Overexpression of Key Sterol Pathway Enzymes in Two Model Marine Diatoms Alters Sterol Profiles in Phaeodactylum tricornutum

Sterols are a class of triterpenoid molecules with diverse functional roles in eukaryotic cells, including intracellular signaling and regulation of cell membrane fluidity. Diatoms are a dominant eukaryotic phytoplankton group that produce a wide diversity of sterol compounds. The enzymes 3-hydroxy-3-methyl glutaryl CoA reductase (HMGR) and squalene epoxidase (SQE) have been reported to be rate-limiting steps in sterol biosynthesis in other model eukaryotes; however, the extent to which these enzymes regulate triterpenoid production in diatoms is not known. To probe the role of these two metabolic nodes in the regulation of sterol metabolic flux in diatoms, we independently over-expressed two versions of the native HMGR and a conventional, heterologous SQE gene in the diatoms Thalassiosira pseudonana and Phaeodactylum tricornutum. Overexpression of these key enzymes resulted in significant differential accumulation of downstream sterol pathway intermediates in P. tricornutum. HMGR-mVenus overexpression resulted in the accumulation of squalene, cycloartenol, and obtusifoliol, while cycloartenol and obtusifoliol accumulated in response to heterologous NoSQE-mVenus overexpression. In addition, accumulation of the end-point sterol 24-methylenecholesta-5,24(24’)-dien-3β-ol was observed in all P. tricornutum overexpression lines, and campesterol increased three-fold in P. tricornutum lines expressing NoSQE-mVenus. Minor differences in end-point sterol composition were also found in T. pseudonana, but no accumulation of sterol pathway intermediates was observed. Despite the successful manipulation of pathway intermediates and individual sterols in P. tricornutum, total sterol levels did not change significantly in transformed lines, suggesting the existence of tight pathway regulation to maintain total sterol content.


Introduction
Sterols are essential triterpenoids that function as regulators of cell membrane dynamics in all eukaryotic organisms [1]. In animals and higher plants, sterols participate in the synthesis of secondary metabolites involved in defense mechanisms, and steroid hormones that regulate growth and development [2]. Due to their presence in ancient sediments, sterol compounds are used as durable biomarkers to track important evolutionary events [3]. Sterols of plant origin, known as phytosterols, are used as nutraceuticals for their cholesterol-lowering effects [4]. Other therapeutic applications such as anti-inflammatory [5] and anti-diabetic activities [6] are currently under research. In order to meet MVA products IPP and DMAPP are subsequently used for the synthesis of squalene, the first committed intermediate in the formation of sterols [33] (Figure 1). In plants, fungi, and animals, squalene is converted into 2,3 epoxysqualene. This reaction is conventionally catalyzed by the enzyme squalene epoxidase (SQE, E.C. 1.14.14.17) [33]. Several studies indicate that SQE is a control point in cholesterol synthesis modulated by sterol levels and post-translationally regulated by cholesterol-dependent proteasomal degradation [33,34]. However, diatoms do not possess a conventional SQE, and instead this step is catalyzed by a recently characterized alternative squalene epoxidase (AltSQE) [16]. The diatom AltSQE belongs to the fatty acid hydroxylase superfamily and differs from the conventional flavoprotein SQE. Whether AltSQE has a similar role to SQE in sterol regulation is not known.
While selective inhibitors for AltSQE are not known, the treatment of diatoms with statins, known to inhibit HMGR enzymes, has resulted in perturbation of isoprenoids metabolism that included an overall decrease of total sterols content [24,35], suggesting that a conventional HMGR enzyme might be involved in the pathway. In this work, we genetically targeted the HMG CoA MVA products IPP and DMAPP are subsequently used for the synthesis of squalene, the first committed intermediate in the formation of sterols [33] (Figure 1). In plants, fungi, and animals, squalene is converted into 2,3 epoxysqualene. This reaction is conventionally catalyzed by the enzyme squalene epoxidase (SQE, E.C. 1.14.14.17) [33]. Several studies indicate that SQE is a control point in cholesterol synthesis modulated by sterol levels and post-translationally regulated by cholesterol-dependent proteasomal degradation [33,34]. However, diatoms do not possess a conventional SQE, and instead this step is catalyzed by a recently characterized alternative squalene epoxidase (AltSQE) [16]. The diatom AltSQE belongs to the fatty acid hydroxylase superfamily and differs from the conventional flavoprotein SQE. Whether AltSQE has a similar role to SQE in sterol regulation is not known.
