Revisiting the role of yeast Sfp1 in ribosome biogenesis and cell size control: a chemostat study

Physiology of the sfp1Δ and reference strains in aerobic ethanol- and anaerobic glucose-limited chemostat cultures

While S. cerevisiae CEN.PK 113-7D has a maximum specific growth rate (μ_max) of 0.40 h⁻¹ when aerobically growing in shake flasks on minimal medium with glucose as carbon source (Flikweert et al., 1999), its sfp1Δ derivative has a much slower μ_max of 0.16 h⁻¹. To prevent growth-rate-related variations obscuring data interpretation when comparing the physiology of the reference and the sfp1Δ mutant, both strains were grown at an identical growth rate in chemostat cultures. The reference strain CEN.PK 113-7D and the isogenic sfp1Δ mutant were therefore grown in chemostat cultures under ethanol- and glucose (anaerobic)-limitation.

Under this cultivation set-up, fresh medium continuously flows into the fermenter, while the same amount of culture broth flushes out so that the operative volume is kept constant. Therefore, the culture is continuously diluted by a factor called the dilution rate (D, h⁻¹). The steady state is reached when the measured growth parameters (in this specific case metabolite consumption/production rate, biomass concentration and budding index) become constant. At steady state, the population is in exponential balanced growth, and the specific growth rate for each growing cell in the population [μ=ln(2) (duplication time)⁻¹] equals the dilution rate (Alberghina et al., 1998; Monod, 1950; Novick & Szilard, 1950).

A major advantage of chemostat cultures compared to shake flasks is that the nutritional/environmental conditions (pH, aeration, specific growth rate and agitation) can be set and tightly controlled by the user. In addition, very slow growth rates can be reached while maintaining continuous exponential growth and an optimal physiology (Regenberg et al., 2006; Saldanha et al., 2004). This approach allowed us to grow the reference strain and the sfp1Δ mutant at the same specific growth rate. This was obtained by always setting an identical dilution rate for both the reference and the mutant strain. Similarly, all the other cultivation parameters were always identical for the two strains (pH, medium composition, temperature and aeration).

By changing the nutrient supply and the aeration setting, two metabolic conditions were analysed, i.e. ethanol (aerobic)- and glucose (anaerobic)-limitation. These two different set-ups were chosen as paradigms for two very different metabolic states, fully respiratory (with no glucose present) and fully fermentative, respectively (Cortassa et al., 1995; Daran-Lapujade et al., 2004; Kappeli, 1986). It would have been of interest to study the condition of respiratory metabolism during aerobic glucose-limitation. However, the mutant strain exhibited sustained metabolic oscillations under this growth condition, making it impossible to reach the required steady state (data not shown).

For each growth condition the dilution rate (D), which is equal to the specific growth rate (μ), was chosen to be close to the maximum growth rate of the sfp1Δ mutant for that specific condition. Ethanol-limited cultures were grown at D=0.10 h⁻¹ while glucose (anaerobic)-limited cultures were grown at D=0.05 h⁻¹, corresponding to a duplication time of 6.9 and 13.8 h, respectively.

As shown in Table 1⇓, the two strains showed a very similar behaviour. As expected, during growth on ethanol both strains displayed a fully respiratory metabolism with a respiratory coefficient (RQ) of ∼0.47 and a yield of ∼13 g biomass (carbon moles)⁻¹. No trace secondary metabolites were found in the medium. These data were consistent with previously published findings for the reference strain and indicated that ethanol was completely metabolized (Daran-Lapujade et al., 2004). During anaerobic glucose-limited growth we observed, as expected, significant production of glycerol and ethanol. Ethanol was generated through alcoholic fermentation as the sole mode of glucose dissimilation. Production of glycerol was required for the non-respiratory oxidation of NADH. As expected, the biomass yield on glucose under this growth condition was very low due to the low ATP yield from alcoholic fermentation. Under both steady-state conditions, the sfp1Δ mutant and the reference strain exhibited very similar phenotypes. The specific substrate and oxygen consumption rates (

and ) as well as the specific production rate of carbon dioxide () and secondary metabolites showed very similar values for both strains. Similarly, the yield of biomass on carbon was identical, indicating a similar efficiency of substrate utilization and a comparable flux distribution.

