Engineering eukaryote-like regulatory circuits to expand artificial control mechanisms for metabolic engineering in Saccharomyces cerevisiae

Native transcription factor promoters express at moderate to low levels

A strong promoter, commonly used in overexpression of metabolic enzymes or primary cellular constitutive proteins, is often used to control the expression of artificial transcription factors (TFs) in synthetic circuits^15,35,37. However, in natural eukaryotic systems, TFs are generally not expressed at high levels³⁸. We wanted to explore whether we could achieve a better outcome by exploiting native TF promoters to mimic natural expression systems. We therefore characterised fourteen TF and regulatory protein promoters.

Yeast strains for promoter characterisation were obtained by transformation of an enhanced yeast green fluorescent protein (yEGFP) expression cassette under the control of promoter of interest via single-copy genome integration (Fig. 1a). Fluorescence was analysed in early exponential phase (when glucose was the carbon source) and in post-exponential phase (when ethanol was the carbon source).

We tested three base constructs using the TEF1 promoter to identify the best one to use for promoter comparison (Fig. 1a). In our previous work³⁹, we used a construct with a modified translation initiation region including a BamHI restriction enzyme site and a URA3 terminator (which also acts as a homologous recombination site to introduce the construct onto the genome). Insertion of a PGK1 terminator improved yEGFP fluorescence by >5-fold (Fig. 1b). removal of the BamHI site and reinstatement of the native translation sequence context (Fig. 1a, Construct 3) further improved yEGFP fluorescence by ~50%. This construct design was used to test regulatory protein promoters.

Compared to the TEF1 promoter (Fig. 1b, construct 3), the 14 tested endogenous regulatory protein promoters were weaker by 1–2 orders of magnitude (Fig. 1c). The promoters of the GCR1 and GCR2 genes, encoding transcriptional activators for glycolytic and ribosomal genes, were >39-fold weaker than the TEF1 promoter, and the promoter of the MIG1 gene, encoding an important repressor in the glucose repression signalling pathway, was further weaker. The promoters showing a similar or weaker strength include the promoters of GLN3 (encoding a transcriptional activator in nitrogen catabolite repression system), TOR1 (encoding a primary protein kinase in Target Of Rapamycin regulatory network), SNF1 (encoding a primary protein kinase in glucose repression regulatory networks), NRG1 (encoding a repressor mediating glucose repression), ROX1 (encoding a repressor in oxygen regulation), HAP4 (encoding a global regulator of respiration), and UPC2 (encoding a ergosterol-sensing activator). This shows that many primary components in natural regulatory networks are expressed at the levels much lower than the levels achievable from strong constitutive promoters.

Among the tested promoters, the promoters of YPK2 (encoding a protein kinase required for cell growth), ADR1 (encoding a transcriptional activator in glucose repression), and HAC1 (encoding a transcriptional activator regulating the unfolded protein response in endoplasmic reticulum) showed a moderate strength, around 10-fold lower than the TEF1 promoter. The promoter of DAL80 (encoding a negative repressor responsive to nitrogen levels) showed induced activities on ethanol (Fig. 1c, d), but not on sucrose (Fig. 1d). The DAL80 promoter could be further induced to a level higher the TEF1 promoter, when yeast was grown on urea, a poor nitrogen source (Fig. 1d). Due to the HAC1 showing a consistent expression output on the exponential growth phase and the ethanol-growth phase and its moderate activity, we chose it to control artificial TFs for the following modular characterisation.