While selective inhibitors for AltSQE are not known, the treatment of diatoms with statins, known to inhibit HMGR enzymes, has resulted in perturbation of isoprenoids metabolism that included an overall decrease of total sterols content [24,35], suggesting that a conventional HMGR enzyme Pharmaceuticals 2020, 13, 481 4 of 19 might be involved in the pathway. In this work, we genetically targeted the HMG CoA reduction and squalene epoxidation steps in P. tricornutum to provide further insights into the nature of the diatom sterol biosynthesis pathway and its regulatory constraints. Considering the demonstrated challenges in genetically down-regulating the essential genes involved in P. tricornutum sterol metabolism [14,16], we chose to investigate these pathway nodes by gene overexpression and subcellular localization.
We generated independent diatom exconjugant lines constitutively expressing (i) either the full-length HMGR, (ii) an N-terminal truncated version of HMGR (tHMGR), or (iii) a heterologous SQE from Nannochloropsis oceanica (NoSQE) over the background of the endogenous AltSQE. By phenotyping these transgenic diatom cell lines, we describe specific changes in several nodes of the sterol biosynthesis pathway and provide evidence for regulatory mechanisms unique to diatom sterol metabolism.

Identification of Putative HMGR from T. pseudonana and P. tricornutum
In contrast to the recently discovered AltSQE, little is known about the diatom HMGR enzyme despite being a major regulatory step in sterol biosynthesis. Previous studies demonstrated that specific HMGR inhibitors alter isoprenoids metabolism in the diatoms P. tricornutum, Haslea ostrearia, and Rhizosolenia setigera [24,35]. Given the presence of many unusual features in diatoms metabolism [13,15,16], we analyzed the conservation of the HMGR sequence among all the diatom species with genomic or transcriptomics sequences available.
The amino acid sequence of Arabidopsis thaliana HMGR enzyme (AT1G76490) was used as query to search against the genome sequence of the diatoms T. pseudonana and P. tricornutum to identify the genes putatively encoding HMGR. Through this analysis, we identified a single copy of a putative HMGR gene located in chromosome 29 in P. tricornutum (Gene ID Phatr3_J16870) [14] and chromosome 4 in T. pseudonana (Gene ID Thaps_33680). In model eukaryotic organisms, HMGR is characterized by the presence of an N-terminal membrane domain and a C-terminal catalytic region. Sequence alignment analysis revealed differences in membrane domain location among model organisms, while the catalytic region is conserved (Figure 2 and Figure S3). The C-terminal catalytic domain of HMGR was highly conserved across all the organisms analyzed ( Figure 2). The catalytic residues Glu559, Asp767, and His866 [27,36], were also found to be present and conserved in T. pseudonana and P. tricornutum ( Figure S3). HMGRs from Saccharomyces cerevisiae and Homo sapiens possess a sterol sensing domain (SSD) in the transmembrane N-terminal region, which is involved in sterol homeostasis [28], therefore, we analyzed HMGR transmembrane region in diatoms to identify similarities with other model organisms. Most of the analyzed HMGR sequences from diatoms possess three trans-membrane helices in the N-terminal domain, except a few with two or no domains predicted (Table S2). In comparison, plants usually have two domains [36]. We found seven transmembrane domains in S. cerevisiae and five in Homo sapiens (Table S2). We did not detect similarities with known yeast and mammals SSDs in the N-terminal region in any of the diatoms analyzed ( Figure 2).

Phylogenetic Analysis of HMGR and Conserved Protein Domains
Based on the alignments of full-length HMGR protein sequences of twenty-eight diatom species retrieved from whole genome and transcriptome assemblies, a maximum likelihood phylogenetic tree was constructed to study evolutionary relationship of HMGR protein sequence among diatoms ( Figure 2). We designated HMGR from yeast, mammals, and plants as outgroups. Pennate and centric diatoms were divided into two different clades ( Figure 2). As expected, HMGR from diatoms of the same genus tended to cluster together. Species from the order Thalassiosirales which includes Thalassiosira and Skeletonema genus were grouped together ( Figure 2). Similarly, HMGR from the diatoms P. tricornutum and Fistulifera solaris that belong to the Naviculales order appear to be closely related ( Figure 2). Interestingly, we did not find a match for HMGR in the transcriptomic sequences of the diatoms: Chaetoceros muelleri, as previously reported [15], Chaetoceros brevis, Chaetoceros debilis, and Chaetoceros curvisetus (Table S2).

