Strains used in this study are listed in Table 1. Plasmids are listed in Supplementary Table 1. Relevant primers are listed in Supplementary Table 2.

Deletion or replacement of a gene in K. marxianus and K. lactis was performed by homologous recombination with the aid of a CRISPR plasmid. Three CRISPR vectors, LHZ296, LHZ301, and LHZ531, were used as backbones to build CRISPR plasmids (Shi et al., 2021). Primers containing 20 bp target sequence were annealed in pairs and inserted into SapI or AarI sites of LHZ296, LHZ301, and LHZ531. Details of resultant CRISPR plasmids were listed in Supplementary Table 1.

In K. lactis and K. marxianus, HMR locus contained a silenced copy of genes determining mating type a (a1, a2), while HML locus contained that of genes determining mating type α (α1, α2, and α3) (Barsoum et al., 2010; Lertwattanasakul et al., 2015). HMRa and HMLα serve as donors during the recombinational process that allows a MATa cell to switch to MATa or vice versa. To prevent mating-type switching in K. marxianus, the HMR locus was removed. A fragment upstream of a1 was amplified by the primer pair STF23F/STF23R and a fragment upstream of a2 was amplified by STF24F/STF24R. Both fragments were ligated and inserted into pMD18-T (Takara, D101A) to obtain LHZ329. Donor sequence was amplified from LHZ329 by STF23F/STF24R and co-transformed with LHZ328 into FIM-1ΔU (MATα) by a lithium acetate-mediated method (Antunes et al., 2000). The resultant strain was named LHP507. Similarly, the HML locus was removed from FIM-1ΔU (MATa) to obtain LHP506. To prevent mating-type switching in K. lactis, the HML locus was removed from LHP567 to obtain LHP568. The HMR locus was removed from LHP567 to obtain an intermediate strain (MATa, ura3Δ, HMRΔ). Then, a CRISPR plasmid LHZ567 was transformed into the intermediate strain. LHZ567 caused DNA double-strand breaks upstream of a1 and a2 in the MATa locus, respectively. The HMLα locus in the intermediate strain provided a template to repair the double-strand breaks and that led to the switch of mating type from MATa to MATα. The resultant strain was named LHP569.

To delete KmTRP1 in LHP506 for the selection of diploid cells, KmTRP1 with the flanking sequence was amplified by primer pair STF40F/STF40R. The fragment was ligated with pMD18-T. The ORF of KmTRP1 in the resultant plasmid was removed by mutagenesis PCR using primer pair STF41F/STF41R to obtain LHZ539. The donor sequence was amplified from LHZ539 by STF40F/STF40R and co-transformed with LHZ538 into LHP506 to obtain LHP508. Similarly, KmHIS3 was deleted in LHP507 to obtain LHP560. KmSTE11, KmSTE7, KmFUS3, or KmSTE5 was deleted in LHP508 to obtain LHP561, LHP697, LHP698, and LHP563 respectively. KlURA3 was deleted in ATCC8585 to obtain LHP567. KlTRP1 was deleted in LHP568 to obtain LHP570. KlHIS3 was deleted in LHP569 to obtain LHP571. The 697-705 aa of KmSTE5 was deleted in LHP508 to obtain LHP1043.

To replace KmSTE11 with KlSTE11, KmSTE11 with the flanking sequence was amplified by the primer pair STF54F/STF54R and ligated with pMD18-T. A linear fragment was amplified from the resultant plasmid by STF81F/STF81R. The fragment was ligated with the ORF of KlSTE11 amplified by STF55F/STF55R through Gibson assembly to obtain LHZ544. In LHZ544, the ORF of KISTE11 was flanked by the natural promoter and terminator of KmSTE11. The donor sequence was amplified from LHZ544 by STF54F/STF54R and co-transformed with LHZ542 into LHP508 to obtain LHP562. Similarly, KmSTE7, KmFUS3 or KmSTE5 in LHP508 was replaced by its ortholog in K. lactis to obtain LHP565, LHP566 and LHP564 respectively. KmSTE11, KmSTE7, KmFus3 or KmSTE5 in LHP508 was replaced by its ortholog in S. cerevisiae to obtain LHP699∼LHP702. KlSTE11, KlSTE7, KlFUS3 or KlSTE5 in LHP570 was replaced by its ortholog in K. marxianus to obtain LHP572, LHP705∼LHP707 respectively, and replaced by its ortholog in S. cerevisiae to obtain LHP708∼LHP711. KmSTE5 in LHP508 was replaced by KlSTE5-KmBD, KmSTE5-KlBD, KlSte5-ES and ScSTE5ΔS to obtain LHP703, LHP704, LHP1044 and LHP1045 respectively. KlSTE5 in LHP570 was replaced by KlSTE5-KmBD, KmSTE5-KlBD, KmSte5ΔS, and ScSTE5ΔS to obtain LHP712, LHP713, LHP1046, and LHP1048. KmSTE5 in LHP707 was replaced by KlSte5-ES to obtain LHP1047.

