Growth medium affects secretion titer
An optimal medium for protein production is composed of nutrients that maximize cell density and protein production per cell, or specific productivity. In our prior work, we found that overexpressing the master regulator HilA increased bulk secretion titer by activating secretion in over 90% of the cell population in conditions that maximized cell density [10]. We observed an increase in specific productivity, however, only when we overexpressed HilA in LB-IM media (Lysogeny Broth containing 17 g/L NaCl) [8]. This finding highlights a unique challenge for protein production via the T3SS—secretion titer is increased in high-salt conditions that limit cell density. As a result, an ideal growth medium for protein production via the T3SS will need to strike the optimal balance between maximizing cell density and specific productivity.
Though numerous environmental inputs to the T3SS are documented [11–13], they were not characterized in the context of our engineered system. Thus, to expand our knowledge of medium components that affect secretion titer, we began by simply measuring expression and secretion titer of a model protein in the common bacterial growth media 2X YT and terrific broth (TB). The model protein, or protein of interest (POI), was the soluble catalytic DH domain from the human protein intersectin-1L fused C-terminal to the T3SS secretion tag SptP to facilitate secretion [7, 21]. SptP-DH-2xFLAG-6xHis was expressed from an “export vector” under the control of the native SPI-1 promoter Psic. It was co-transformed with a “secretion activation vector” containing a cassette for inducible HilA overexpression into S. enterica Typhimurium ASTE13. This two-plasmid system enabled T3SS activation and POI expression with a single induction event [7, 10]. SptP-DH secretion titers were measured relative to LB-L using semi-quantitative western blotting.
Bulk secretion titer increased 3-fold in 2X YT and 5.5-fold in TB relative to LB-L (Fig. 1A). 2X YT and TB increased expression of SptP-DH 2.5-fold and 4.5-fold (Fig. 1C, Supplementary Fig. 1) suggesting that an increase in secretion titer correlates with an increase in expression of the target protein. The increase in secretion titer closely matched the increase in expression in 2X YT, but the relative increase in secretion titer was 20% higher than the relative increase in SptP-DH expression in TB (Fig. 1C). This discrepancy suggested that some components in TB caused a higher percentage of expressed protein to be secreted. We will henceforth refer to the fraction of expressed protein that is secreted as “secretion efficiency”.
TB contains two defined components that are not present in LB-L or 2X YT: 0.4% w/v glycerol and 89 mM potassium phosphate (Fig. 1A). We hypothesized that adding those components to LB-L could recapitulate the effects observed in TB. Surprisingly, separate addition of those components to LB-L revealed a competing dynamic—glycerol repressed secretion, while phosphate buffer promoted secretion. Combining the two components in LB-L, however, had a synergistic effect, matching the increase in secretion titer and efficiency observed in TB (p < 0.001 for separate addition, p = 0.6 for LB-L with glycerol and phosphate)(Fig. 1B-C).
Non-ionic carbon sources decrease secretion titer
SPI-1 T3SS transcriptional activity is repressed in the presence of glucose [23], so the discovery that glycerol negatively impacted secretion titer led us to screen a panel of carbon sources in LB-L. Our goal was to determine the optimal carbon source for maximum secretion titer, as well as to develop an understanding of how these carbon sources might be affecting secretion titer through heterologous protein expression and T3SS expression. Therefore, we evaluated SptP-DH secretion titer, expression, and T3SS transcriptional activity for each carbon source in LB-L. Secretion titer and expression were measured relative to LB-L using semi-quantitative western blotting. Transcriptional activity over time was measured by performing flow cytometry on strains with GFP integrated into one of the SPI-1 inv, prg, or sic loci. The inv, prg, and sic operons encode regulatory, structural, and natively secreted proteins, respectively [24].
SptP-DH secretion titer decreased relative to no added carbon source in the presence of all carbon sources tested (Fig. 2A), and SptP-DH expression decreased in the presence of all carbon sources except glycerol (Fig. 2B). Cell density was approximately constant across all conditions (Mean OD600nm ± S.D., Supplementary Table 5), indicating that the added carbon sources altered SptP-DH expression and secretion titer by affecting T3SS function. T3SS transcriptional activity, as assessed by flow cytometry, supported this reasoning (Fig. 2C, Supplementary Fig. 4). Glucose had a strong negative effect on expression and secretion titer and caused lower transcriptional activity that ceased earlier than all other carbon sources. Glycerol, conversely, caused the smallest decrease in secretion titer and had no effect on expression relative to no added carbon source, and transcriptional activity in that condition was higher than and showed similar dynamics to no added carbon source.