Phylogenetic Analysis of HMGR and Conserved Protein Domains
Based on the alignments of full-length HMGR protein sequences of twenty-eight diatom species retrieved from whole genome and transcriptome assemblies, a maximum likelihood phylogenetic tree was constructed to study evolutionary relationship of HMGR protein sequence among diatoms ( Figure 2). We designated HMGR from yeast, mammals, and plants as outgroups. Pennate and centric diatoms were divided into two different clades ( Figure 2). As expected, HMGR from diatoms of the same genus tended to cluster together. Species from the order Thalassiosirales which includes Thalassiosira and Skeletonema genus were grouped together ( Figure 2). Similarly, HMGR from the diatoms P. tricornutum and Fistulifera solaris that belong to the Naviculales order appear to be closely related ( Figure 2). Interestingly, we did not find a match for HMGR in the transcriptomic sequences of the diatoms: Chaetoceros muelleri, as previously reported [15], Chaetoceros brevis, Chaetoceros debilis, and Chaetoceros curvisetus (Table S2).

Expression and Subcellular Localization of Putative HMGR and tHMGR
While the core reactions in sterol synthesis being conserved in T. pseudonana and P. tricornutum [15], both diatoms produce a distinctive profile of sterol compounds which variates differently upon changing environmental conditions and chemical inhibitors treatment [15,20]. Additionally, the sterol metabolism of the centric diatom T. pseudonana has not been explored to the same depth as the model pennate P. tricornutum. To investigate the subcellular localization and evaluate the effect of overexpression of the rate-limiting enzyme HMGR on sterol accumulation, T. pseudonana and P. tricornutum were transformed with episomes containing their respective putative HMGR copy fused to mVenus, driven by a constitutive promoter (Figures 1, S1 and S2). Episomes are maintained extrachromosomally and therefore enable more consistent expression required for metabolic studies [37]. The trans-membrane domains of HMGR enzymes in mammals and yeast have been reported to contain a sterol sensing domain (SSD) that regulates expression and degradation of the enzyme [38].

Expression and Subcellular Localization of Putative HMGR and tHMGR
While the core reactions in sterol synthesis being conserved in T. pseudonana and P. tricornutum [15], both diatoms produce a distinctive profile of sterol compounds which variates differently upon changing environmental conditions and chemical inhibitors treatment [15,20]. Additionally, the sterol metabolism of the centric diatom T. pseudonana has not been explored to the same depth as the model pennate P. tricornutum. To investigate the subcellular localization and evaluate the effect of overexpression of the rate-limiting enzyme HMGR on sterol accumulation, T. pseudonana and P. tricornutum were transformed with episomes containing their respective putative HMGR copy fused to mVenus, driven by a constitutive promoter (Figure 1, Figures S1 and S2). Episomes are maintained extra-chromosomally and therefore enable more consistent expression required for metabolic studies [37]. The trans-membrane domains of HMGR enzymes in mammals and yeast have been reported to contain a sterol sensing domain (SSD) that regulates expression and degradation of the enzyme [38]. Although our results suggest that diatom HMGRs lacks an SSD (Figure 2), we designed an N-terminal truncated version, tHMGR, to evaluate whether an unknown regulatory sequence is present in N-terminal region affecting activity of HMGR in diatoms. This tHMGR sequence encoded solely the C-terminal catalytically active region of the enzyme.
The mVenus fluorescence was measured by flow cytometry and used as an indirect proxy of enzyme expression, since each expression system was C-terminal fused with mVenus protein.