To measure the strength of the interaction between Ste5 and kinases, a series of the pGADT7-Ste5 plasmids and pGBKT7-Ste7 plasmids were constructed. Sequences encoding KmSTE5, KlSTE5, ScSTE5, KlSTE5-KmBD, KmSTE5-KlBD, KmSTE5ΔS, KlSTE5-ES, and ScSTE5ΔS was amplified and cloned into EcoRI site of pGADT7 (Clontech, 630442) by Gibson assembly. ORFs of Ste7 orthologs were amplified and cloned into EcoRI site of pGBKT7 (Clontech, 630448) separately by Gibson assembly.

Kluyveromyces marxianus, K. lactis, and S. cerevisiae cells were grown in YPD medium (2% peptone, 1% yeast extract, 2% agar for plates). Synthetic complete (SC) media were prepared as described before (Amberg, 2005). ME plates (2% malt extract,3% agar) were prepared for the mating of K. lactis cells (Zonneveld and Steensma, 2003). The concentration of agar in ME plates was adjusted to 5% to optimize the mating of K. marxianus in this study.

Protein sequences of ScSTE11 (YLR362W), ScSTE7 (YDL159W), ScFUS3 (YBL016W), and ScSTE5 (YDR103W) were downloaded from the Saccharomyces Genome Database (SGD). Protein sequences of KmSTE11 (KLMA_30080), KmSTE7 (KLMA_70246), KmFUS3 (KLMA_20041), KmSTE5 (KLMA_20171), KlSTE11 (KLLA0_B13112g), KlSTE7 (KLLA0_C16577g), KlFUS3 (KLLA0_E10539g), and KlSTE5 (KLLA0_F12023g) were downloaded from the KEGG. Domains of Ste11, Ste7, Fus3, and Ste5 in K. marxianus and K. lactis were determined by alignment with their counterparts in S. cerevisiae. The range of domains were determined as follow: ScSte11 SAM (aa 21–83), ScSte11 RBL (aa 116–236), ScSte11 kinase (aa 415–717) (Yerko et al., 2013); KlSte11 SAM (aa 11–73), KlSte11 RBL (aa 107–228), KlSte11 kinase (aa 439–730); KmSte11 SAM (aa 11–73), KmSte11 RBL (aa 107–228), KmSte11 kinase (aa 445–735); ScSte7 kinase (aa 191–466) (Won et al., 2011); KlSte7 kinase (aa 166–442); KmSte7 kinase (aa 155–431); ScFus3 kinase (aa 13–309); KlFus3 kinase (aa 13–327); KmFus3 kinase (aa 13–322); ScSte5 RING (aa 177–229), ScSte5 PH (aa 388–518), ScSte5 Fus3BD (aa 288–315), ScSte5 VWA (aa 593–786), ScSte5 Ste7BD (aa 744–895) (Good et al., 2011); KlSte5 RING (aa 128–181), KlSte5 PH (aa 331–450), KlSte5 VWA (aa 498–683), KlSte5 Ste7BD (aa 647–746); KmSte5 RING (aa 136–189), KmSte5 PH (aa 345–464), KmSte5 VWA (aa 512–702), and KmSte5 Ste7BD (aa 660–763).

