Urea is a drop-in nitrogen source alternative to ammonium sulphate in Yarrowia lipolytica

Summary Media components, including the nitrogen source, are significant cost factors in cultivation processes. The nitrogen source also influences cell behavior and production performance. Ammonium sulfate is a widely used nitrogen source for microorganisms’ cultivation. Urea is a sustainable and cheap alternative nitrogen source. We investigated the influence of urea as a nitrogen source compared to ammonium sulfate by cultivating phenotypically different Yarrowia lipolytica strains in chemostats under carbon or nitrogen limitation. We found no significant coherent changes in growth and lipid production. RNA sequencing revealed no significant concerted changes in the transcriptome. The genes involved in urea uptake and degradation are not upregulated on a transcriptional level. Our findings support urea usage, indicating that previous metabolic engineering efforts where ammonium sulfate was used are likely translatable to the usage of urea and can ease the way for urea as a cheap and sustainable nitrogen source in more applications.


INTRODUCTION
Over the past decade, nonconventional yeasts as host organisms for biotechnology applications have been on the rise. Advanced genetic engineering tools and decreasing prices for omics analysis allow more studies on non-model organisms. The oleaginous yeast Yarrowia lipolytica is one example of such an organism. The initial interest was focused on the outstanding lipid production capacity of Y. lipolytica, 1,2 which can be increased to up to 90% of its dry weight. 3 However, this host has also been shown to be suitable for non-lipid products, e.g. terpenoids such as carotenoids and polyketides like flavonoids. 4 For the production of these molecules, fat-free strains have been engineered that lack the essential genes for storage lipid production. [5][6][7] Oleaginous yeasts, such as Y. lipolytica, can accumulate at least 20% of their dry weight as lipids in wild-type strains. One of the primary triggers for lipid production is nitrogen starvation. Nitrogen limitation (high C/N ratio) induces adenosine monophosphate (AMP) depletion by AMP deaminase, which inhibits the isocitrate dehydrogenase of the tricarboxylic acid (TCA) cycle. This inhibition leads to an excess of citrate in the mitochondria, which is transported into the cytosol by malate/citrate transferase. In the cytosol, citrate is cleaved to form acetyl-coenzyme A (CoA) by ATP-citrate lyase and is further converted to fatty acids. 8 Together with the carbon source, the nitrogen source is one of the main cost factors in the media of largescale cultivation. 9 The most widely used nitrogen source in microbial cultivation is ammonium sulfate, produced by sulfuric acid treatment of ammonia. Industrial ammonia production is mostly via the energy-and carbon-intense Haber-Bosch process that fixes nitrogen with hydrogen at high temperature and pressure. 10 Urea is an interesting alternative that is more sustainable since it can be extracted from urine, and cost calculations showed a lower cost per mol nitrogen for urea than that for ammonium sulfate. 11 In addition, urea can be produced from municipal solid waste in an economical and environmental way. 12 In contrast to ammonium sulfate, urea utilization does not acidify the media, which requires less base addition during pHcontrolled large-scale cultivation. However, changing the media composition can affect cell behavior and the production performance of microorganisms. [13][14][15][16] In Y. lipolytica, the nitrogen source was also linked to dimorphic growth. 17 To further understand the cells' reaction to different nitrogen sources, it is not enough to limit observation to single parameters (e.g., lipid content, hyphenation, and growth rate); we need to monitor the cell as a whole. One approach for studying cell behavior on a genome-wide level is transcriptomic analysis. Transcriptomics maps the abundance of all mRNA molecules of a cell (i.e. the transcriptome), thereby determining the total gene expression. Comparison between different conditions can then reveal changes in cell behavior.
In this study, we performed transcriptomic analysis of Y. lipolytica cultivations where we varied three variables: (a) we used three strains differing in their lipid accumulation ability; (b) cultivated in either carbonor nitrogen-limiting conditions; (c) with urea or ammonium sulfate as nitrogen source. With this experimental setup, we aimed to investigate whether a response to the nitrogen source might differ depending on the amount of lipid accumulation.

