of Chrostek and Teixeira [4]). However, some variation is either maintained or constantly generated, as selection for Octomom copy number could be reversed either by direct reverse selection or by relaxing selection in high Octomom copy number stocks ( Fig. S6 and Fig. S8 of Chrostek and Teixeira [4]).
Rohrscheib et al. report substantial changes in Octomom copy number throughout Drosophila life. The average Octomom copy number increases more than two fold in the first eight days of adult life and subsequently decreases more than four fold. These large directional changes could be explained by heterogeneity in wMelPop Octomom copy number between flies. According to our model, during early adult life of Drosophila, wMelPop with higher Octomom copy numbers proliferate more than wMelPop with lower Octomom copy numbers. Consequently, over time their contribution to the pool of total wMelPop in the host population increases and total mean Octomom copy number increases. Next, flies carrying wMelPop with high Octomom copy numbers die faster and, therefore, at later time points these Wolbachia are depleted from the pool of total wMelPop and mean Octomom copy number decreases. This and similar explanations, based on conclusions of Chrostek and Teixeira [4], are discussed and dismissed by Rohrscheib et al. However, when we simulate Octomom copy number variation in pooled samples from a heterogeneous population, we can observe initial increase of mean Octomom copy number followed by a later decrease (Fig 1 in this text), similar to the dynamics reported by Rohrscheib et al.
3) The most puzzling result presented in Rohrscheib et al. is that average Octomom copy number equals 1 to 1.5 at various temperatures and it is about 0.5 copy at 18˚C. These estimates are much lower than estimates from our previous publications [4,5] and from an independent report on wMelPop in Drosophila melanogaster [6] co-authored by some of the authors of the paper under discussion. wMelPop strains with zero copies of Octomom were only reported in Aedes aegypti cells and mosquitoes (wMelPop-CLA and wMelPop-PGYP) [6].
This striking difference in wMelPop Octomom copy numbers estimates between this and previous reports requires a more detailed characterization and validation.
1. None of the graphs or tables with Octomom qPCR data shows control wMel variants with only one Octomom copy in the genome. Presenting the data for these lines is necessary to validate and interpret qPCR results.
2. The calculation of Octomom copy number is crucial for this discussion. It is not clear how the authors calculate specific Octomom copy numbers based on the qPCR results for WD0550 and WD0508. To determine Octomom copy numbers per genome the authors state that they used the delta-delta Ct method [7], but cite the paper on the Pfaffl method [8]. These are similar but not equivalent methods. The Pfaffl method requires determination of the efficiencies of the primers and includes them in the formula, while the deltadelta Ct assumes that the efficiencies are equal to 2 (100%). Nonetheless, both methods use a calibrator sample and the calculated fold changes are expressed relative to this calibrator [7,8]. The use of any of these two methods to calculate gene copy number variation requires a calibrator sample with known ratio of gene of interest to the reference gene [9]. In this particular case a calibrator with known Octomom copy number is required. Otherwise, both methods produce adimensional values, appropriate to compare the magnitude of change between samples within a single experiment, but inappropriate to estimate Octomom copy number. Since Rohrscheib et al. did not specify what they used as a calibrator, it is impossible to determine what the presented qPCR results represent.
3. Finally, sequencing of the wMelPop used in this study would show if it is otherwise identical to the Drosophila melanogaster wMelPop sequenced in [5] and [6] or if it acquired other mutations, similarly to the wMelPop-CLA and wMelPop-PGYP in Aedes aegypti cells and mosquitoes [6].

4) When calculating
Wolbachia levels in Drosophila the authors also report Wolbachia to host genome estimates: "reaching an average maximum density of approximately 181.6 Wolbachia genomes to 1 Drosophila genome." It is not explained how these estimates are calculated. Again, to calculate Wolbachia to host genomes ratios using the Pfaffl or delta-delta Ct methods a calibrator sample with known Wolbachia to host genomes ratio should have been used, which is not mentioned in the paper.