Trans-activating modules have variable effects on transcriptional activity from different hybrid promotes

The trans-activating domain (TAD) binds to transcriptional co-regulator proteins to refine and determine the magnitude of the response. Ottoz and colleagues¹⁹ used a chimeric construct approach to characterise four TADs in yeast: VP16 (the alpha-gene-transactivating factor (α-TIF) from the alphaherpesviruses gene UL48 promoter), B112, B42, and the Gal4 activation domain. VP16A showed the strongest activation capacity¹⁹. To expand the pool of available elements, we characterised two more TAD modules, a ten-WD-repeat sequence (10*WD)⁴⁰ and a Gcn4p trans-activating domain (Gcn4 aa107-144; Gcn4^A). Mediators are evolutionarily conserved and form the coactivation complex generally required for eukaryotic RNA polymerase II-dependent transcription⁴¹. Recruiting mediators to upstream activation sequences in promoters may allow transcription, even under inhibitory conditions⁴². To better understanding transactivation effects of mediators, we included the tail mediator 3 (Med3)⁴² and the tail mediator 15 (Med15)⁴² in the comparison.

We expressed synthetic chimeric TFs consisting of an N-terminal zinc-finger (ZIF) DNA binding domain (Zif268), a linker harbouring a SV40 nuclear localisation sequence (NLS)⁴³, and TAD (or mediator). VP16^A was included as a reference. We selected the HAC1 promoter, which delivers moderate and stable expression during both the exponential and ethanol growth phases (Fig. 1c), to control the expression of synthetic TFs (Fig. 1a). This provides a TF expression system that mimics natural expression levels. We used a single-copy centromeric plasmid pRS414⁴⁴ to introduce TF expression cassettes into yeast. The reference was the empty vector plasmid pRS414.

Synthetic cis-regulatory reporter cassettes were constructed using yEGFP as the reporter gene and introduced into yeast via genome integration at the URA3 locus (Fig. 1a). For a better understanding of trans-activation effects of artificial TFs, either the weak core CYC1 promoter or the strong TEF1 were fused to a synthetic DNA sequence (truncated from the synthetic P15 promoter, published previously²⁰) including four Zif268 binding elements (Fig. 2a, b; P_{CYC1+4×[Z268]} and P_{TEF1+4×[Z268]}). Reference promoters were constructed by inserting the synthetic sequence with Z268 elements removed (P_CYC1(−) and P_TEF1(−)).

The cis-regulatory reporter cassettes and the trans-activator-expressing cassettes were sequentially transformed into the CEN.PK2-1C strain. The yEGFP fluorescence from cells in exponential and in ethanol growth phases was measured.

The CYC1 core promoter, P_CYC1(−), showed a very low basal activity (Fig. 2a). Insertion of Z268 elements (P_{CYC1+4×[Z268]}) increased basal expression mildly in the absence of an artificial TF. A variety of responses were observed in the presence of synthetic TFs with different TADs. Gcn4^A showed insignificant activation effects in the exponential growth phase, and a weak effect in the ethanol growth phase. VP16^A, 10*WD, Med3, and Med15 showed >3-fold activation in the cells at the exponential phase. Med3 and Med15 drove significantly higher expression on ethanol than VP16^A.

A short synthetic linker sequence with no biological relevance was into the TEF1 promoter to provide a landing pad for additional cis-acting elements (Fig. 2b). Insertion of this sequence resulted in a decrease in yEGFP expression (P_TEF1(−); Fig. 2b). Addition of the Z268 element sequence did not change expression in the exponential phase, but led to a two-fold increase in expression on ethanol (Fig. 2b). Very similar results were observed for expression modification in the presence of the different TAD/mediator-containing constructs (Fig. 2a, b), such that the data from the P_{TEF1+4×[Z268]} promoter responses positively correlated with the effects observed in the P_{CYC1+4×[Z268]} promoter (Supplementary Figure 1). The strongest trans-activation was observed with Med15, which drove 30–50% higher expression on both exponential and ethanol phases than VP16^A. Gcn4^A showed the weakest trans-activating capacity, with 10*WD being slightly stronger. VP16^A and Med3 showed similar activating capacity, in the cells at the exponential growth phase. In contrast to the others, Med3 drove higher expression on ethanol than during the exponential growth phase.