RESULTS
From the Y. lipolytica genome, potential homologs could be identified to reconstruct the ammonium and urea assimilation pathways. The pathways are not dissimilar from the model yeast Saccharomyces cerevisiae, indicating a conserved mechanism, although the number of homologous genes occasionally differed (e.g. DUR3, Table S1). Briefly, ammonium and urea are transported into the cell by MEP1,2,3 and DUR3, respectively ( Figure 1). Ammonium dissociates into ammonia, and the released proton is transported back into the medium by the plasma membrane H + -ATPase (PMA1) under the consumption of one ATP per proton. 18 Urea is converted into two ammonia molecules by a urea amidolyase (Dur1_2). DUR1_2 is a multifunctional enzyme with urea carboxylase and allophanate hydrolase activity. The first activity is biotin-dependent and converts urea into allophanate under the consumption of one ATP and bicarbonate. The second activity consumes water and releases CO 2 and two ammonia molecules, making this process more energy efficient than the usage of ammonium from the media. 19 Ammonia is then incorporated into glutamate and glutamine by the NADP-dependent glutamate dehydrogenase (GDH1) and glutamine synthetase (GLN1), respectively. Glutamate and glutamine both serve as the starting points for amino acid production. In addition, glutamate can be converted back by the NAD-dependent glutamate dehydrogenase (GDH2) to ammonia and a-ketoglutarate, which links the nitrogen metabolism to the TCA cycle. 20 The gene expression of both nitrogen pathways has been shown to be regulated by the available nitrogen source in S. cerevisiae. 21,22 Cell physiology does not change with urea compared to ammonium sulfate as a nitrogen source We investigated the effect of urea versus ammonium sulfate as a nitrogen source in three different Y. lipolytica strains. The strain OKYL049 is a genetically engineered obese strain carrying DGA1 overexpression (YALI1_E38810g) and are1 deletion (YALI1_F09747g). 23 The strain JFYL007, 24 also referred to as Q4 strain, is incapable of synthesizing the storage lipid triacylglycerol or sterol esters due to the deletion of four genes, Dare1 (YALI1_F09747g), Dlro1 (YALI1_E20049g), Ddga1 (YALI1_E38810g), and Ddga2 (YALI1_D10264g). 5 OKYL029 25 has no modifications of the lipid production and displays iScience Article a wild-type lipid phenotype. In all strains, hyphae formation was abolished by deleting MHY1 (YALI1_B28150g) to ease the bioreactor cultivations and ensure similar cell morphology despite different culturing conditions. 25 We performed chemostat cultivations at a dilution rate of 0.1 in a minimal medium under carbon or nitrogen limitation (C/N ratio 3 or 116) with urea or ammonium sulfate as equimolar nitrogen sources. The pH was maintained at a value of 5 by the addition of potassium hydroxide. In contrast to ammonium sulfate, urea requires significantly less base addition during cultivation ( Figure S1), which can further lower process costs.
The cell physiology was largely unaffected by the nitrogen source in our chemostat cultivations ( Table 1). The nitrogen assimilation from ammonium sulfate has a higher cost (1 ATP per ammonia) compared to urea (½ ATP per ammonia). We reasoned that this might have an impact on the biomass, lipid content, their corresponding yields, or the specific uptake rate of glucose (r-Glucose). However, only the obese strain (OKYL049) showed significant changes in biomass in nitrogen limitation. The other strains showed no statistically significant changes (p value <0.01) in any of the measured parameters between ammonium sulfate and urea.
The fatty acid composition is often of interest when Y. lipolytica is applied to produce lipid derivatives, e.g. food oils. 26 In these cases, nitrogen starvation is used to trigger lipid production to achieve high lipid titers. Under nitrogen-limiting conditions, the obese strain (OKYL049) showed significant changes in C16:0 and C18:2 between the two nitrogen sources. In addition, we observed a significant change toward lower saturation and longer chain length (C16/C18) in urea compared to ammonium sulfate. The storage-lipidfree Q4 strain (JFYL007) showed significant changes in the C18:0 and C18:1, which did not result in an overall significant change in saturation or chain length of the fatty acids.