5) The authors characterize a new pathogenic wMel strain, determine that it has low Octomom copy numbers, and use this as an argument for the Octomom copy number being unrelated to pathology. Postulating that amplification of Octomom causes wMelPop pathogenicity does not imply that all pathogenic wMel variants or Wolbachia strains have Octomom amplification. Therefore, this result is irrelevant for the discussion on wMelPop pathogenicity. 6) Finally, in the discussion the authors are not accurate in the description of our previous experiments. They state that in our setup we selected for wMelPop pathogenesis: "While both selected for increased or decreased wMelPop pathology for a minimum of 14 generations, based on high/low Octomom copy number [22] or on survival [30],. . .", and imply that therefore the phenotypes we observed are due to genetic differences in the host. We did not select for increased or decreased wMelPop pathology. We selected strictly for Octomom copy number. Therefore, we controlled this number before determining pathology, hence our claim of  [4]. Both datasets include flies raised and kept at 25˚C. The script (S1 Text) and data (S1 and S2 Datasets) for the simulation are provided as supplementary data. The black line indicates the average Octomom copy number of all the simulated samples. The smooth conditional mean is included for illustrative purposes. causality. The assays were performed at several generations, starting at generation 3 (S2 Table in Chrostek and Teixeira [4]). Based on this protocol, it would be improbable to select for host mutations that, both, change host lifespan and correlate with the Octomom copy number-to the point that selection for one Octomom copy number completely reverts the phenotype. Moreover, we performed the selection experiment twice, in two different host genetic backgrounds, which renders the selection on the host even more unlikely.
In summary: 1. Octomom copy number determining pathogenicity is compatible with the previously known fact that temperature influences wMelPop pathogenicity.
2. That both factors determine pathogenicity does not imply that they influence each other.

Simulation of Octomom copy number variation over time in pools of flies from a heterogeneous population
The data used for this simulation were previously published in [4]. Wolbachia levels data were published as Fig. 3C (Dataset S6) and survival data as Fig. S5G (Dataset S4). All flies used in these experiments were raised and kept at 25˚C. In this simulation the Octomom copy number of wMelPop is assumed to be fixed throughout time and equal to two or twelve. The initial simulated population of flies was composed of 50% flies carrying wMelPop with two Octomom copies per genome and 50% flies carrying wMelPop with twelve Octomom copies. The proportion of these flies in the population at different time points was calculated from the survival data. Proportion of flies carrying wMelPop with two Octomom copies in the population at a given time point is equal to percentage of these flies surviving at that time point divided by the sum of the percentages of survival of flies carrying wMelPop with either Octomom copy number. Proportion of flies carrying wMelPop with twelve Octomom copies is equal to one minus the proportion of flies carrying wMelPop with two copies of Octomom. Thirty samples of five flies per time point were sampled in the simulation. The odds of sampling a fly carrying two or twelve copies of Octomom, at each time point, correspond to the proportion of flies carrying wMelPop with either Octomom copy number.
The levels of Wolbachia in flies carrying wMelPop with either Octomom copy number, at a given time point, was calculated from the statistics of the linear regression analysis of Wolbachia levels. For each wMelPop set a log-linear regression was performed using lm in R. The estimated intercepts and slopes were used to calculate levels of Wolbachia at each time point.
For each sample of five flies the Octomom copy number estimate is equal to the sum of the products of Wolbachia levels times Octomom copy number for each fly, divided by the sum of the Wolbachia levels estimates for each fly.
The script to simulate these samples was written in R [10]. Due to experimental variation the survival of flies carrying wMelPop with two Octomom copies was slightly lower that flies carrying wMelPop with twelve copies during the first days of the assay. To test the influence of this variation in the outcome of the simulation, survival data can be manually edited to ensure higher (100%) survival of flies carrying wMelPop with two Octomom copies in the initial time points of the simulation. This change does not affect the clear initial increase in average Octomom copy number with time. A suggestion on how to perform this alteration is embedded in the script as a comment.