In summary, the artificial trans-activators increased the expression from the hybrid CYC1 promoters by larger fold-changes than that from the hybrid TEF1 promoters, but because the core CYC1 promoter expression was so weak, the absolute expression increases from the hybrid TEF1 promoters were larger. Med3 drove an increased activity at the ethanol-growth phase and Med15 had the strongest trans-activating capacity. Overall, these results provide insight on the variation of trans-activation effects of different protein modules on different core promoters. The stronger activation from Med15 than that from VP16^A might be because Med15 activates directly, whereas VP16^A recruits Med15 as a co-mediator to activate⁴⁵.

Additional trans-activation modules do not increase the maximal activities of the Sk.GAL2 promoter

Previously, we have shown that the GAL1 promoter is stronger than glycolytic promoters when induced on galactose³⁹. We further characterised a set of GAL promoters from other Saccharomyces species, and found even stronger GAL promoters, including the GAL2 promoters from S. eubayanus (Se.GAL2) and S. kudriavzevii (Sk.GAL2)⁹. As shown above, promoter hybrids with cis-regulatory elements can be boosted by synthetic trans-activators (Fig. 2b). We applied the same mechanism to the SkGAL2 promoter. The PZ4 cis-element ‘TA GAG TGA GAC GTT’ is found naturally in the Sk.GAL2 promoter, but not in the Se.GAL2 promoter or in other GAL promoters we have previously characterised⁹.

To develop an integrated synthetic trans-activating module to boost activity of the GAL2 promoter, we designed three synthetic Zif domains using ToolGen’s Zinc Finger Module set⁴⁶ (ZifPZ42 targeting ‘GAG TGA GAC GTT’, ZifPZ43 targeting ‘TGA GAC GTT’, and ZifPZ44 targeting ‘GAG TGA GAC’) and one TALE domain, TALPZ4 (recognising ‘TA GAG TGA GAC GTT’) based on the dHax3 scaffold⁴⁷. These DNA-binding domains were separately cloned into the N-terminal of the Med3 or Med15 modules to generate synthetic trans-activators. Only ZifPZ42 and ZifPZ43 (Fig. 2b)-containing trans-activators exhibited trans-activation on the hybrid TEF1–PZ4 promoter (P_{TEF1+4×[PZ4]}), with the ZifPZ43-containing trans-activators being stronger. However, by observing multi-well plate cultures, we saw that the strains expressing ZifPZ43-Med3 and ZifPZ43-Med15 grew slower compared to the reference. Strains expressing ZifPZ42-Med15 and ZifPZ43-Med15 also exhibited a non-homogeneous population (Supplementary Fig. 2). Zif268-Med3 showed a stronger activation on P_{TEF1+4×[Z268]} on the ethanol growth phase; this was not seen for ZifP643-Med3 with P_{TEF1+4×[PZ4]} (Fig. 2b). By analysing the sequences inserted into the promoters using the YEASTRACT web-service⁴⁸, we found the sequences containing the Z268 and the PZ4 elements also included binding sites for different endogenous yeast TFs (Supplementary Fig. 3). This may indicate a synergistic activation effect of synthetic Zif268-Med3 with endogenous TFs on the hybrid TEF1 + [Z268] promoter, whereas such effect is not present for ZifPZ43-Med3 on the hybrid TEF1 + [PZ4] promoter.

We chose the ZifPZ43 trans-activator to characterise potential combinatorial effects of additional trans-activation components on the GAL promoter. We modified the Sk.GAL2 promoter by inserting one to four additional PZ4 elements, and evaluated the expression outputs from the Sk.GAL2 promoter and its mutants in the presence of the synthetic trans-activators ZifPZ43-Med3 or ZifPZ43-Med15. The Se.GAL2 promoter (no PZ4 element) was used as a reference. Expression of a yEGFP reporter cassette was examined in the presence and absence of a synthetic trans-activator using a gal80Δ background strain in early exponential phase on glucose or ethanol (Fig. 2c). Expression of ZifPZ43-Med3 or ZifPZ43-Med15 did not appear to impact on growth rate in these strains.