Deletion of SFP1 causes major transcriptional changes during growth on glucose

A genome-wide transcriptional analysis was performed for the reference and the sfp1Δ strains for each growth condition. All data presented in this work were derived from three independent chemostat cultures. The average cv for all genes for triplicate experiments was below 0.19 for each of the four conditions. Furthermore, the levels of ACT1 and PDA1, common loading standards for conventional Northern analysis, varied by less than 9 % over the different cultures. For all tested conditions 5828 transcripts (∼95 % of annotated ORFs) were above the detection limit. Comparison of transcript levels between the sfp1Δ and the reference strains revealed, for each of the two conditions, the genes whose expression significantly changed (SAM, FC threshold=2 and FDR=1 % as described in Methods). In this study we will refer to the genes whose expression levels were higher and lower in the mutant strain compared to the reference strain as ‘up regulated’ and ‘down regulated’ , respectively. This terminology does not imply any hypothesis regarding regulatory mechanisms.

Comparison of the sfp1Δ and the reference strain during ethanol-limited growth revealed 91 transcripts whose expression significantly changed (33 up- and 58 downregulated; see Supplementary Table S1, available with the online version of this paper). No enrichment for functional categories or features (localization, complexes, etc.) could be identified among genes either up- or downregulated under this condition. Very similar results were obtained after lowering the FC threshold to 1.5 when performing SAM analysis.

During anaerobic glucose-limited growth 217 genes displayed significantly changed expression (see Supplementary Table S2, available with the online version of this paper), 62 being upregulated and 155 downregulated (Fig. 1a⇓). Again, no enrichment for functional categories was found among the genes that were upregulated. However, close to 50 % of the downregulated genes (72 genes) belonged to only four different functional categories that were significantly enriched compared to their corresponding genome-wide distribution (see table in Fig. 1a⇓). The ribosome biogenesis cluster (‘RiBi regulon’) was the largest, containing genes involved in processing of rRNA and assembly and transport of mature ribosomal subunits (Jorgensen et al., 2004). Genes encoding translation initiation factors (TIF) and RNA polymerase I (RNA Pol I) were also significantly represented, as was the pyrimidine nucleotide metabolism (PNM) cluster. The remaining 83 downregulated genes represented miscellaneous features. Surprisingly, no ribosomal protein (RP) genes were present among the downregulated sequences under any of the analysed conditions.

In the first round of analysis, SAM software was used to find significant changes (see Methods) because it allowed a stringent statistical control over the complete dataset. Once the over-represented functional categories were found, we focused on these categories to try and find if more genes belonging to them were downregulated, although to a lesser extent, in the complete dataset (all the genes above the expression threshold were considered; see Methods). For this we performed a t-test on all those genes that showed a FC lower than −1.5. Thus, a significant enrichment was observed for each of the four functional groups (Fig. 1b⇑). The same analysis was performed for the RP category, but again this cluster was not significantly represented (Fig. 1b⇑).

RiBi and RP are independent transcriptional clusters

The lack of downregulation of RP genes in the sfp1Δ strain during glucose-limited growth was surprising. Indeed, previous data suggested that Sfp1 was involved in both RiBi and RP transcriptional regulation (Jorgensen et al., 2004; Marion et al., 2004). To obtain a quantitative measure of such a different behaviour we compared our dataset of anaerobic glucose-limited growing cultures with a published dataset, referring to an sfp1Δ and an isogenic reference strain growing on rich medium in shake flasks (Jorgensen et al., 2004). As shown in Fig. 2⇓, in both datasets RiBi, TIF, RNA Pol I and PNM gene clusters emerged as the functional categories highly represented among the downregulated genes. However, shake flask cultures of the sfp1Δ mutant showed a significant downregulation of the RP cluster compared to the reference strain (123 of 129 RP genes showed a FC lower than −1.5). In our dataset, only 11 RP genes of the 114 ORFs displayed a FC below −1.5. It is important to note that in our experimental setup, as previously described, the reference and the mutant strains were growing at the same specific growth rate. This was not the case for the dataset derived from cultures growing in shake flasks using rich medium. Under these conditions the mutant strain grew much more slowly than the reference strain (Jorgensen et al., 2002). Our data suggest that the RP and RiBi genes may be regulated, at least partly, through independent mechanisms. In particular, Sfp1 appeared responsible for the transcriptional regulation of genes involved in ribosome biogenesis (RiBi, TIF, RNA Pol I and PNM) but not of RP genes.