Under carbon-limiting conditions (C/N ratio 3), storage lipid production is not triggered and the three strains have a similar lipid content. Therefore, most of the extracted fatty acids are expected to originate from phospholipids (mostly lipid membranes) and free fatty acids. Under this condition, we observed a significant change between urea and ammonium sulfate in C16:1 and C18:1 for the strains OKYL029 and JFYL007. For the latter, we also found a significant change toward longer chain length in urea compared to ammonium sulfate. We did not observe any significant changes under carbon limitation in the obese Displayed is the mean G SD of at least three replicates. Significance was calculated between the two nitrogen sources ammonium sulfate (AS) and urea (U), with a two-tailed homoscedastic t-test. a Indicates a significant change (p value <0.01) between the ammonium sulfate and urea conditions. iScience Article strain (OKYL049) between the nitrogen sources. Overall, these results indicate that the fatty acid composition of the storage lipids (mostly triacylglycerol) can be slightly influenced by the nitrogen source, while the membrane and free fatty acid composition are not consistently affected throughout all three strains.
As a next step, we performed RNA-seq analysis to probe whether any transcriptional changes occurred that might influence the phenotype beyond our measured parameters.

Genotype and C/N ratio dominate transcriptional changes
A principal component analysis (PCA) was first performed to assess overall similarities and dissimilarities between the expression profiles of the samples ( Figure 2).
In each of the conditions and strains, the replicates clustered well together, demonstrating the reproducibility of the experimental setup. We found that the nitrogen source ( Figure 2A) only resulted in minor separation across the samples. Meanwhile, as the C/N ratio severely affected cell physiology (e.g. lipid content, Table 1), this also significantly separated the samples in the PCA, with PC1 explaining 59% of the variance. Cell physiology was further affected by the strain genotype, primarily in nitrogen limitation (C/N ratio 116), which was reflected in the triacylglycerol-free strain separating further away from the other two strains. In carbon limitation (C/N ratio 3), the strains showed only minor variance, indicating that the choice of nitrogen source has little effect on gene expression when available in copious amounts. Overall, these results showed that the C/N ratio resulted in the biggest differences between samples while the nitrogen source only attributed to minor variance.

Gene ontology analysis revealed no coherent systemic response to the nitrogen source
To probe whether a systemic regulatory response could be observed, we performed gene ontology (GO) term analysis and compared the resulting GO terms in a heatmap to identify GO terms that are relevant in multiple strains and conditions ( Figures S2-S4).
Similar to the results of differentially expressed genes, we only found a few GO terms that were identified in multiple strains and conditions. We did not find any GO term shared between all six comparisons (three strains and two C/N ratios). For carbon limitation, we identified three GO terms that were found in all three strains: protein-macromolecule adaptor activity, copper ion binding, and yeast-form cell wall. For nitrogen Displayed is the mean G SD of at least three replicates. Significance was calculated between the two nitrogen sources ammonium sulfate (AS) and urea (U), with a two-tailed homoscedastic t-test. a Indicates a significant change (p value <0.01) between the ammonium sulfate and urea conditions.

OPEN ACCESS
Since the C/N ratios had a significant impact on the overall gene expression ( Figure 2B), we were also interested in the clusters of DE genes that shared a coherent response in all three strains under the same condition (carbon or nitrogen limitation) (cluster B and C). Cluster B only contained three uncharacterized proteins, and cluster C contained two uncharacterized proteins as well as an S-(hydroxymethyl) glutathione dehydrogenase (YALI1_F13170g).
As the Q4 strain (JFYL007) showed a very different behavior in contrast to the other two strains ( Figure 2C and Table 1), we further inspected the clusters of genes that shared a coherent response across the more similar strains, OKYL029 and OKYL049, in the same conditions (cluster D and E). 16 of the 21 genes of clusters D and E were uncharacterized proteins. For the remaining proteins, we could not reconcile a coherent response to the change in nitrogen source.