Despite that GAL promoters and the expression of the Gal4 trans-activator are repressed by Mig1p in the presence of glucose, basal expression from GAL promoters was observed in strains grown on glucose (Fig. 2c). Consistent with previous observations⁹, the Sk.GAL2 promoter showed a higher basal expression on glucose than the Se.GAL2 promoter. As noted, the Se.GAL2 promoter does not have a PZ4 element, and expression of trans-activators ZifPZ43-Med3 or ZifPZ43-Med15 did not increase yEGFP expression from the Se.GAL2 promoter. However, expression of Zif268-VP16^A led to a two-fold increase of yEGFP fluorescence, despite that the Se.GAL2 promoter does not contain Z268 element. Nevertheless, GFP fluorescence was still very low in this strain on glucose.

For strains harbouring the Sk.GAL2 promoter and its PZ4-element mutants growing on glucose, activators ZifPZ43-Med3 and ZifPZ43-Med15 increased yEGFP expression (Fig. 2c). There was a positive correlation between the number of PZ4 elements and the observed GFP fluorescence (Supplementary Fig. 4). In contrast to our previous findings with the TEF1 promoter (Fig. 2b), the Med3 mediator drove a stronger activation than the Med15 mediator. Zif268-VP16^A drove a ~2-fold increase in the expression from Sk.GAL2 4×[PZ4] promoter and Sk.GAL2 5×[PZ4] promoter (Fig. 2c; but not from other ‘PZ4’-only Sk.GAL2 promoters). This may indicate that there is a non-specific trans-interaction between Zif268 and the sequences inserted into Sk.GAL2 4×PZ4 promoter and Sk.GAL2 5×[PZ4] promoter. The Sk.GAL2 5×[PZ4] promoter in the presence of ZifPZ43-Med3 showed ~3-fold stronger expression output, in comparison with the commonly-used TEF1 promoter with the same yEGFP-reporter construct (Construct 1, Fig. 1), and 1.7-fold stronger than the TDH3 promoter³⁹. This shows the strong trans-activation effects from the artificial trans-activator-cis element interaction on the hybrid GAL promoter on glucose.

On ethanol, the Se.GAL2, Sk.GAL2, and Sk.GAL2 + [PZ4] promoters were activated to a very high level (Fig. 2c), at least seven-fold higher than the TEF1 promoter (Construct 1, Fig. 1). Neither ZifPZ43-Med3 nor ZifPZ43-Med15 significantly trans-activated the Sk.GAL2 + PZ4 promoters on ethanol.

In addition to the Sk.GAL2 + [PZ4] promoters, we also constructed a Sk.GAL2 + Z268 hybrid promoter, and tested the trans-activation effects of artificial trans-activators Zif268-VP16^A, Zif268-Med3, and Zif268-Med15. Expression of these trans-activators increased the yEGFP activity from the Sk.GAL2 + [Z268] promoter by at least five-fold (Fig. 2c) on glucose. Zif268-Med3 showed the strongest activation effect, resulting in the expression of Sk.GAL2 + Z268 promoter on glucose being near to the levels on ethanol. These findings further confirmed the strong trans-activation capacity of artificial trans-activators.

However, by observing the cultures in multi-well plates, we saw that Sk.GAL2 + Z268 promoter-testing strain (gal80Δ background) with Zif268-Med3 and Zif268-Med15 grew slower compared to the reference, and the Zif268-Med15-expressing strain did not grow on ethanol. Such effects were not seen in the TEF1/CYC1 + [Z268] promoter-testing strains (GAL80-wildtype; Fig. 2). We postulated that the growth impairment might be caused by untargeted trans-activation from Zif268-Med3 and Zif268-Med15 in gal80Δ background.

In summary, artificial trans-activators can significantly activate the GAL promoter via the orthogonal cis-elements on glucose. However, they did not further increase the activities of the GAL promoters on ethanol, indicating that these activators do not significantly increase its maximum expression output of the GAL promoter.