The quantitative mating assay was performed as described before with modifications to accommodate Kluyveromyces species (Miller and Kirchmaier, 2014). LHP560 served as a MATα tester strain for K. marxianus, and LHP571 served as a tester strain for K. lactis. Fresh cells of the experimental MATa strain and MATα tester strain were grown in YPD liquid medium at 30°C overnight until the optical density at 600 nm (OD₆₀₀) was above 12. Cells were washed and suspended in H₂O at an OD₆₀₀ of 1. Excess amount of tester cells compared with experimental cells was required in the quantitative mating assay in S. cerevisiae (Miller and Kirchmaier, 2014). In this study, the optimal ratio between experimental cells and tester cells in the mating of K. marxianus was determined as 1:2 and the same ratio was applied in the mating of K. lactis. For K. marxianus, 0.75 mL MATa cells and 1.5 mL MATα cells were mixed and loaded into a syringe attached to a Swinnex Filter Holder (Millipore, SX0002500). For K. lactis, 0.5 mL MATa cells and 1 mL MATα were mixed and loaded. The mixture of cells was collected on a nitrocellulose filter (Whatman, 7184-002). The filter, cells side up, was placed onto a ME plate and incubated at 30°C. Unless indicated, the incubation time for K. marxianus cells was 4 h, and that for K. lactis cells was 10 h. The filter was then transferred to a 50 mL centrifuge tube containing 4 ml H₂O. Cells were resuspended thoroughly by the vortex. The cells were serial diluted and plated on YEPD for total colony-forming units and on SC-His-Trp plates to select for diploid cells. The mating efficiency was calculated as the number of diploid cells divided by one-third of the number of total colony-forming units since the experimental MATa strain was mixed with MATa tester strain in a 1:2 ratio. Three different colonies of an experimental strain were assayed in parallel.

Two-tailed t-tests were performed to determine if there is a significant difference between samples. P-values of less than 0.05 were considered statistically different.

A Matchmaker GAL4 Two-Hybrid System was applied in this study (Clontech, PT3247-1). pGADT7-Ste5 plasmids were transformed into AH109 separately and selected on SC-Leu plate. pGBKT7-Ste7 plasmids were transformed into Y187 separately and selected on SC-Trp plate. A Trp⁺ transformant was mated with a Leu⁺ transformant on YPD and diploid cells were selected on SC-Leu-Trp plates.

The quantitative β-galactosidase assay was performed as described in the Yeast Protocols Handbook (Clontech, PT3024-1). Diploid cells were grown in liquid culture and lysed by freeze/thaw cycles in liquid nitrogen. Activities in the cell lysate were measured by using ONPG (Solarbio, O8040-1g) as a substrate. 1 unit of β-galactosidase is defined as the amount which hydrolyzes 1 μmol of ONPG to o-nitrophenol and D-galactose per min per cell. To remove the autoactivation of pGBKT7-Ste7, the unit measured for a diploid containing a pGBKT7-Ste7 plasmid and a void pGADT7 vector was subtracted from the unit of a diploid containing the same pGBKT7-Ste7 plasmid and a pGADT7-Ste5 plasmid. Three different colonies of one diploid strain were subjected to the assay and activities were measured in duplicates.

Fresh K. marxianus and K. lactis cells were grown in YPD liquid medium overnight until the OD₆₀₀ was above 12. Cells of 300 mL overnight cultures were harvested and frozen at −80°C. RNA was extracted from frozen cells using a Quick-RNA Fungal/Bacterial Miniprep kit (Zymo Research, R2010) and were reverse transcribed using a PrimeScript RT Reagent Kit (Takara, RR037A). The qPCR was performed using TB Green Premix Ex Taq (Takara, RR820A). Primers used in qPCR were listed in Supplementary Table 2.

Same as S. cerevisiae, K. marxianus and K. lactis have two mating types designated as MATa and MATα. In responding to harsh environments, haploid cells of opposite types mate to produce diploid and form spores (Zonneveld and Steensma, 2003; Cernak et al., 2018). In a quantitative mating assay of K. marxianus, mixed haploid cells started to form diploid after 3 h, when the number of cells was three times the initial number, the mating efficiency was only 0.093% after 4 h when the number of cells duplicated twice. The efficiency jumped to 0.762% after 5 h (Figure 2A). Similar to K. marxianus, K. lactis haploid cells started to produce diploid cells after the number of cells was three times the initial number. It took longer (5 h) for K. lactis cells to finalize mating, probably because K. lactis grows slower than K. marxianus (Lane and Morrissey, 2010). When cells were duplicated about two times after 10 hours, the mating efficiency of K. lactis was 3.39%, which was much higher than that of K. marxianus (Figure 2B). The efficiency climbed to 9% after 11 h. However, K. lactis and K. marxianus are both poor maters compared to S. cerevisiae. Wild-type S. cerevisiae cells exhibit mating efficiencies between 15 and 90% (Miller and Kirchmaier, 2014). Different mating behaviors of three species suggested the divergence in regulating mating response, including that in the pheromone pathway.