The existence of such a differential regulation was further underlined by an in silico promoter analysis. The promoters of the 155 genes that were downregulated during glucose-limited growth were specially enriched for the presence of RRPE and PAC elements [RRPE=GAAAAWWTT, PAC=GMRATGARNT (Hughes et al., 2000); see Methods]. It has been previously shown that these elements are typical of the gene clusters involved in ribosome biogenesis, while they are not present in RP gene promoters (Hughes et al., 2000; Wade et al., 2001; Wang et al., 2004). We found that among the 155 analysed promoters, 106 and 55 contained RRPE and PAC, respectively, suggesting the existence of a functional link between Sfp1 activity and the genes containing these promoter elements. Finally, it should be noted that not all the genes that were downregulated and that contained RRPE and PAC elements in their promoters were involved in ribosome biogenesis. These genes represented miscellaneous features whose phenotype, if any, did not clearly emerge from our analysis.

Analysis of cell size and cell cycle parameters

In glucose-grown batch cultures the main phenotypic trait resulting from SFP1 deletion is a very reduced cell size (whi phenotype) accompanied by a slow growth rate (Cipollina et al., 2005; Jorgensen et al., 2002). This size phenotype is less evident during batch growth on ethanol, when the growth rate defects of the mutant are marginal (Cipollina et al., 2005). We wondered whether the whi phenotype was maintained in continuous cultures where nutrient-dependent modulation of cell size can be evaluated separately from growth-rate-dependent variables. Therefore the average cell size for each strain and growth condition was measured by a flow cytometric approach on FITC-stained cells (see Methods). This gives quantitative information on the protein content distribution of the cells in the analysed populations. It has been reported previously that this parameter is a good estimate of the cell size (Alberghina & Porro, 1993; Vanoni et al., 1983).

During ethanol-limited growth the average size was 285.0 and 174.3 for the reference and mutant strain, respectively (the cell size is expressed as channel number, relative units; Fig. 3a⇓). During glucose-limited growth the average size of the reference and the mutant strains was 326.0 and 254.2, respectively (Fig. 3b⇓). Surprisingly, under both growth conditions the protein distributions of the mutant strain were shifted towards lower values. Thus, the sfp1Δ strain was characterized by a lower general ‘size setting’ compared to the reference population, with smaller dimensions both at birth and at division [dividing and newborn cells are mainly represented in the right- and left-hand sides of the histograms, respectively (Porro et al., 1997)]. The cell volume distributions, measured with a Coulter counter analyser, were in good agreement with the cell protein distributions (data not shown).

The bimodal distribution of the reference population growing under ethanol limitation (Fig. 3a⇑) reflects a high heterogeneity in cell size (cv=0.38) and a population structure which is typical of budding yeast growing slowly in poor media with a low budding index (BI) (Alberghina et al., 1991; Vanoni et al., 1983). The sfp1Δ strain growing on ethanol showed a right-skewed size distribution reflecting a comparable degree of heterogeneity (cv=0.43), although a true bimodal shape was not detected.

To assess whether defects in size setting of the mutant strain correlate with alterations in cell-cycle progression we determined the percentages of cells in G1 and in S+G2+M phases in our cultures. We used two independent approaches: manual determination of BI and flow cytometric analysis. The BI represents the percentage of budded cells in the population which are in the S+G2+M phase, with a DNA content greater than 1C. The remaining part of the population, namely the unbudded cells, is in the G1 phase, with a DNA content of 1C. The flow cytometric analysis was performed on PI-stained cells, whose fluorescence intensity is proportional to the DNA content. The percentage of G1 cells was determined from dot plots of FSC and PI fluorescence on the x and y axes respectively. As shown in Table 2⇓, the two approaches gave very similar results. The BI was ∼28 % for ethanol-limited cultures and ∼50 % for glucose (anaerobic)-limited cultures. This latter value was unexpectedly high; however, high BI values appear to be typical of populations growing under anaerobiosis (L. Brambilla, personal communication). From the BI it was also possible to calculate the duration of the budded and unbudded phases (Vanoni et al., 1983) (Table 2⇓). Intriguingly, all the data clearly indicated that while cell cycle parameters changed according to the two conditions, the sfp1Δ mutant did not exhibit any significant difference in the distribution throughout the cell cycle phases compared to the reference strain, in particular in the G1 phase where the main control over cell size takes place. Therefore, these findings indicate that the size defects and the alterations in cell cycle progression can be ‘uncoupled’ and are fully consistent with previous publications showing that an extended G1 phase does not always result in entering Start at a smaller cell size (Jorgensen & Tyers, 2004; Popolo et al., 1982).