Eukaryote-like trans-repression mechanisms can be used to build effective repression circuits in yeast

Trans-repression is another natural mechanism for transcriptional regulation. In bacteria, a repressor commonly functions by binding to the elements near the transcription start site to prevent transcription from the target promoter^49,50. A similar mechanism can be reconstructed in yeast by the insertion of bacterial repressor-binding sites near to ‘TATA-box’ or transcription initiation sequence in the target promoter and the expression of the cognate bacterial repressor^6,15,21. However, intensive effort, including construction and screening of large libraries, was required to obtain the ideal hybrid repressible promoter; in addition, a strong constitutive promoter is required to express bacterial repressors^6,15. In contrast to bacterial repressors, transcriptional repressors in eukaryotes commonly mediate gene repression through a series of regulatory mechanisms including recruiting histone deacetylase complexes to direct nucleosome formation and blocking co-activators^51,52. These mechanisms may be more relaxed about the position of cis-acting elements in target promoters and can potentially be exploited to develop a trans-repressing toolset.

We characterised the trans-repressing capacity of four protein modules, including Sin3 C-terminal domain (Sin3 aa600-1536; Sin3^C)⁵³, Tup1⁵⁴, Cyc8⁵⁴, and Mig1 C-terminal domain (Mig1 aa123-504; removal of DNA-binding domain; Mig1^C)⁵⁵. Sin3^C mediates repression through its interactions with histone deacetylase complexes⁵³, Tup1/Cyc8 through recruiting histone deacetylase complexes and masking binding of co-activators⁵⁴, and Mig1^C is thought to recruit the Tup1/Cyc8 complex⁵⁵. Sin3^C, Tup1, and Cyc8 have been previously tested in various studies by fusing them to a DNA-binding domain, but Mig1^C has not^53,54.

We first attempted to characterise the fusion of Zif268-SV40^NLS with the C-terminal Sin3^C, Tup1, or Mig1^C. The HAC1 promoter was used to control the fusion expression. yEGFP controlled by P_{TEF1+4×[Z268]} was used as the reporter (Fig. 3a). However, yeast transformants of Zif268-Sin3^C and Zif268-Tup1 constructs grew extremely slowly on agar plates, and the colonies could not be picked after more than a week of incubation. We were able to characterise the strain expressing Zif268-Mig1^C. We observed two distinct cell populations: in one population, the yEGFP expression was severely repressed, whereas in the other, no repression (relative to the control construct) was observed (Supplementary Fig. 5).

The unstable expression profiles in the Zif268-containing promoter constructs suggested off-target effects rendering the constructs unusable. We therefore developed an alternative approach using TetR elements as the DNA-binding domain for characterisation of the four trans-repressing modules. A cis-regulatory reporter cassette was constructed by replacing the Z268 elements in P_{TEF1+4×[Z268]} with TetO elements. The control construct in the absence of a repressor module showed ~2-fold higher expression in the exponential phase than in the ethanol phase and did not respond to tetracycline addition (Fig. 3a).

Yeast expressing TetR-derivative repressors (TetR-Sin3^C, TetR-Tup1, TetR-Cyc8, and TetR-Mig1^C) grew normally on plates and showed strong repression on the TEF1 + 4×[TetO] promoter in the absence of tetracycline. TetR-Tup1 was most efficient, showing >100-fold repression and eliminating yEGFP fluorescence in > 90% cell population (Fig. 3a; Supplementary Fig. 5). TetR-Cyc8, and TetR-Mig1^C showed the repression similar to TetR-Tup1. The repression was maintained in the cells at the ethanol-growth phase, even for the cells expressing TetR-Mig1^C, although Mig1 itself is a glucose-dependent repressor. Repression on ethanol by TetR-Mig1^C might be due to the inclusion of a nuclear localisation sequence (Fig. 1a), which dominates the glucose-dependent nuclear localisation of Mig1⁵⁶. Addition 50 μM tetracycline can de-repress yEGFP expression in the whole cell population (Supplementary Fig. 6). TetR-Sin3^C was less efficient, not eliminating yEGFP fluorescence in the whole population (Fig. 3a; Supplementary Fig. 6). Yeast expressing TetR-repressors did not show a homogenous population: ~ 8% population deviated from the major population and showed higher yEGFP fluorescent levels under the conditions without tetracycline additions (TetR-Tup1 in Supplementary Fig. 5). We postulated that this might be caused by the loss of repressor-expressing plasmids: centromeric plasmids show a 3% loss rate per generation⁵⁷.