As a visual examination of the flow cytometry data suggests, the maximum transcriptional activity, represented by maximum fluorescence, and the length of transcriptional activation correlated with SptP-DH expression and secretion titer (Supplementary Fig. 4A-B). Flow cytometry measures transcriptional activity on a per cell basis, so its results were compared to SptP-DH expression and secretion titer per cell. Spearman correlations revealed that SptP-DH expression per cell correlated with both the level of transcriptional activation at T3SS loci (Supplementary Fig. 4B), and the duration of T3SS activation (Supplementary Fig. 4A). SptP-DH secretion titer per cell correlated only with the duration of T3SS activation. The strength of the correlations varied by locus and parameter—invE activity was least predictive for either expression or secretion titer, while maximum activation and length of activation at sipC was most predictive. Taken together, these results suggest that prolonged T3SS activity is a critical factor in maximizing secretion titer, but that alone does not explain the observed decreases in secretion titer. That effect might also be explained by acidification of the medium (Supplementary Table 5), an environmental condition known to repress secretion via the SPI-1 T3SS [12, 13].
Secretion titer increases with ionic content
Phosphate is a unique buffer species because in addition to providing buffer capacity, it contributes significantly to the ionic content of the medium. To decouple the effects of buffering and increased ionic content, we compared secretion titer in media containing a base of 10 g/L tryptone and 5 g/L yeast extract supplemented with a range of concentrations of potassium phosphate, 3-(N-morpholino)propanesulfonic acid (MOPS), sodium chloride, or MOPS supplemented with sodium chloride (Supplementary Table 6). We selected MOPS as an alternative buffer species because it is the main buffer component in a defined medium explicitly designed for S. enterica cultivation [25], it has a minimal contribution to ionic strength, and it is one of Good’s buffers [26]. To mimic the simultaneous contributions of buffering and ionic content inherent to phosphate, we supplemented MOPS with sodium chloride. Finally, to control for changes in ionic content in the absence of a buffering agent, we tested sodium chloride alone. We monitored ionic content by measuring conductivity. The concentrations of sodium chloride with and without MOPS were chosen to match the conductivities of the specified concentrations of potassium phosphate. SptP-DH expression and secretion titer were compared to that in LB-L with no additives using semi-quantitative western blotting.
Expression and secretion titer increased with buffer and salt concentration (Fig. 3A, Supplementary Fig. 5). Plotting relative expression and secretion titer against conductivity revealed a linear relationship, irrespective of species (Fig. 3B, Supplementary Fig. 5). The correlation was stronger for secretion titer (R2 = 0.8) than expression (R2 = 0.7). The highest concentrations of buffers and salts resulted in only a twofold relative increase in expression and secretion titer compared to the nearly six-fold increase observed in media with a combination of phosphate and glycerol (Fig. 1B), indicating that high ionic content is but one component of an optimal medium formulation.
Carbon sources and buffers have a synergistic effect on expression and secretion titer
Independent addition of buffers, salts and carbon sources showed that secretion titer increases with ionic content and decreases with carbon sources that cause acidification of the culture. If acidification of the culture was the sole cause of decreased secretion titer, addition of a buffer would return secretion titer to levels at least equivalent to no added carbon source. Potassium phosphate and glycerol in LB-L had a synergistic effect, however, suggesting that high ionic strength, buffering, and added carbon sources are a critical combination for increased secretion titer.
To determine if the synergistic effect could be a result of any combination of carbon source and buffer at high ionic strength or whether the effect was specific to the combination of potassium phosphate and glycerol, we measured secretion of SptP-DH in LB media containing potassium phosphate, MOPS, sodium chloride, and MOPS plus sodium chloride with added glucose or glycerol. We chose these two carbon sources because glucose had a strong negative effect on SptP-DH expression and secretion titer, while glycerol alone had no effect on SptP-DH expression, a moderate negative effect on secretion titer, and in combination with phosphate buffer caused a significant increase in expression and secretion titer (Fig. 1, Supplementary Fig. 1). In addition to representing the spectrum of effects observed in this study, glucose and glycerol are prominent carbon sources in industry and in research studies. The potassium phosphate and MOPS concentrations were 90 mM to match the concentration of potassium phosphate in TB, and the concentrations of sodium chloride with and without MOPS buffer were chosen to approximate the conductivity of 90 mM potassium phosphate (Supplementary Table 7). Expression and secretion titer of SptP-DH were measured in comparison to LB-L with no added carbon source using semi-quantitative western blotting.