Based on the sequence identity, orthologs of Ste5 and associated kinases, as well as domains in these proteins, can be recognized in K. lactis and K. marxianus (Figure 2C). In K. marxianus, deletion of genes encoding Ste5, Ste7, Ste11, and Fus3 caused mating defects (Figure 2D). Similar results were reported in K. lactis previously (Kawasaki et al., 2008). Therefore, Ste5 and kinases from both Kluyveromyces species play similar roles in the mating pheromone pathways as their counterparts in S. cerevisiae. The hierarchy of the sequence identities of full-length Ste5 and associated kinases among K. lactis, K. marxianus and S. cerevisiae was determined as follow: Fus3 > Ste11 > Ste7 > Ste5 (Figure 2C). As the least conserved protein, KlSte5 and KmSte5 display similar sequence identities (24%) to ScSte5. KlSte11 and KmSte11, KlSte7 and KmSte7, KlFus3, and KmFus3 also displayed similar sequence identities to their counterparts in S. cerevisiae, respectively.

To investigate functional homology of Ste5 and associated kinases, the gene encoding Ste5, Ste7, Ste11, or Fus3 in K. lactis or K. marxianus was replaced by its ortholog in S. cerevisiae at its natural locus (Figure 3A), or was swapped between K. lactis and K. marxianus (Figure 3B). The functional replaceability of an ortholog in the new host was indicated by the mating efficiency of the modified host relative to that of the natural host.

The functional replaceability of ScSte11 and KlSte11 in K. marxianus was 125 and 118%, respectively (Figures 3C,E). The functional replaceability of ScSte11 and KmSte11 in K. lactis was 86 and 98%, respectively (Figures 3D,F). High functional replaceability of Ste11 orthologs in the interspecies complementation (all above 80%) suggested the function of Ste11 was well conserved in K. marxianus, K. lactis and S. cerevisiae.

Replacing the STE7 genes in K. marxianus and K. lactis by their orthologs in S. cerevisiae reduced the mating efficiencies to 46 and 52%, respectively (Figures 3C,D). The result suggested the functional replaceability of ScSte7 was reduced to a similar level in K. marxianus and K. lactis. The functional replaceability of KlSte7 in K. marxianus was 61% (Figure 3E), while the functional replaceability of KmSte7 in K. lactis was 169% (Figure 3F), suggesting the intrinsic property of KmSte7 in promoting mating was superior to that of KlSte7.

Among Ste5 and three associated kinases, Fus3 was the most conserved protein in terms of the full-length sequence identity. However, the functional replaceability of ScFus3 and KlFus3 in K. marxianus was 23 and 27%, respectively, which was among the lowest functional replaceability observed in this study (Figures 3C,E). Therefore, the extent of full-length sequence identity did not perfectly match the extent of functional replaceability. Regarding the identities between domains, the hierarchy of the average identity between the kinase domains in Fus3, Ste11, and Ste7 of K. lactis, K. marxianus, and S. cerevisiae was Ste11 > Ste7 > Fus3 (Figure 2C). The same hierarchy of functional replacement of three kinases was observed in the complementation assay in K. marxianus (Figures 3C,E). Therefore, compared to the identities between full-length protein, the identities between kinase domains better matched the extent of functional replaceability of MAPKs. However, there were exceptions. The identity between the kinase domain of KmSte11 and KlSte11 (93.8%) was slightly higher than that of KmSte7 and KlSte7 (93.5%), but the replaceability of KmSte11 in K. lactis was only half of that of KmSte7 in K. lactis (Figure 3F). Thus, the extent of sequence identity, either between full-length proteins or between domains, did not necessarily indicate the extent of functional replaceability.