In summary, the low variance identified in the PCA, the low number of DE genes, and the low overlap between them indicate that the nitrogen source (urea or ammonium) has minimal effect on the overall transcriptome.
Genes associated with urea transport and degradation are not significantly induced by urea compared to ammonium sulfate We further explicitly investigated if the expression of genes directly associated with urea metabolism are nitrogen source dependent (Figure 4). Y. lipolytica has three genes that are annotated as homologs of urea amidolyases (YALI1_B19217g, YALI1_E08620g, and YALI1_E41754g). However, none showed a significant iScience Article (adjusted p value 0.05) different expression between the two nitrogen sources in either of the tested C/N ratios. There are four genes with homology to the S. cerevisiae urea transporter DUR3 (YALI1_B05609g, YALI1_C22751g, YALI1_E33888g, and YALI1_E39848g). Gene expression data from two of them (YALI1_C22751g and YALI1_E33888g) were removed during the filtering step because gene counts were too low (both have a lower protein sequence similarity to ScDUR3 than the other homologs). Of the remaining two genes, we only observed a significantly different expression for YALI1_B05609g. Cultivation at C/N ratio 116 with urea instead of ammonium sulfate increased expression of this gene by 2.22-and 2.85fold in the OKYL029 and OKYL049 strains, respectively. A similar trend was found for the ammonium transporter and ammonium permeases. Six genes have homology to S. cerevisiae MEP1,2,3. Also here, gene expression data from two of the genes were removed during filtering (YALI1_A02653g and YALI1_A20201g), while three other genes did not show any significant different expression (YALI1_E32180g, YALI1_F17337g, and YALI1_F22568g) in any strain or C/N ratio. Only YALI1_B18292g showed a modest 0.57-fold increase in strain OKYL049 in carbon limitation (C/N ratio 3). The V-ATPase PMA1 (YALI1_B28659g) that is mainly responsible for the regulation of pH homeostasis was not DE in any strain under any condition. None of the four genes involved in the glutamine pathway (YALI1_F00821g, YALI1_B26112g, YALI1_F23664g, and YALI1_E11943g) showed a significantly different expression.
These results indicate that none of the urea amidolyase, the urea or ammonium transporters, or genes of the glutamine pathway are induced by the presence of urea or ammonium on a transcriptome level in Y. lipolytica.

DISCUSSION
Most Y. lipolytica metabolic engineering studies have relied on the use of ammonium sulfate as a nitrogen source. While it can be argued that urea is an economically and environmentally attractive alternative nitrogen source, 12 it would be detrimental if the cellular response to the use of urea would undermine the  iScience Article significant process that has been made to develop Y. lipolytica as a promising microbial cell factory. In that light, it has been reassuring to observe that a thorough analysis with diverse lipid phenotypes and cultivation conditions revealed no significant differences in the cell physiology or transcriptome upon the use of urea instead of ammonium sulfate. Previous studies have found significant differences in gene expression when using a simple nitrogen source (ammonium sulfate) and a complex nitrogen source (peptone plus yeast extract). 29 We did not find major expression changes in our comparison of ammonium sulfate and urea, suggesting that the organic nitrogen source urea is very similar to the simple inorganic nitrogen source ammonium sulfate, and the difference in assimilation energetics does not majorly impact cell growth. Other studies that used urea as a nitrogen source in Y. lipolytica before did not report any growth differences 25 or an improvement of growth in urea compared to other nitrogen sources. 11 In addition, first experiments using synthetic and real human urine as nitrogen source for the cultivation of Y. lipolytica showed promising results with similar growth and biomass formations compared to ammonium sulfate, highlighting the future potential of urea as a nitrogen source. 11 The minor changes observed in the fatty acid profile under carbon limitation (C/N ratio 3) most likely derive from changes in membrane fatty acids composition, changes that become visible when the contribution of the storage lipids (triacylglycerols or sterol esters) to the lipid content is very low. These changes could be a stress response aimed at adjusting membrane fluidity 30 and connected to the acidification process of ammonium assimilation. However, we did not observe this change in the obese strain OKYL049. In addition, we did not find overall changes in the saturation or chain length of the fatty acids in any of the strains in carbon limitation.