Three clones per expression system with the highest median mVenus signal were chosen for full scale experiments. In P. tricornutum, time course median mVenus fluorescence in mVenus control clones was 10-fold compared to WT, which confirms the effectiveness of the chosen promoter ( Figure S4). Conversely, median mVenus fluorescence clones expressing HMGR-mVenus and tHMGR-mVenus was 1.3-fold compared to WT, indicating an apparent regulation process occurring over the fused proteins ( Figure S4). In T. pseudonana, median mVenus fluorescence in HMGR-mVenus, tHMGR-mVenus, and mVenus control clones appear similar to WT signal, suggesting that low expression was achieved using the EF2 promoter ( Figure S5). However, confocal microscopy images confirmed expression of mVenus in both diatom species (Figure 3). Different cellular localizations were observed for each genetic construct. Images of control cell lines showed mVenus expression localized in the cytoplasm (Figure 3), while no mVenus fluorescence was detected in WT diatoms. mVenus fluorescence in exconjugants overexpressing HMGR-mVenus was detected around the chloroplast, suggesting that putative HMGR is localized in the endoplasmic reticulum, ER, which tightly surrounds the chloroplasts in diatoms [39,40]. Conversely, tHMGR-mVenus localizes in the cytoplasm, consistent with truncation of the N-terminal membrane domain ( Figure 3).
terminal catalytically active region of the enzyme.
The mVenus fluorescence was measured by flow cytometry and used as an indirect proxy of enzyme expression, since each expression system was C-terminal fused with mVenus protein. Three clones per expression system with the highest median mVenus signal were chosen for full scale experiments. In P. tricornutum, time course median mVenus fluorescence in mVenus control clones was 10-fold compared to WT, which confirms the effectiveness of the chosen promoter ( Figure S4). Conversely, median mVenus fluorescence clones expressing HMGR-mVenus and tHMGR-mVenus was 1.3-fold compared to WT, indicating an apparent regulation process occurring over the fused proteins ( Figure S4). In T. pseudonana, median mVenus fluorescence in HMGR-mVenus, tHMGR-mVenus, and mVenus control clones appear similar to WT signal, suggesting that low expression was achieved using the EF2 promoter ( Figure S5). However, confocal microscopy images confirmed expression of mVenus in both diatom species (Figure 3). Different cellular localizations were observed for each genetic construct. Images of control cell lines showed mVenus expression localized in the cytoplasm (Figure 3), while no mVenus fluorescence was detected in WT diatoms. mVenus fluorescence in exconjugants overexpressing HMGR-mVenus was detected around the chloroplast, suggesting that putative HMGR is localized in the endoplasmic reticulum,ER, which tightly surrounds the chloroplasts in diatoms [39,40]. Conversely, tHMGR-mVenus localizes in the cytoplasm, consistent with truncation of the N-terminal membrane domain ( Figure 3).  T. pseudonana cells, scale bars correspond to 10 µm. Wild type (WT) as negative control and control cell lines that only expressed mVenus. 3-hydroxy-3-methylglutaryl-coenzyme A reductase, HMGR-mVenus; truncated HMGR, tHMGR-mVenus; squalene epoxidase from N. oceanica, NoSQE-mVenus. Scale bars correspond to 10 µm.

Influence of HMGR and tHMGR Expression on Sterol Levels in T. pseudonana and P. tricornutum
Although expression of transgenes appeared to be low according to mVenus fluorescence levels, we proceeded to identify changes in sterol profiles in the transgenic lines. Sterols were extracted from exconjugants in the mid-exponential phase, which was the time period with the maximum observed mVenus fluorescence ( Figures S4 and S5) with enough biomass to sample for sterol extraction (determined to be 48 h growth for T. pseudonana and 72 h for P. tricornutum). After 75 h, cell density of P. tricornutum HMGR-mVenus and tHMGR-mVenus was 1.4 times lower than WT, while no growth impairment was observed for mVenus exconjugants ( Figure S6). No differences in chlorophyll levels and in effective quantum yield of PSII were observed in P. tricornutum exconjugants ( Figures S7 and S8). Similarly, no growth impairment, chlorophyll levels, or differences in effective quantum yield of PSII compared to WT were observed for T. pseudonana exconjugants (Figures S9-S11).