The repression was maintained in the cells at the ethanol-growth phase, even for the cells expressing TetR-Mig1^C, although Mig1 itself is a glucose-dependent repressor. Repression on ethanol by TetR-Mig1^C might be due to the inclusion of a nuclear localisation sequence (Fig. 1a), which dominates the glucose-dependent nuclear localisation of Mig1⁵⁶.

In summary, repression capacities of the four trans-repressing domains were sequenced as Tup1 ~ Mig1^C ~ Cyc8 » Sin3^C, and a low-level expression of a TetR fusion with Tup1, Mig1C, or Cyc8 repressor can fully repress the expression of a strong promoter inserted with TetO elements.

Tetracycline-mediated repression of GAL promoters

To develop a mechanism that can fully repress GAL promoters-controlled metabolic pathways on demand, we used the tetracycline de-repressible circuit to control the expression of the GAL repressor Gal80p. In the resulting strain, TetR-Tup1 was expressed under the control of the HAC1 promoter, the TEF1 + [TetO] promoter was integrated upstream of the GAL80 gene to control its expression, and a yEGFP gene was introduced under the control of the GAL1 promoter to report the regulatory effects of tetracycline (Fig. 3b). The strain was first grown on ethanol to relieve Mig1p-mediated glucose repression on GAL promoters^2,9, in the absence or presence of tetracycline. In the absence of tetracycline, a very high level of yEGFP fluorescence was detected (Fig. 3c and Supplementary Fig. 6), showing a fully de-repression state of the GAL1 promoter^2,39. In the presence of 125 μM tetracycline, yEGFP fluorescence was repressed by >200 fold to very minimal fluorescence.

We then grew the strain on nutrient agar containing 20 g L⁻¹ glucose as the carbon source. Consistent with our previous observation that 20 g L⁻¹ glucose in agar is not sufficient to repress the expression from GAL promoters⁹, the GAL1 promoter was moderately de-repressed in the absence of tetracycline (Fig. 3c and Supplementary Fig. 6). Addition of 125 μM tetracycline repressed the yEGFP expression by >200 fold.

In summary, by integrating tetracycline-inducible expression circuit to control GAL80 expression, we successfully deployed a tetracycline-mediated mechanism to repress GAL promoters.

A heat-mediated response circuit for 37 °C induction of GAL promoters

Temperature is a primary parameter to control during yeast cultivation. Although the preferred temperature for S. cerevisiae is 30 °C, growth at higher temperature can potentially reduce the cooling costs for exothermic fermentation processes⁵⁸, and increased temperatures are well tolerated. Transition to an increased bioprocess temperature upon entry into the production phase (when fast growth is no longer required) would therefore make a convenient tool for regulatory control and may make the bioprocess more economical. To exploit this potential, we designed a heat-inducible circuit to modulate expression from GAL promoters in strains growing on glucose. Our model employs a heat-inducible degron⁵⁹ to regulate the stability of the glucose-dependent repressor Mig1p (Fig. 4b).

We first used the yEGFP reporter to characterise the previously-developed heat-inducible degron (H.degron)⁵⁹ in our system. The full heat-inducible degron module comprises four parts: a ubiquitin moiety (Ubi) that is cleaved off by ubiquitin C-terminal hydrolase after translation, an ‘N-end rule’ residue Arg, a linker ‘HGSGTMV’, and mouse dihydrofolate reductase P66L temperature sensitive mutant (DHFR*). The heat-inducible degron Ubi4-RHGSGTMV-DHFR* was fused to the N-terminus of yEGFP, and the fusion (UBI4-RHGSGTMV-DHFR*-yEGFP) was expressed under the control of TEF1 promoter and the PGK1 terminator (Fig. 4a). yEGFP was characterised as the control. Two resulting yeast strains, expressing yEGFP and UBI4-RHGSGTMV-DHFR*-yEGFP, were grown at 30 °C or at 37 °C. Consistent with previous studies^59,60, elevation of growth temperature to 37 °C did not slow the growth. In the yEGFP control, fluorescence was 1.2-fold lower at 37 °C (Fig. 4c), indicating that elevated temperature did not dramatically affect yEGFP abundance.