Ste5 was the least conserved protein, in terms of full-length protein and domains (Figure 2C). However, Ste5 obtained the highest average functional replaceability in four complementation pairs, supporting the conclusion above. The functional replaceability of ScSte5 in K. marxianus and K. lactis was 116 and 102%, respectively (Figures 3C,D). Notably, swapping KmSte5 and KlSte5 in K. marxianus and K. lactis significantly promoted mating efficiency in K. marxianus and K. lactis to 247 and 137%, respectively (Figures 3E,F). The mechanism underlying this unusual phenotype was investigated afterward.

Significant improvement of mating efficiency by replacing native protein with its ortholog was not only observed in the cases of Ste5, but also in the replacement of KlSte7 by KmSte7 (Figure 3F). qRT-PCR analysis indicated that mRNA levels of STE5 or STE7 orthologs were similar to those of native counterparts (Supplementary Figure 1). It suggested that improved mating efficiency was not due to the change of expression levels.

The function of a protein is largely determined by its interactions with other proteins. In S. cerevisiae, the strongest interaction among Ste5 and associated kinases was between Ste5 and Ste7 (Choi et al., 1994). The change of Ste5-Ste7 interaction might underlie the improved mating efficiency by replacing natural Ste5 or Ste7 with its orthologs. To test this hypothesis, interactions between Ste5 and Ste7 of K. marxianus, K. lactis, and S. cerevisiae was measured by a quantitative β-galactosidase assay in a yeast two-hybrid system (Figure 4A).

When the relative mating efficiencies after replacing natural Ste7 with its ortholog were plotted against the relative strength of the interaction between Ste7 ortholog and natural Ste5, a negative correlation was observed (R² = 0.85) (Figure 4B). It suggested that Ste5-Ste7 interaction negatively regulated the mating efficiency in K. marxianus and K. lactis. In S. cerevisiae, the C terminal portion of Ste5 (744–895 aa) was required to interact with Ste7 (Inouye et al., 1997). The Ste7-binding region was recognized in KmSte5 and KlSte5 by sequence similarity (Figure 4C and Supplementary Figure 2). To investigate the relationship between Ste5-Ste7 interaction and mating efficiency, the Ste7-binding region in KmSte5 was replaced by that from KlSte5 and the resultant chimeric protein was named KmSte5-KlBD (BD was short for binding-domain). mRNA level of KmSTE5-KlBD was similar to that of KmSTE5 (Supplementary Figure 1). The strength of the interaction between natural KmSte5 and KmSte7 was 22.78, and that between KmSte5-KlBD and KmSte7 was 14.92. While the relative mating efficiency of wild-type K. marxianus cells is 100%, that of the K. marxianus cells carrying Ste5-KlBD increased to 232% (Figure 4E). In a reciprocal test, the Ste7-binding region in KlSte5 was replaced by that from KmSte5 to obtain KlSte5-KmBD (Figure 4D). mRNA level of KlSTE5-KmBD was similar to that of KmSTE5 (Supplementary Figure 1). The strength of the interaction between KlSte5 and KmSte7 was 8.13, and that between KlSte5-KmBD and KmSte7 increased to 11.04. The relative mating efficiency of K. marxianus cells carrying KlSte5 was 247% and that number reduced to 164% in the cells carrying KlSte5-KmBD (Figure 4E).

In K. lactis, KlSte5 was also replaced by two chimeric Ste5, separately. mRNA levels of KlSTE5-KmBD and KmSTE5-KlBD in K. lactis was similar to that of KlSTE5 (Supplementary Figure 1). The interaction strength between KlSte5 and KlSte7 was 13.74, and that between KlSte5-KmBD and KlSte7 increased to 20.30. The relative mating efficiency of K. lactis cells carrying KlSte5-KmBD dropped to 74%, compared to wild-type K. lactis. The strength of the interaction between KmSte5 and KlSte7 was 29.99, and that between KmSte5-KlBD and KlSte7 increased to 31.77. The relative mating efficiency of K. lactis cells carrying KmSte5 was 137%, and that of cells carrying KmSte5-KlBD reduced to 66% (Figure 4E). The results suggested that, in K. marxianus and K. lactis, the strengthened interaction between Ste7 and Ste5 attenuated mating, while the weakened interaction improved mating.

