Table 1.
VOC model | Gene | KO | Produce (p) or consume (c) | Variable (enzyme/VOC efflux/physiochemical) | VIP | PLSR weight | |
---|---|---|---|---|---|---|---|
w1 | w2 | ||||||
Acetate | na | na | na | acetone-C2 efflux | 1.51 | 0.61 | 0.28 |
ACH1 | K01067 | p | acetyl-CoA hydrolase | 1.21 | 0.48 | 0.15 | |
aarC | K18118 | c | succinyl-CoA:acetate CoA-transferase | 1.05 | 0.21 | −0.23 | |
na | na | na | soil moisture (%) | 1.02 | −0.42 | – | |
poxL | K00158 | p | pyruvate oxidase | 0.92 | 0.17 | −0.53 | |
poxB | K00156 | p | PDH-quinone | 0.90 | 0.18 | −0.50 | |
acyP | K01512 | p | acylphosphatase | 0.86 | – | 0.28 | |
acs | K01895 | c | acetyl-CoA synthase | 0.82 | −0.21 | 0.12 | |
pct | K01026 | p | propionate CoA-transferase | 0.82 | −0.28 | – | |
ackA | K00925 | c | acetate kinase | 0.75 | – | 0.46 | |
R2- total 0.91 (6 comp.) | 0.51 | 0.20 | |||||
Acetone | na | na | na | acetate-C2 efflux | 1.36 | 0.61 | – |
adc | K01574 | p | acetoacetate decarboxylase | 1.05 | 0.47 | – | |
na | na | na | soil moisture (%) | 1.00 | −0.44 | – | |
adh1 | K18382 | c | NAD+-dependent secondary alcohol dehydrogenase | 0.80 | 0.36 | – | |
acmA | K18371 | c | acetone monooxygenase | 0.64 | −0.29 | – | |
R2- total 0.76 (1 comp.) | 0.76 | – | |||||
Diacetyl+ | na | na | na | soil moisture (%) | 1.12 | – | 0.99 |
butA | K03366 | c | meso-butanediol dehydrogenase | 0.86 | –0.99 | – | |
R2- total 0.23 (2 comp.) | 0.15 | 0.08 |
PLSR models were created with single response variables (fluxes for each of the 13C-enriched VOCs) and predictor variables that included expression of genes known to produce or consume each 13C-enriched VOC and soil moisture (%). The VIP values represent the degree of importance of each predictor variable on the response variable. The PLSR weights (w) depict the magnitude and direction of contribution of each predictor variable (+ or − relationship) on the response variable along the first and/or second component (w1 and w2), with a dash indicating a lack of contribution to that component. Genes or physicochemical properties that were determined to not contribute to each model are not shown (see Methods). 13C-enriched VOC fluxes and gene expression data are from three collars that received 13C2-pyruvate at 0, 6 and 48 h post pyruvate injection during pre-drought and drought conditions (n = 18). The total percentage of variance explained (R2) from predictor variables for all components (comp., shown in parentheses) in the optimized model is shown in the bottom row for each VOC model in bold. The R2 broken down by the first and second components are shown below w1 and w2, respectively, in bold. Acetone-C2 efflux was included in the model for acetate-C2, and acetate-C2 efflux was included in the model for acetone-C2 due to the high correlation between acetate-C2 and acetone-C2.
na, not applicable; +, C4H6O2 represents diacetyl and/or one other unidentifiable compound.