Comments from curator
Phenotyping data come from the authors'
Dryad deposit (
doi:10.5061/dryad.js408), file
allnutrients.csv ("Line means for all phenotypes from study"; "bacterial load is natural log load per fly ... Data from linear models as described in manuscript"), cited in the Data Availability statement of the paper (
journal page, also on
PMC). Only these line means were published (no per-sample data), so the study has a summary dataset only.
Design: each DGRP line was split and reared in parallel on two diets that differ only in their glucose content (5% Brewer's yeast and 1% agar, with 10% glucose for the high-glucose diet or 2.5% glucose for the low-glucose diet) for at least three generations, and the experimental flies were reared egg to adult on that diet. Males 3-6 days old were pricked in the thorax with a needle dipped in a suspension of
Providencia rettgeri (strain Dmel, A
600 = 1.0; about 1,000 bacteria per fly) and kept at 24 °C (12:12 h light:dark). About 24 h later, they were homogenised in groups of 3 and plated, and the colony-forming units (CFU) were counted; the load is the natural log of the CFU per fly. The infections were done in 3 blocks per diet by 3 researchers, giving 6-9 samples (18-27 flies) per line and diet: 1,429 samples (4,287 flies) on the low-glucose diet and 1,396 samples (4,188 flies) on the high-glucose diet, from 172 lines (the exact set of lines varied slightly among blocks). The authors estimated line means with mixed linear models (SAS;
Wolbachia status, diet, line within
Wolbachia status, infector, plater, block within diet and diet × line). All flies were males.
Summary dataset (sex M, 152 lines; values copied as written in allnutrients.csv):
-
mn_PrettLoad_HiGluc and
mn_PrettLoad_LoGluc: line means (authors' mixed model) of the
P. rettgeri load, as the natural log of the CFU per fly, 24 h after infection, on the high-glucose (151 lines) and low-glucose (151 lines) diets. Higher values mean poorer resistance to infection. The correlation between the two diets is r = 0.69 (paper: r = 0.69), and loads are 2.32 times higher on the high-glucose diet (paper: about 2.4).
-
mn_PrettLoad_Pooled: the file's "load.main" column (152 lines), the line estimate of the authors' models with the data of both diets combined (the "pooled" analyses and GWAS of the paper). It is not the average of the two diet means (on average 0.47 lower).
-
Nutritional indices of uninfected males (3 pools of 10 males per line and diet): free glucose (mn_Glucose_*), glycogen (mn_Glycogen_*), free glycerol (mn_Glycerol_*), triglycerides (mn_TAG_*), soluble protein (mn_Protein_*) and wet mass (mn_WetMass_*), each on the high-glucose (_HiGluc, 147 lines) and low-glucose (_LoGluc, 150 lines) diets and with both diets (_Pooled, the file's ".main" columns, 152 lines). They are line means of the abundance of each metabolite per fly (authors' mixed model). The Dryad README gives "mg/fly" for the metabolites and "mg" for the weight, but the values (e.g. a wet mass of 589-875 per fly, like the "meanweight(ug)" column of the G3 paper) are in µg, the unit used here. The genotype-by-diet interaction of the paper (and its SNP-by-diet GWAS) concerns the difference between the diets: a line's sensitivity to dietary glucose can be derived as the _HiGluc minus the _LoGluc value. The authors' GWAS used the individual samples, not these line means.
The paper infected 172 lines ("the exact number and composition of lines varied slightly among replicate blocks"), but allnutrients.csv has line means for 152 lines only (the same lines as G3 Table S2, i.e. the lines with nutritional indices); the loads of the other lines were not published. Lines are numbered as RAL lines in the file (101 = DGRP_101); all are DGRP lines of DGRPool. The file's last row ("Line" followed by NA) and its "Line2" column (the line name again, as line_101) are not uploaded.
Nutritional indices and DGRPool study 32: the companion paper Unckless, Rottschaefer & Lazzaro 2015 (
G3 5:417-425,
doi:10.1534/g3.114.016477), which is
DGRPool study 32 (from its Table S2), reports the same assays of the same lines, but as metabolite per mg of fly (least-squares means of a fixed-effect model). The values of allnutrients.csv are per fly (mixed model, as in the Methods of this paper), so they differ from study 32 for every line (correlation between the two 0.76-0.93 for glucose, glycogen, glycerol and protein; about 0.99 for wet mass, which is close but not equal; and -0.53 to 0.24 for triglycerides). This paper's analyses use the per-fly values: the correlation between the load and free glucose on the high-glucose diet is r = -0.176 (P = 0.033) with allnutrients.csv (paper: r = -0.18, P = 0.033), but r = -0.142 (P = 0.086) with study 32. They are therefore uploaded here too. The principal components of the nutritional indices (Table 3) are not in the Dryad file and are not uploaded.
Not uploaded: the GWAS results (4 large files of the Dryad deposit), the RNAi knockdowns of the mapped genes (S1 Table; not DGRP lines) and the per-line boxplots of S1 Fig (a figure).
DGRPool study 16 (Howick & Lazzaro 2017) measured
P. rettgeri load in the DGRP in a separate experiment.