A previous study reported a change in the fatty acid profile when OKYL029 was cultivated with urea in nitrogen-limiting conditions compared to ammonium sulfate. 25 Under this condition, we only observed a change in the fatty acid profile for the obese strain OKYL049. However, the observed trends in saturation and fatty acid chain length were inverted compared to the previous study. The previously reported change in fatty acid profile was possibly due to a pH change throughout the shake flask cultivation, 31 instead of the assimilation of urea per se. Because the ammonium metabolism is acidifying the media, the pH at the end of the cultivation might have been different in cultures with ammonium sulfate compared to urea. Our chemostat cultures were pH-controlled, preventing any influence of different pH. In additon, it is  Table S1. Genes marked with x have been removed during the filtering of the gene counts; genes marked with/did not show any significant (adjusted p value <0.05) change between the nitrogen sources. iScience Article noteworthy that the changes in the fatty acid profile due to the nitrogen source are minor compared to the changes introduced by genetic engineering to increase lipid production (OKYL029 vs OKYL049). Furthermore, these changes (overexpression of DGA1 and Dare1) did not aim to modify the fatty acid profile, while multiple studies have shown that targeted engineering e.g. of the desaturases can massively alter the fatty acid profile. 26,32,33 Our results show that urea can be used as a drop-in nitrogen source instead of ammonium sulfate without massively altering the storage lipid profile.
Urea is converted in a one-enzyme two-step reaction to ammonium and further to glutamate and glutamine. Yeast cells have pathways to regulate nitrogen metabolism, which are often sensed by measuring the cellular glutamine level. 34 Because we did not observe any significant changes in gene regulation downstream of glutamate, the glutamine level is likely unaffected by ammonium sulfate or urea as nitrogen sources.
Therefore, the minimal overall transcriptomic changes were anticipated and confirmed by only small overlaps of DE genes of different strains and conditions. The number of uncharacterized proteins shows the limitation of omics approaches for unconventional organisms. However, with increasing numbers of studies conducted in Y. lipolytica and advancements in protein function prediction, we expect this limitation to reduce gradually.
The near absence of transcriptional induction of any genes directly related to ammonium and urea transport and urea conversion was not anticipated. In Candida albicans, the urea transporter DUR3 is highly induced by the presence of urea on a transcriptional level. 35 In S. cerevisiae DUR3 and the urea amidolyase DUR1,2 are induced by the intermediate of the multifunctional enzyme Dur1,2p, allophanate. 36 We only observed a significant upregulation of one of the four genes homolog to DUR3. This indicates that YALI1_B05609g is the functional homolog of DUR3 with a similar regulation as in related yeast species. However, this upregulation was only observed in nitrogen limitation and not in carbon limitation conditions, in which the initial urea concentration was higher (2.4 g/L versus 0.21 g/L urea in nitrogen limitation). This indicates that the upregulation of DUR3 might be intertwined with the transcriptional response to nitrogen limitation.
Taken together, our physiological data and transcriptomics analysis show that the cell behavior in urea is very comparable to that in ammonium sulfate. Urea is an economically and environmentally attractive nitrogen source that does not acidify the media and requires low base addition during cultivation, further reducing the costs of the process. These findings can open the way for future studies and industrial applications using urea as a sustainable alternative nitrogen source in Y. lipolytica.

Limitation of the study
A limitation of transcriptomic studies in unconventional yeasts is the high amount of uncharacterized genes and proteins. Most of the DE genes identified in this study are uncharacterized. We anticipate that the knowledge gap between conventional and unconventional yeasts will be reduced over the next years, allowing for further insights that are currently out of reach.
In addition, this study is based on transcriptomics analysis and does not take posttranslational modification and regulation into account. Furthermore, we only conducted the study in chemostats with a dilution rate of 0.1. It is possible that the cell behavior could change at different dilution rates.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following: