Derive the respiratory-chain kcats from the data, and compare with the Xia et al. (2022) series - #15

Summary

Rebuilds this study's results with respiratory-chain kcats derived from the data and with enzyme complexes treated as one unit. Then runs the same analysis, for comparison, on the glucose-limited chemostat series of Xia et al. (2022): eight conditions from μ = 0.027 to 0.379 h⁻¹, of which μ0.33 and μ0.38 secrete ethanol.

From proteomics to enzyme caps (overflow.complexes, overflow.flexibilize):

1. Complex levelling. Each complex gets one abundance: the median per copy over its measured subunits. Proteins that serve several complexes or also work alone keep their own measurement.
2. Minimum usage. A complex the model cannot run at the dilution rate for less than its measured abundance is raised to what it needs.
3. Reaching the dilution rate. One LP finds the least relaxation that reaches the dilution rate: it minimises the summed fold increases of the measured caps, one per complex, and lets the protein pool grow at 1000 times that price.
4. Measured rates. The same LP, run under the measured exchange rates, makes the least further relaxation that reproduces them.

Respiratory-chain kcats (overflow.respiratory_kcats, a new step): for complex III, complex IV and ATP synthase, the apparent turnover in each condition is the sampled flux (conventional model at the measured rates, no kcats) over the levelled complex abundance. The highest across a dataset's conditions becomes the kcat, which overflow.build --respiratory-kcats applies to every isozyme variant. No other kcat changes.

The condition that sets a kcat runs that complex at full capacity, so each derived kcat is a lower bound, and capacity usage of the respiratory chain is relative to the busiest condition of the same dataset.

Running the Xia series. The series is built with GAM 55.3, as this study, and with the measured rates held within 16% in the build and in sampling (8% and 5% in this study). Its measured O2 uptake exceeds what complete oxidation of its measured glucose consumes by 7–46%, and holding growth, glucose, CO2 and O2 together needs a band of up to 15.4% (μ0.15). The series' μ = 0.044 condition is not included.

Files

• results/: this study, rebuilt (sampling unchanged). The build summary gains the modelled ethanol rate.
• literature/xia2022/:
  • the series in the pipeline's formats (data/, scripts/prepare_data.py, with Ptot from Supplementary Table 6);
  • a wrapper that runs any step on it (scripts/run_step.py);
  • the outputs of its run that the comparison reads, with its own derived kcats;
  • comparison/, with six figures and six tables covering both datasets.
• overflow.run_sampling --rate-tolerance sets the band around the measured rates in sampling.

Results

The respiratory limit of a condition is the highest glucose uptake its measured proteome can respire completely at the condition's growth rate. Above it, the model has to secrete overflow carbon. Caps the build raised, in any of its steps, show that the measurement understates what the condition needs and are lifted for this analysis.

Glucose and ethanol in mmol/gDW/h. Respiratory-chain usage: enzyme mass used over the measured capacity of its enzymes. CN: this study; μ: Xia et al. (2022).

• Respiratory-chain usage at D = 0.1 (CN4 20%, μ0.10 34%). The difference is not in the respiratory flux, which is the same in both: complex III carries 5.6 and 5.5 mmol/gDW/h. The Xia series measures more of each complex at μ0.10 than this study does at CN4: 2.3× complex III, 1.4× complex IV and 2.2× ATP synthase. Its complexes therefore turn over less than half as fast, but usage is expressed against each dataset's own busiest condition, and the busiest Xia condition (μ0.38) turns over 3.0–3.6 times slower than CN75. With this study's kcats applied to the Xia series, μ0.10 uses 10.6% of its respiratory chain (0.47 of 4.42 mg/gDW), against 19.7% at CN4 (0.46 of 2.32 mg/gDW): the same enzyme mass over twice the measured capacity.
• hGR stays the single condition at D = 0.29: it is the average of chemostats at 0.26 and 0.32 h⁻¹, which are not available separately.
• The last condition before overflow: CN22 and μ0.28. Both respire all their glucose and are the fastest of their series to do so.
  • CN22 takes up 1.48 of a 1.57 limit (94%). The limit is set by succinate dehydrogenase: raising its cap 1.2-fold raises the limit by 11%, while raising the whole respiratory chain 1000-fold leaves it unchanged. CN22 also secretes formate (0.05 mmol/gDW/h), the one condition without ethanol that has a C1 overflow.
  • μ0.28 takes up 3.14 of a 4.32 limit (73%). No single complex sets that limit: raising the respiratory chain or the TCA cycle 1000-fold leaves it unchanged, lifting RPE1 or glycolysis raises it by 5.4%, and raising every measured cap 1.2-fold raises it by 16%. Its respiratory chain turns over at 0.74–0.76 of the series' maximum. The series does not report formate, so a C1 overflow cannot be compared.
• When overflow starts. Conditions without ethanol take up 29–94% of their limit, and CN75, hGR and μ0.38 take up 1.3–1.5 times theirs. Three conditions sit at their limit, within 6% and inside the band around the measured rates: CN22 (94%) without ethanol, and CN38 (94%) and μ0.33 (96%) with it. In this study the measured onset of overflow lies at a glucose uptake of 2.0 ± 0.3, where the limits of CN38 (2.13) and CN75 (2.19) fall. At fast growth, μ0.33 and μ0.38 share one limit, 5.06, set by complex IV; μ0.33 takes up just under what it can respire while μ0.38 takes up a third more. As each dataset's kcats are its own highest apparent turnover, this is a consistency check rather than an independent test.
• ATP supply. The respiratory share of ATP is 84–87% in every condition without ethanol and at CN38, and falls as ethanol rises: 81% at CN75 and μ0.33, 74% at hGR and 73.5% at μ0.38.

The figures and tables are in literature/xia2022/comparison/.