In P. tricornutum overexpressing HMGR-mVenus, squalene levels were ten times higher in HMGR-mVenus exconjugants than in WT and mVenus controls. Moreover, a 3-fold increase in cycloartenol and a 2.5-fold obtusifoliol accumulation was detected compared to WT ( Figure 4). However, we did not detect the intermediate 2,3 epoxysqualene. Levels of end-point sterol campesterol decreased 2-fold, whereas no significant differences were observed in brassicasterol, the most abundant sterol in P. tricornutum. We detected traces levels of end point sterol 24-methylcholesta-5,24(24 )-dien-3β-ol in the WT and mVenus controls, which is typically found in centric diatoms and has not been reported in P. tricornutum [18]. Levels of this end-point sterol were 17 times higher in two independent exconjugant lines expressing HMGR-mVenus. Total sterol levels were not affected despite significant changes in individual sterols (Figure 4).
of P. tricornutum HMGR-mVenus and tHMGR-mVenus was 1.4 times lower than WT, while no growth impairment was observed for mVenus exconjugants ( Figure S6). No differences in chlorophyll levels and in effective quantum yield of PSII were observed in P. tricornutum exconjugants ( Figures S7 and S8). Similarly, no growth impairment, chlorophyll levels, or differences in effective quantum yield of PSII compared to WT were observed for T. pseudonana exconjugants ( Figures S9-S11).
In P. tricornutum overexpressing HMGR-mVenus, squalene levels were ten times higher in HMGR-mVenus exconjugants than in WT and mVenus controls. Moreover, a 3-fold increase in cycloartenol and a 2.5-fold obtusifoliol accumulation was detected compared to WT ( Figure 4). However, we did not detect the intermediate 2,3 epoxysqualene. Levels of end-point sterol campesterol decreased 2-fold, whereas no significant differences were observed in brassicasterol, the most abundant sterol in P. tricornutum. We detected traces levels of end point sterol 24methylcholesta-5,24(24′)-dien-3β-ol in the WT and mVenus controls, which is typically found in centric diatoms and has not been reported in P. tricornutum [18]. Levels of this end-point sterol were 17 times higher in two independent exconjugant lines expressing HMGR-mVenus. Total sterol levels were not affected despite significant changes in individual sterols (Figure 4). Since expression of mVenus, HMGR-mVenus, and tHMGR-mVenus in T. pseudonana did not appear to be effective based on flow cytometry data ( Figure S5), observed changes on sterol profiles may not be directly related to the overexpression of the targeted enzymes. T. pseudonana cell lines transformed with HMGR-mVenus construct exhibited a decrease in the minor sterols fucosterol and isofucosterol relative to WT control ( Figure S12). However, isofucosterol reduction was also detected in the control expressing only mVenus ( Figure S12). No intermediates were detected. Total sterol levels remained similar in the three independent cell lines studied; no significant differences were observed in comparison to the WT control ( Figure S12).
No changes in total sterol content were observed in T. pseudonana cell lines transformed with tHMGR-mVenus ( Figure S12). No intermediates were detected. Significant changes in less abundant sterol compounds occurred in transformants including controls expressing only mVenus.

Heterologous Expression of a Stramenopile Putative SQE
Diatoms have been reported to employ an alternative squalene epoxidase (AltSQE) that is different from the conventional SQE found in other eukaryotes [16]. It has been observed that artificially altering the expression of this enzyme in P. tricornutum is particularly challenging [16], suggesting that a strict regulation of endogenous AltSQE may be occurring as is the case for SQE in [33,34]. Therefore, we hypothesized that the expression of a heterologous, conventional SQE that could override endogenous regulation, would influence final sterol levels. Consequently, a heterologous putative SQE from the Stramenopile N. oceanica (NoSQE, Nanoce ID 521007) was expressed in the diatoms T. pseudonana and P. tricornutum.
Total sterol content of T. pseudonana and P. tricornutum exconjugants transformed with NoSQE-mVenus constructs remained unchanged compared to WT and mVenus controls (Figure 4 and Figure S12). However, in P. tricornutum overexpressing NoSQE-mVenus, we observed significant differences in both end-point sterols and sterol intermediates. Downstream intermediates cycloartenol and obtusifoliol exhibited a 1.8-fold increase compared to WT control (Figure 4), while no differences in squalene were observed. The intermediate 2,3 epoxysqualene was not detected in either diatom species. In contrast to HMGR-mVenus overexpression, campesterol increased by 3-fold ( Figure 4). However, the major end-point sterol brassicasterol remained unchanged.

