Comments from curator
Phenotyping data comes from the authors'
figshare dataset
10.6084/m9.figshare.21657503, cited in the Data Availability statement of the paper (
journal page), archive
Data and Scripts.7z:
-
Data_DGRP.csv: viral load of each vial (block, virus and inoculum, DGRP line,
pastrel allele, fold change), used for the raw dataset.
-
Data_DGRP_Deltas.csv: the authors' change in viral load during coinfection, per pair of vials, used for the checks.
The list of lines (S1 Table) and the number of vials per line and inoculum (S6 Table) are in the
Supporting information of the paper.
Males of 25 DGRP lines (stocks from J. Day and F. Jiggins; 15 lines with the resistant G allele and 10 with the susceptible A allele of the
pastrel A2469G SNP) were reared at a controlled larval density, collected at 0-1 day and aged 7-8 days on cornmeal food. They were inoculated under CO
2 by septic pin prick in the pleural suture with one of three inocula:
Drosophila C virus (DCV, isolate DCV-C) or cricket paralysis virus (CrPV), each mixed 1:1 with Ringer's solution (single infection), or DCV and CrPV mixed 1:1 (coinfection, same dose of each virus). The flies were kept in vials of 7 to 20 males (mean 12.7) at 22°C and snap frozen 2 days (± 2 h) after inoculation. Each vial gives one RNA sample, in which the viral load was measured by qRT-PCR (two technical replicates, corrected for plate effects) relative to
RpL32 by the ΔΔC
T method: ΔC
T = C
T virus − C
T RpL32, ΔΔC
T = 40 − ΔC
T, fold change = 2
ΔΔCT. The experiment was run in 3 blocks, with about 4 vials per line, inoculum and block.
The raw dataset has
one row per coinfected vial, with the single-infection vials of the same line and block stacked on the same rows (300 rows, 12 per line, sex M: the 295 coinfected vials, plus 5 rows that hold only single-infection vials):
-
DCV_LogLoad_Coinf and
CrPV_LogLoad_Coinf: DCV and CrPV loads of the same coinfected vial. Both viruses were measured on the same RNA sample: in each block, the DCV and CrPV coinfection rows of the authors' file list the same sequence of lines, and the two loads of a row deviate from their line mean together (within-line correlation r = 0.47, against 0.00 for vials paired at random), as expected from their shared
RpL32 normalisation. Two coinfected vials have no CrPV value (DGRP_437 in block 2, DGRP_386 in block 3).
-
DCV_LogLoad_Single and
CrPV_LogLoad_Single: loads of vials inoculated with DCV alone or CrPV alone. These are other vials of the same line and block, stacked on the rows.
-
ExpBlock (obsolete): the experimental block (1-3) of all the vials of the row, a random effect in the authors' models. DCV loads differ markedly between blocks (block 3 in particular), as noted in the paper.
The loads are
log10 of the authors' fold change (column Foldchange), written at full precision. The paper analyses and plots only this scale, and the fold changes range from 17 to 2.6 × 10
9, so means on the fold scale would be dominated by single vials. The DGRPool line means are therefore on the scale of the line means of Fig 1. The
pastrel allele of the lines (coded C/T in the file for the G/A alleles of the paper) is a DGRP genotype and is not uploaded.
Change in viral load during coinfection (model 2 of the paper, Table 1 "Change", Fig 1): the authors computed it per pair of vials as log10 coinfection load − log10 single-infection load, pairing the k-th single-infection vial with the k-th coinfected vial of the same line and block (Data_DGRP_Deltas.csv). The single-infection vials were placed on the rows so that, on each row, DCV_LogLoad_Coinf − DCV_LogLoad_Single and CrPV_LogLoad_Coinf − CrPV_LogLoad_Single are exactly the authors' values (291 pairs for each virus). It is not uploaded as a separate phenotype.
Checks: the file has 1,181 loads from 888 vials (295 with DCV alone, 298 with CrPV alone, 295 coinfected with a DCV value and 293 with a CrPV value), for the 25 lines of S1 Table. Means and heritabilities computed from these values (one-way ANOVA within blocks, h
2 = half the line variance over half the line variance plus the residual variance) match Table 1: CrPV single infection 4.67 and 0.12 (Table 1: 4.68 and 0.13), CrPV coinfection 4.27 and 0.12 (4.28 and 0.13), DCV single infection 6.18 and 0.31 (6.17 and 0.32), DCV coinfection 6.63 and 0.27 (6.62 and 0.28). The lines with large changes match the text: CrPV decreased about 14-fold in DGRP_320 and DGRP_852, and DCV increased 37-, 129- and 142-fold in DGRP_892, DGRP_386 and DGRP_738.
Of note:
- S6 Table gives 12 CrPV vials for DGRP_073 (11 in the file), and 11 coinfected vials for DGRP_386 and DGRP_437 (12 DCV values in the file, one of them without a CrPV value).
- The Table 1 means of the change (DCV 0.33, CrPV −0.27) are model estimates: the means of the authors' per-vial changes are 0.46 and −0.42. The authors' script (Models.R) fits model 2 on log10 of these changes, which are already log10 differences, so their negative values (181 of 291 for CrPV, 117 for DCV) become missing; this may explain part of the difference.
- The paper reports 8,618 inoculated DGRP flies, i.e. 9.7 flies per vial for the 888 vials of the file (887 in S6 Table), while its methods give a mean of 12.7 flies per vial (perhaps over both experiments). The number of flies of each vial is not in the file.
Not uploaded: the across-species experiment (47 Drosophilidae species, Data_Drosophilidae.csv), which has no DGRP lines.