SUPPLEMENTARY TEXT

Each trait was assigned to a general disease type or a non-disease trait using information about studies provided in the dataset. Sometimes the same GWAS variant was found associated with more than one disease belonging to same disease type, potentially inflating statistics. To avoid such cases, yet to recognize the value of variants found associated with the same disease type in independent studies, we selected entries with the unique combination of the following fields: (1) name of the variant, (2) PubMed record ID and (3) disease type (Supplementary Table 1). Therefore, the same GWAS variant was counted more than once if it was associated with different disease types or with the same disease type but in more than one independent publication. Overlap with genomic annotations was done in a hierarchical fashion: a SNP overlapping multiple genomic annotations was assigned to the annotation with a higher priority. The order of priorities were (from highest to lowest): CDS's, UTR's, promoters of known genes, introns of known genes, lincRNAs (exons and introns), vlincRNAs and remaining Intergenic space, where CDS's (portions of genes that code proteins), UTR's (untranslated gene regions on each side of CDS) and introns were taken from UCSC Genes table

regarding other growth phases.We observed consistently slower growth in liquid culture, and smaller colony sizes of ΔrsmA compared to ΔwspF, but the slight changes in growth during log phase do not fully account for this difference.We therefore monitored the growth of these strains throughout their culturing.Due to the highly auto-aggregative nature of ΔrsmA, accurately measuring the optical density of the liquid cultures was not possible.We therefore used a ΔrsmA Δpel Δpsl strain, which does not auto-aggregate (data not shown).
Consistent with the previous report, only a slight difference was observed in the doubling time of ΔrsmA Δpel Δpsl mutant compared to PAO1 wild-type and Δpel Δpsl mutant (Fig. S7).However, the ΔrsmA strain was defective in growth during lag phase, delaying its arrival at log and stationary phases by 2-3 h compared to wild-type strains.Mulcahy et al. (2008) reported that PAO1 wild-type out-competed ΔrsmA by 4 hours of co-culture.This could be due to the extended lag phase of ΔrsmA.Pessi et al. (2001) reported that ΔrsmA mutants exhibited enhanced production of the extracellular pigment pyocyanin.We, however, observed a complete lack of pigment production (pyocyanin and pyoverdine) in ΔrsmA compared to PAO1 wild-type (Fig. S8).Furthermore, we observed an identical pigment defect phenotype in ΔwspF and inducible Psl over-expression strains, both of which are auto-aggregative in culture (data not shown).
Interestingly, the non-auto-aggregating strains ΔrsmA Δpsl and ΔrsmA Δpel Δpsl demonstrated wild-type levels of pigment production (Fig. S8), indicating that the pigment phenotype is likely linked to auto-aggregation and is not directly regulated by RsmA.
Secondary site mutation frequencies that give rise to suppressor mutants of ΔcsrA strains have been documented to be very high in other species (Altier et al., 2000;Timmermans and Van Melderen, 2009).We observed the appearance of suppressor mutants in the ΔrsmA background to be high especially in rich medium (50% -100% suppressor mutants in overnight cultures; Fig. S9).Defective growth may be one of the pleoitropic effects that favours the appearance of the suppressor mutants.These suppressors lost the rugose small colony variant (RSCV) morphologies (Fig. S9).
Interestingly, the suppressors had altered phenotypes in auto-aggregation in liquid cultures and flagella-dependent motility (data not shown).The suppressor mutants do not revert to the RSCV morphologies and are stable, as multiple passages failed to produce RSCVs Oligonucleotide primer Sequence (5' to 3')

Fig
Fig. S4.RsmA-His 6 eluted from the Ni-NTA column at high millimolar range of imidazole with high purity.Western blot analyses using antibodies raised against E. coli CsrA and nickel-activated derivative as His 6 motifspecific probe (HisProbe-HRP, Pierce) confirmed that the protein band on Coomassie-stained SDS polyacrylamide gel (top panel) is RsmA-His 6 .

Fig. S5 .
Fig. S5.RsmA-His 6 construct is fully active in vivo compared to wild-type RsmA.Over-expression of RsmA-His 6 showed identical translational psl activity compared to wildtype RsmA over-expression.RsmA-His 6 was able to complement ΔrsmA, and reduce psl translation lower than PAO1 level.The y-axis unit is described as β-galactosidase activity divided by total mg of protein from the cell lysates.VC = vector control (pUCP18).

Fig. S7 .
Fig. S7.ΔrsmA mutants are growth defective.ΔrsmA mutants are auto-aggregative in liquid culture, rendering optical density measurements inaccurate.We therefore compared the growth curve between wild-type and ΔrsmA in the Δpel Δpsl strain background, which do not aggregate in liquid.ΔrsmA mutants appear to have a similar doubling time in log phase and final cell density at stationary phase, but are defective in recovering from lag phase.Data sets were compiled from three independent cultures for each strain grown in VBMM at 37 o C.

Fig. S8 .
Fig. S8.Loss of pigment production in ΔrsmA is dependent on cellular auto-aggregation.Strains were grown in VBMM overnight at 37 o C. Photographs of the cultures indicate that non-aggregating strains ΔrsmA Δpsl and ΔrsmA Δpel Δpsl are fully proficient in pigment production at wild-type level, indicating that pigment biosynthesis is not under RsmA regulation, but is repressed when cells are auto-aggregating.

Fig. S9 .
Fig. S9.ΔrsmA suppressor mutants arise at high frequency during culture.ΔrsmA was grown overnight in LB broth and plated on VBMM + Congo Red.Black arrows indicate parental ΔrsmA RSCV colonies while white arrows indicate suppressor mutants (ΔrsmA rev) that have lost the RSCV properties on the plate.The ΔrsmA rev phenotype is stable, as multiple passages do not give rise to a RSCV phenotype (data not shown).