HMGR is Largely Conserved among Diatoms and Lacks a Conventional Sterol Sensing Domain
To investigate sequence characteristics of the rate-limiting HMGR enzyme in diatoms, we identified the genes putatively encoding the enzyme HMGR from 28 different diatom species. While a putative HMGR homologue was detected in 28 of the diatom species, the failure to detect HMGR transcripts in transcriptomic sequences of some diatoms belonging to Chaetoceros genus (Table S2) may support the hypothesis that that those diatoms may solely rely on the MEP pathway to produce isoprenoids [15]. The lack of obvious HMGR transcripts may also have occurred due to low expression or down-regulation of these and other genes related to the MVA pathway under the conditions in which RNA sequencing was performed. Nevertheless, conserved HMGR genes were detected in many diatoms for which genomic or transcriptomic data are available. These HMGR genes diverge between pennate and centric diatoms (Figure 2), which are separated by 90 million years of divergent evolution [41]. Presence of putative HMGR in most of the diatom species analyzed is an indicator of presence of MVA pathway operating in diatoms, as it was previously reported by transcriptomics analysis in the diatoms P. tricornutum and T. pseudonana [15]. The MEP pathway appears functional in the two model diatoms, which indicates that both cytosolic MVA and plastidial MEP are simultaneously operating in P. tricornutum and T. pseudonana, as is the case in plants [15,22]. Presence of both pathways could represent an advantage for terpenoids production, due to a potentially higher metabolic flux and greater pool of available precursors [17].
The organization of the N-terminal transmembrane domain of diatom HMGR differs significantly to their mammals and yeast counterparts. While most of the diatoms analyzed in this study presented three transmembrane domains (Figure 2, Table S2), mammals possess five and yeast, seven. The presence of transmembrane domains is likely related to anchoring the protein within the ER membrane, but the consequences of this structural difference for diatom HMGR in terms of regulation of enzyme expression and activity are unknown. In mammals and yeast, HMGR possess a SSD involved in sensing oxysterol molecules that activate feedback regulation leading to degradation of the protein [28,42]. Despite the lack of a conventional sterol-sensing domain in the HMGR enzymes of diatoms (Figure 2), several studies have shown a transcriptional response of MVA enzymes to perturbations in sterol metabolism [11,15]. The fact that diatom sequences do not possess a canonical SSD opens several possibilities, including that HMGR may not play the same regulatory role in diatoms as it does in other organisms; i.e., the MVA pathway may be regulated through a different mechanism that does not involve HMGR feedback regulation. Another possible explanation is that HMGRs of diatoms and plants possess a non-conventional SSD sequence, a motif with different characteristics than the already described in mammals and yeast.

HMGR Overexpression Lead to Accumulation of Sterol Pathway Intermediates in P. tricornutum
In this study, we investigated the response of diatoms to genetic targeting of sterol biosynthesis, through manipulation of the MVA rate-limiting enzyme HMGR. The fluorescent localization of extra-chromosomally expressed HMGR-mVenus to the membrane surrounding the plastid is consistent with ER localization of proteins, including AltSQE, from previous studies in P. tricornutum [16,39] and T. pseudonana [43]. Therefore, our study provides evidence that diatom HMGR is localized to the ER, just as in mammals, yeast, and higher plants [28,44]. The truncation of the native HMGR sequence resulted in cytoplasmic localization, demonstrating that the signals for protein targeting are in the N-terminal portion of the protein sequence.
In some cases, overexpression of HMGR has been showed to slightly increase sterol and intermediates production in other organisms. The overexpression of HMGR from the plant Panax ginseng in Arabidopsis thaliana resulted in a nearly two-fold increase of sitosterol, campesterol, and cycloartenol, while levels of squalene and stigmasterol did not significantly change [45]. In a more recent study, Lange et al. [46] independently over-expressed all genes participating in the MVA pathway, obtaining a significant increase in total sterols when expressing HMGR (3.4-fold) and 3-hydroxy-3-methylglutaryl-co-enzyme-A synthase, HMGS (2-fold). The overexpression of native HMGR in Arabidopsis (HMG1) led to high levels of HMGR mRNA, but only a slight increase in HMGR activity, and no changes in leaf sterol levels [47]. Heterologous expression of HMGR from Hevea brasiliensis in tobacco, however, showed an increase in HMGR transcript and total sterol from leaves [48].