Continuous cultures: subsiding growth rate effects

The advantages of using chemostat cultures for physiological and metabolic studies have been widely recognized (Brauer et al., 2005; Hoskisson & Hobbs, 2005; Porro et al., 2003; Wu et al., 2006). Nevertheless, they have not yet found wide application in molecular and cell biology.

One of the advantages of this approach is that cells are continuously growing exponentially, even at very slow growth rates such as those used in this study. No evidence of starvation characterizes these slow-growing cultures and the transcription of ribosomal protein genes, an indicator of cellular fitness, is fully active (see Supplementary Table S3, available with the online version of this paper) (Regenberg et al., 2006; Saldanha et al., 2004).

The use of chemostat cultures was the only strategy that allowed us to grow the reference and the sfp1 null mutant strains at the same, albeit slow, specific growth rate, with any desired carbon source, while maintaining an optimal cell physiology.

Data reported here, while confirming some previous findings on the role of Sfp1 in yeast physiology, revealed new features which otherwise would have remained masked by growth-rate-dependent effects. In fact, the slow-growing phenotype of an sfp1Δ mutant during growth in shake flasks ‘per se’ generates a number of effects that are difficult to uncouple from those directly generated by the absence of SFP1. Our findings indicate that the mutant and reference strains share a similar physiology when growing at the same specific growth rate. Consistently, the absence of SFP1 did not affect the efficiency of carbon utilization, the activity of the fermentative and respiratory pathways or the global flux distributions. Therefore, the chosen cultivation approach was revealed to be a good starting point for further studies concerning the activity of Sfp1 as a transcriptional regulator and modulator of size setting in response to different nutrients.

Sfp1 is a positive regulator of ribosome biogenesis during growth on glucose

Previous studies have suggested that Sfp1 is a transcriptional activator of the RiBi gene family and of RP genes (Jorgensen & Tyers, 2004; Marion et al., 2004; Wang et al., 2004). We observed the downregulation of the RiBi cluster, but not of RP genes, in glucose-limited sfp1Δ cultures, indicating the involvement of Sfp1 in the regulation of RiBi but not RP genes (Fig. 2⇑).

This finding was recently confirmed by additional experiments in which glucose pulses were added to carbon-limited cultures of both the sfp1 null mutant and the reference strain. The dynamics of the transcriptional response induced by such perturbation were analysed over time. Preliminary results indicated that the primary transcriptional target of Sfp1 is indeed the RiBi cluster and not the RP genes. Furthermore, these data strongly indicated that the lack of involvement of SFP1 in RP gene transcriptional regulation was not due to the slow growth rate applied in the present work. On the contrary, this phenomenon was evident after the glucose pulse when the growth conditions were comparable to growth in shake flasks (preliminary data, not shown).

At first glance these data appeared to be in contrast with previously reported results for an sfp1Δ strain. However, they were consistent with the observation that overexpression of SFP1 results in a delay between the induction of the RiBi and the RP genes, which gave the first indication of the Sfp1-dependent transcription of the RiBi cluster (Jorgensen et al., 2002).

For the RP genes, Rap1 together with the Fhl1–Ifh1 complex is the main transcriptional regulator (Klein & Struhl, 1994; Martin et al., 2004; Rudra et al., 2005). The activity of all these factors is regulated via partially overlapping pathways (involving TOR and PKA), allowing a coordinate expression of all the components required for ribosome biosynthesis. This explains the similar behaviour that the RiBi and RP gene clusters usually show. However, the use of independent but parallel and partially redundant regulatory pathways is often strategically employed by cells to fine-tune transcriptional regulation. We hypothesize that this is the case for the RiBi and RP clusters. Downregulation of the RP genes observed in sfp1Δ mutants during growth in shake flasks might be a secondary effect due to growth rate defects. Finally, several expression profiles published elsewhere (Saldanha et al., 2004; Wade et al., 2001) together with the observed differences in the promoter regions lend further support to our hypothesis on the differential regulation of RiBi and RP gene clusters.