Tagging Ubi4-RHGSGTMV-DHFR* decreased fluorescence at 30 °C by 4.5-fold compared to the control yEGFP (Fig. 4c), suggesting either disruption of transcription/translation or basal degradation by the degron. At 37 °C, fluorescence decreased by 16.2-fold. These data demonstrate the functionality of Ubi4-RHGSGTMV-DHFR* as an effective heat-inducible degron (H.Degron).

In summary, the basal degradation (or impaired expression) at 30 °C caused by Ubi4-RHGSGTMV-DHFR* (H.Degron) is significant; moreover, heat-inducible degradation by this degron at 37 °C, although dramatically decreasing yEGFP abundance, did not result in a full depletion of target yEGFP (Fig. 4c). Incomplete degradation might be problematic for application to control the abundance of a transcription factor, because transcription factors can function effectively at a relatively low expression level. To apply the H.Degron to regulate the abundance of Mig1p for the proposed 37 °C-mediated de-repression of GAL promoters, it will be important to overcome these two limitations, otherwise a sharp temperature-induced boundary effect cannot be achieved.

To deploy the 37 °C-mediated de-repression module on GAL promoters, we used a gal80Δ strain expressing yEGFP under the control of the GAL1 promoter and the URA3 terminator. In this strain, we modified the MIG1 locus by fusing the H.Degron to the N-terminus of Mig1p and introducing an alternative promoter to control the expression strength. We tested three promoters to provide variable expression strength with the aim of titrating the transcription of H.Degron-MIG1 to obtain a sharp boundary effect: the TEF2 promoter (strong)³⁹, the HAC1 promoter (medium-weak; Fig. 1c), and the NRG1 promoter (weak; Fig. 1c). The resulting yeast strains were grown on glucose at 30 °C or 37 °C to examine yEGFP expression from the GAL1 promoter (Fig. 4d). In these cases, cultivation at 37 °C did not result in slowed growth. The strain with the TEF2 promoter construct showed a very low level of yEGFP fluorescence at 37 °C, indicating the expression from the TEF2 promoter was too high and the H.Degron was not sufficient to deplete Mig1p. The strain with the NRG2 promoter showed an induced level of yEGFP fluorescence at 30 °C, indicating that the expression from the NRG2 promoter was not sufficient and the basal degradation of H.Degron caused the depletion of Mig1p. Fortunately, the strain with the HAC1 promoter construct showed very low level yEGFP fluorescence at 30 °C and good induction 37 °C (Fig. 4d and Supplementary Fig. 6). This confirmed the functionality of 37 °C-mediated de-repression of GAL promoters.

Expanded regulatory circuits for conditional control of sesquiterpene (nerolidol) production

Previously, we developed yeast strains for high-level sesquiterpene production through metabolic engineering^17,61,62. To develop these strains, GAL promoters, coupled with the deletion of GAL repressor gene GAL80, were used to control the expression of the genes from the mevalonate pathway and the genes for production of the sesquiterpene trans-nerolidol. The resulting strain NLD401 produced ~1.7 g L⁻¹ nerolidol in two-phase flask cultivation at 30 °C¹⁷. In the current work, we developed an isogenetic yeast strain (N9R5HIRBU), in which the synthetic genetic modules for tetracycline-mediated repression and 37 °C-mediated de-repression of GAL promoters were also established (Fig. 5a). In this strain, Yellow Fluorescence-Activating and absorption-Shifting Tag (YFAST) and Actinidia chinensis (kiwifruit) nerolidol synthase (AcNES1) were expressed through a 2A sequence-mediated bicistronic expression mechanism (YFAST-2A-AcNES1)¹⁷. The YFAST-2A-AcNES1 was under the control of the GAL2 promoter. This construct not only contributed to an improved nerolidol production, but also allowed us to monitor the GAL2 promoter activities by measuring YFAST fluorescence throughout the whole cultivation as a proxy for GAL2 promoter-driven expression. We evaluated the effects of tetracycline and elevated temperature on nerolidol production and the GAL promoter activities in this strain via two-phase dodecane-overlayed flask cultivation.