Buffering was essential to maintain secretion in the presence of glucose and glycerol, and the combination of buffering and increased ionic content was necessary to increase expression and secretion titer (Fig. 4A). LB-MOPS with glucose or glycerol produced secretion titers similar to LB-L with no added carbon source, while secretion titer decreased in LB-NaCl with glucose or glycerol. The decrease in secretion titer in LB-NaCl with glucose or glycerol was accompanied by acidification of the extracellular environment, as observed in Fig. 2 (Supplementary Table 7). Both LB-(KH2PO4/K2HPO4) and LB-(MOPS + NaCl) with added glucose or glycerol provided at least a threefold increase in secretion titer. LB-(MOPS + NaCl) appeared to increase secretion titer primarily by increasing expression, while LB-(KH2PO4/K2HPO4) increased expression but also increased secretion efficiency (Fig. 4B, Supplementary Fig. 6). LB-(KH2PO4/K2HPO4) with glycerol had a specific beneficial effect—the increase in secretion titer was statistically higher than the other combinations of carbon source, buffer, and salt (p < 0.05, Fig. 4).
Carbon sources and buffers affect T3SS transcriptional activity
Flow cytometry performed on ASTE13 strains containing transcriptional fusions of GFP at the sic, prg, and inv loci in the media listed in Supplementary Table 7 showed that SPI-1 transcriptional activity varied with medium composition (Fig. 5). Transcriptional activity increased with conductivity for no added carbon source. In media containing glucose or glycerol, the trends were more complex. Transcriptional activity was repressed in LB-NaCl and highest in LB-(MOPS + NaCl). LB-(KH2PO4/K2HPO4) with glucose or glycerol caused higher transcriptional activity at the prg operon than either the inv or sic operons. Relative to media with no added carbon source, glucose and glycerol also prolonged transcriptional activation in buffered media while causing an early decrease in unbuffered media. Histograms for the data in Fig. 5 are available in Supplementary Figs. 7–9.
We again compared maximum transcriptional activity and duration of T3SS activation to expression and secretion titer on a per cell basis, as we did for the data in Fig. 2. Expression per cell corresponded more strongly with maximum transcriptional activity (Supplementary Fig. 10A), while secretion per cell corresponded more strongly with prolonged activation of T3SS loci (Supplementary Fig. 10B). Again, the strength of the correlations varied by locus and parameter. Expression and secretion per cell showed the strongest correlation with maximum mean fluorescence at the prgH locus, and length of activation at the sipC locus had the strongest correlation with secretion and expression per cell. Neither transcriptional activity nor dynamics at the invE locus correlated well with expression or secretion titer per cell.
In prior work, we showed that knocking out a protein in the SPI-1 T3SS tip complex, SipD, increased secretion titer twofold in strains with hilA overexpression [27]. An ideal T3SS production platform would combine all features that increase secretion titer, so we evaluated if ∆sipD was additive with an optimized medium. To determine if the effect was general, we selected a variety of test proteins in addition to DH: magainin-1 (MAG1), an antimicrobial peptide; 14B7*, an scFv against the protective antigen of the anthrax toxin [28, 29]; and recombinant human growth hormone (rhGH, mature somatropin). All proteins were cloned in the format SptP-POI-2xFLAG-6xHis and secreted in ASTE13 WT and ∆sipD strains. We added 90 mM NaCl to LB-KH2PO4/K2HPO4 with glycerol to match the conductivity of LB-L supplemented with glycerol and potassium phosphate (Supplementary Table 7). We will refer to this medium as LB-ES for “enhanced secretion”.
The ∆sipD strain improvement and the optimized medium were indeed additive for all proteins tested (Fig. 6, Supplementary Fig. 11). Secretion titer increased by varying amounts, but the minimum increase provided by the combination of ∆sipD and LB-ES was six-fold above a WT strain in LB-L. Total protein expression showed a different pattern from secretion titer. SptP-MAG1 and SptP-DH followed similar expression patterns, and the apparent effect of ∆sipD and LB-ES was to increase secretion efficiency for these proteins. SptP-rhGH and SptP-14B7* expression increased by a surprising eight- and fourteen-fold, however. SptP-rhGH expression increased by the same fraction as secretion titer in each condition, suggesting that secretion efficiency remained constant. SptP-14B7* secretion titer increased by a much smaller margin than expression in both WT and ∆sipD strains with LB-ES. We hypothesize that the discrepancy was caused by loss of expressed protein to insoluble aggregates, preventing secretion.