In this study, overexpression of endogenous putative HMGR-mVenus in P. tricornutum resulted in an increased accumulation of the intermediates squalene, cycloartenol, and obtusifoliol and decrease in the end-point sterol campesterol. However, we did not detect 2,3 epoxysqualene, indicating possible differences on the catalytic rates of the enzymes squalene epoxidase and cycloartenol cyclase (Figure 1). While accumulation of squalene is detectable, 2,3 epoxysqualene seems to be rapidly converted into cycloartenol. Accumulation of 2,3 epoxysqualene has been reported before by chemically inhibiting cycloartenol cyclase with Ro 48-8071 [14,15].
Accumulation of intermediates suggests that in P. tricornutum, overexpression of HMGR-mVenus boosted production of presumably MVA-derived intermediates, IPP, and DMAPP that are subsequently converted into squalene (Figure 1). Even though MVA products also serve as building blocks of other isoprenoids [49], perturbation of the rate-limiting step catalyzed by HMGR was sufficient to cause accumulation of downstream intermediates committed to sterol biosynthesis, such squalene, cycloartenol, and obtusifoliol. However, this metabolic bottleneck did not translate into overall increase of sterol compounds; on the contrary, levels of the end-point sterol campesterol were reduced and brassicasterol levels remained unchanged.
Although fluorescence levels in P. tricornutum exconjugants expressing the target enzymes were considerably lower compared to the expression of mVenus alone ( Figure S5), phenotypic changes in terms of sterol profiles indicate that the level of expression was sufficient to cause perturbations in the metabolic pathway ( Figure 4). Moreover, since all the enzymes expressed are membrane proteins, as confirmed by confocal microscopy (Figure 3 and Figure S3), fluorescent signal could have been hindered and no direct correlation with expression could be assumed. The absence of detectable sterol pathway intermediates in T. pseudonana transformants ( Figure S12) may be related with the promoter chosen for overexpression of the target enzymes, as higher expression may be necessary to trigger the accumulation of intermediates observed in P. tricornutum. The fluorescence signal for mVenus control in P. tricornutum (Phatr3_J49202 promoter) was around ten times higher than in T. pseudonana mVenus control (EF2 promoter) throughout the full-scale experiment ( Figures S4 and S5), suggesting that use of stronger promoters for metabolic engineering of T. pseudonana should be considered. Levels of fucosterol and isofucosterol decreased in T. pseudonana lines transformed with a putative endogenous HMGR-mVenus ( Figure S12). However, since similar results were observed for isofucosterol in the control expressing only mVenus, it is uncertain if the observed reduction was a direct consequence of putative HMGR-mVenus overexpression alone. No intermediates were detected in T. pseudonana overexpressing putative HMGR-mVenus. End-point sterols 24-methylenecholesta-5,24(24')-dien-3β-ol, cholesterol, campesterol, and total sterol levels were statistically indistinguishable from those obtained with untransformed WT and mVenus control transformants ( Figure S12).
Results suggest that there are several regulation points in sterol biosynthesis in diatoms, including the MVA pathway, conserved core, and specialized downstream reactions. It is also possible that MEP responds to an alteration on the MVA pathway, rebalancing IPP and DMAPP pools, and metabolic cross-talk between this two pathways could potentially occur in diatoms. Despite the core reactions in sterol synthesis being conserved in T. pseudonana and P. tricornutum [15], we observed different responses in lines overexpressing putative HMGR-mVenus of these two model diatoms (Figure 4 and Figure S12). As previously mentioned, these results might be related with the promoters chosen for each species, or could indicate differences in sterol regulation between centric and pennate diatoms that correlates with divergence between putative HMGR from both diatom groups ( Figure 2). Differences in sterol profiles from T. pseudonana and P. tricornutum have been suggested to occur in downstream reactions of sterol synthesis [15], and responses to alteration on a key point of MVA pathway suggest possible divergences in regulation mechanisms.