In flask cultivation at 30 °C and the absence of tetracycline, strain N9R5HIRBU (+ tetracycline-repressible/heat-inducible regulatory modules: Fig. 3b and Fig. 4b) showed a specific maximum growth rate (µ_max) of 0.3 h⁻¹ and a maximum cell density at OD₆₀₀ of ~14 at 72 hr. N9R5HIRBU produced ~1.76 g L⁻¹ nerolidol at 72 hr (Fig. 5e). The dynamics of YFAST fluorescence were also similar in both strains: a very low expression in the first 12 hr when glucose was the carbon source and a gradually induced expression after 12 hr when glucose was depleted (Fig. 5c). These physiological features were similar to the isogeneic strain NLD401 published previously (w/o tetracycline-repressible/heat-inducible regulatory modules; gal80Δ)¹⁷. These data show that the ‘OFF’ state of tetracycline-mediated repression on GAL promoters is tight, producing a gal80Δ-like genotype and phenotype.

Addition of 125 μM tetracycline at 30 °C did not significantly alter growth characteristics in the exponential phase (Fig. 5b and d), however much more biomass was produced during the post-exponential phase, resulting in a maximum OD₆₀₀ of ~22 (Fig. 5b). Consistent with tetracycline-mediated repression on yEGFP expression from the GAL1 promoter (Fig. 3), YFAST expression in N9R5HIRBU was essentially fully repressed by 125 μM tetracycline throughout the whole cultivation (Fig. 5c). Only ~0.03 g L⁻¹ nerolidol was produced at 72 hr, 60-fold lower than in the absence of tetracycline at 30 °C (Fig. 5e). These data demonstrate that the ‘ON’ state of tetracycline-mediated repression on GAL promoters provided a fine mechanism for nearly full repression on GAL promoter-controlled heterologous synthetic pathways. It also demonstrates that nerolidol production delivers a metabolic burden which decreases biomass production. The tetracycline repression module can facilitate strain maintenance during laboratory manipulation by relieving the metabolic burden caused by heterologous synthetic pathways.

We then characterised strain N9R5HIRBU at 37 °C in the absence of tetracycline to investigate the effects of elevated temperature on nerolidol production. Pre-culturing was performed at 37 °C to prepare the seed inoculum. A decrease in exponential growth rate and biomass accumulation was observed at 37 °C, with a maximum OD₆₀₀ of ~11 (Fig. 5b and d). YFAST fluorescence at 0 h was ~20-fold higher at 37 °C than at 30 °C, and was continuously induced during the exponential and post-exponential growth phases (Fig. 5c). This suggests that heat-inducible protein degradation might not be efficient for complete depletion of Mig1p, but does provide a moderate relief of glucose repression. Reflecting the induced expression from the GAL promoter, strain N9R5HIRBU produced ~ 0.5 g L⁻¹ nerolidol at 24 h and 37 °C, ~6-fold improvement compared to that at 30 °C; and produced ~2.54 g L⁻¹ nerolidol at 72 h, a 45% improvement on the original NLD401 strain and on the N9R5HIRBU strain grown at 30 °C (Fig. 5e). Consistent with this, the specific nerolidol production rates in the cells grown at 37 °C were significantly higher than those in the cells grown at 30 °C, with a 9-fold increase in the first 24 h and a 49% increase from 24 hr to 72 h (Fig. 5f). These data demonstrate that the integration of 37 °C-mediated induction circuits for the GAL promoters resulted in improved production of nerolidol-producing strains at an elevated temperature.