The observed expression defects of the genes related to rRNA transcription (the RNA Pol I cluster) and processing (RiBi cluster) might well explain the phenotype of the mutant strain. This is characterized by low amounts of rRNA and a delay in rRNA processing steps (Fingerman et al., 2003). Moreover, consistent with previous data on Sfp1 localization (Jorgensen et al., 2004; Marion et al., 2004), our findings indicated that during growth on glucose, but not on ethanol, Sfp1 acts as activator of RiBi cluster transcription, thus sustaining the growth potential of the cell.

The signal inducing the transcriptional activity of Sfp1 in response to the available carbon source remains to be elucidated. In anaerobic glucose-limited chemostats, while glucose is efficiently transported and a sustained glycolytic flux is active inside the cell, the extracellular residual glucose concentration is low (<0.3 mM) compared to the K_m of the high-affinity sensing systems (Ozcan et al., 1996; Rolland et al., 2000). This suggests that high levels of extracellular glucose are not required for stimulating Sfp1 activity. The positive signal could be mainly generated at the intracellular level and sustained by fermentative metabolism.

The mechanisms through which Sfp1 exerts its transcriptional control remain unknown. Attempts to prove physical interaction of Sfp1 with promoters of RiBi genes have failed (Fingerman et al., 2003; Jorgensen et al., 2004). So far, only weak interactions with some RP gene promoters have been identified, but this point is still controversial (Fingerman et al., 2003; Jorgensen et al., 2004; Marion et al., 2004). It is possible that Sfp1 acts through indirect mechanisms, modulating the chromatin structure or acquiring and regulating other activators. Similar mechanisms of transcriptional control have recently been shown for other important factors involved in RP and rRNA gene regulation (Li et al., 2006; Rohde & Cardenas, 2003; Tsang et al., 2003). Such strategies would allow coordinated transcriptional control and be compatible with the low expression levels of Sfp1 itself (Ghaemmaghami et al., 2003).

Deletion of SFP1 causes a small size phenotype

We have shown that SFP1 is involved in cell size control. This role seems to be independent of the metabolic status of the cell and the growth rate. In fact, the sfp1Δ mutant showed a whi phenotype during growth on both ethanol and glucose. This suggests that the control of size setting might act through a different pathway compared to the transcriptional control of the RiBi cluster that is promoted by Sfp1 mainly during glucose-limited growth. We specifically searched for common genes or genes with a shared function with expression patterns that were significantly changed under both growth conditions; no statistically valid results were obtained that could explain the observed whi phenotype. However, an inspiring suggestion emerging from this analysis was related to the expression of two RNA Pol I genes, RPA49 and RPA14. It has been previously shown that the deletion of each of them caused a strong whi phenotype (Jorgensen et al., 2002). We observed that RPA14 is downregulated in sfp1Δ cultures during ethanol-limited growth (FC=−3.2) but not during glucose-limited growth (FC=−1.2). Vice versa, RPA49 is downregulated during glucose- but not ethanol-limited growth (FC=−2.4 and 1.0, respectively). These two RiBi genes, encoding potential Start repressors (Jorgensen et al., 2002), could be the link between the observed transcriptional alterations of the sfp1Δ mutant and the small size phenotype.

To see whether the size defects might be connected with alterations in cell cycle progression we performed a more detailed analysis of the cell cycle phases. Surprisingly, no alterations were found in the mutant strain, indicating that deletion of SFP1 can affect size setting without altering the distribution throughout the cell cycle. It has been hypothesized that an intimate link exists between ribosome biogenesis, cell cycle control and size setting (Jorgensen et al., 2004; Sudbery, 2002). Our results suggest that these phenotypes can be partially uncoupled if the growth rate defects of the mutant are removed. We suggest that Sfp1 may act at two regulatory levels, modulating ribosome biogenesis and size setting. This is further supported by the observation that the presence of defects in ribosome biogenesis is not enough to explain the generation of a whi phenotype (Rudra et al., 2005).

Our results, together with previous findings, provide strong evidence that Sfp1 plays a key role in linking growth potential with cell size control. A future challenge lies in further studying the molecular mechanisms that are at the base of this control in order to understand the complex regulatory networks that, in all eukaryotes, govern growth, size setting and cell-cycle progression.