In S. cerevisiae, a Ste5 (V763A, S861P) mutant that was unable to bind Ste7 displayed a clear mating defect, indicating Ste5-Ste7 interaction is required for the positive output of the pheromone pathway (Inouye et al., 1997). This conclusion does not necessarily contradict our assumption that Ste5-Ste7 interaction is negatively correlated with mating efficiency. Ste5 is composed of modular interaction domains, some of which mediate activating or inhibitory regulation with the same binding partner (Good et al., 2011). For example, ScSte5 uses both VWA and FBD domains to interact with ScFus3. The former interaction is required for the allosteric activation of ScFus3 by ScSte7 (Good et al., 2009), while the latter one is believed to attenuate the pathway output (Bhattacharyya et al., 2006). Therefore, Ste5 may use specific motifs inside the Ste7-binding region to mediate inhibitory regulation of the mating pathway. In the Ste7-binding region of ScSte5, a negatively charged segment (DEHDDDDEEDN) was located in the overlapped region of the VWA domain and Ste7-binding domain (Figure 5A). Notably, major differences between the Ste7-binding region of three Ste5 orthologs were around the negatively charged segment (Supplementary Figure 2). In KmSte5, the last asparagine of the segment was changed into a non-polar glycine and seven residues downstream the segment was absent. In KlSte5, half of the segment, as well as the downstream residues were absent (Figure 5B). To investigate the relationship between the segment and improved mating efficiency, the segment of ScSte5 and KmSte5 was reduced to the same length of that of KlSte5 to obtain ScSte5ΔS (S was short for the segment) and KmSte5ΔS, respectively. Meanwhile, the segment of KlSte5 was extended to mimic the segment of KmSte5 and the resultant mutant was named KlSte5-ES (ES was short for the extended segment). mRNA levels of ScSTE5ΔS, KmSTE5ΔS, and KlSTE5-ES in K. marxianus and K. lactis were similar to those of native counterparts (Supplementary Figure 1).

The relative mating efficiency of K. marxianus cells carrying KmSte5ΔS was 2 times higher than that of cells carrying wild-type Ste5. Similarly, the mating efficiency of K. marxianus cells carrying ScSte5ΔS was 44% higher than that of cells carrying ScSte5. Therefore, shortening the negatively charged segment in KmSte5 and ScSte5 improved the mating efficiency in K. marxianus (Figure 5C). On the other hand, the relative mating efficiency of K. marxianus cells carrying KlSte5-ES was 40% lower than that of cells carrying KlSte5. Therefore, extending the negatively charged segment in KlSte5 reduced the mating efficiency in K. marxianus (Figure 5C).

Similar results were observed in K. lactis. The relative mating efficiency of K. lactis cells carrying ScSte5ΔS was 45% higher than that of cells carrying ScSte5. Notably, the relative mating efficiency of K. lactis cells carrying KmSte5ΔS was 4 times higher than that of cells carrying KmSte5 (Figure 5D). Meanwhile, the relative mating efficiency of K. lactis cells carrying KlSte5-ES was 26% lower than that of cells carrying wild-type KlSte5 (Figure 5D).

The negatively charged segment composed a binding interface between Ste5 and Ste7. In S. cerevisiae, deletion of this region interfered with the ScSte5-ScSte7 interaction and dramatically diminished Ste7-to-Fus3 phosphorylation which was a positive output of the pheromone pathway (Good et al., 2009). Unexpectedly, ScSte5ΔS displayed improved interaction with KmSte7 and KlSte7 (Figures 5C,D). The opposite effects of shortening the segment of ScSte5 on its interaction with Ste7 might be related to the divergence between ScSte7 and KmSte7/KlSte7. Similarly, shortening the segment in KmSte5 also improved its interaction with KmSte7 and KlSte7 (Figures 5C,D), suggesting the function of the segment in ScSte5 and KmSte5 was conserved. Meanwhile, KlSte5-ES displayed improved interaction with KmSte7 and KlSte7 (Figures 5C,D). The results suggested that extending the length of the segment in KlSte5 improved its interaction with Ste7 and that might contribute to the reduced mating efficiency.

In summary, our results suggested that the length of a negatively charged segment in the Ste7-binding domain of Ste5 was negatively correlated with mating efficiency in Kluyveromyces. Extending the segment of KlSte5 improved its interaction with KmSte7/KlSte7, while shortening the segment of KmSte5 and ScSte5 also improved their interaction with KmSte7/KlSte7. Thus, the segment of KlSte5, not those of KmSte5 and ScSte5, might mediate the relationship between attenuated Ste5-Ste7 interaction and improved mating.