Overexpression of tHMGR does not Circumvent Native Regulatory Mechanisms
Additional strategies have been developed to overcome regulation by HMGR and increase MVA carbon flux in other eukaryotes. Truncation of HMGR to remove N-terminal membrane and SSD domain was first reported in plants and yeast with the aim to express only the catalytic domain of HMGR and avoid regulatory effects [50,51]. Although HMGRs of plants do not contain an SSD sequence (Figure 2), expression of an N-terminal truncated HMGR has been reported to increase sterol levels. Expression of tHMGR from hamster in tobacco resulted in augmented sterol content in leaf tissue [52]. Constitutive expression of tHMGR from Hevea brasiliensis in tobacco resulted in an 11-fold increase of seed HMGR activities and 2.4-fold increase in total seed sterol content [53]. However, overexpression of a tHMGR has not always been effective at altering sterol content; the overexpression of tHMGR in yeast resulted in accumulation of squalene and no changes in ergosterol, the final sterol in yeast [50,51]. Likewise, in this study we did not observe a statistically significant alteration in total sterols of T. pseudonana and P. tricornutum after constitutive expression of a tHMGR-mVenus ( Figure 4 and Figure S12). Yet, accumulation of the intermediates squalene, cycloartenol, and obtusifoliol was observed in P. tricornutum. Interestingly, levels of those intermediates were higher when expressing HMGR-mVenus, suggesting that truncation may have affected enzyme activity, performance, or access to substrates. Although we observed changes in intermediates and minor sterol levels in P. tricornutum expressing HMGR-mVenus and tHMGR-mVenus, total sterol levels remained unchanged. These results suggest that diatoms have a tight sterol regulation system that may not be related to the conventional regulation model that involves SSD but rather a complex system with several regulation points not only in the MVA pathway but further down the sterol metabolic pathway.

Levels of End-Point Campesterol Increased after Heterologous Expression of NoSQE in P. tricornutum
Diatoms possess a distinct AltSQE [16] catalyzing the conversion of squalene into 2,3 epoxysqualene which is then transformed into cycloartenol, the first committed step towards the production of steroids (Figure 1). Whether the presence of an AltSQE confers diatoms and other microeukaryotes with specific biological advantages is not yet known. Similarly, AltSQE and SQE are mutually exclusive and, to date, no organisms have been found to naturally harbor both [16].
The genetic manipulation of SQE and squalene synthase, SQS, has been extensively used for enhanced production of squalene and triterpenoids [54][55][56]. Point mutations in the SQE gene (ERG1) in yeast resulted in accumulation of squalene [57]. Similarly, accumulation of squalene was observed in the green microalgae Chlamydomonas reinhardtii after knocking down the SQE gene, while co-transformation lines with SQE-overexpression and SQE-knockdown yielded similar amounts of squalene [58]. This is the first study to investigate the response of diatoms to the expression of a conventional SQE. We did not observe any significant changes in growth and photosynthetic phenotypes of T. pseudonana and P. tricornutum expressing heterologous NoSQE (Figures S6-S11). This suggests that in diatoms there is no apparent toxicity or physiological reason for the mutual exclusivity between AltSQE and conventional SQE. Confocal microscopy images of lines expressing NoSQE-mVenus fusion proteins, revealed that heterologous NoSQE was proximal to the chloroplasts, indicating that diatoms could recognize the ER signal peptide on the heterologous NoSQE, localizing it in the ER membrane ( Figure 3 and Figure S3), just as the endogenous AltSQE (Pollier et al., 2019) and native SQE enzymes are in other species [59,60].
Significant accumulation of cycloartenol (1.8-fold) and obtusifoliol (1.8-fold) intermediates, but not of 2,3 epoxysqualene, was obtained for P. tricornutum lines expressing NoSQE-mVenus ( Figure 4b). These intermediates occurred after the formation of 2,3 epoxysqualene, which is the product of the reaction catalyzed by SQE ( Figure 1). As expected, we did not observe increased accumulation of squalene, which is the substrate for SQE, contrary to accumulation obtained by expressing HMGR-mVenus which is upstream of squalene production ( Figure 1). Nevertheless, heterologous expression of NoSQE-mVenus resulted in a 2-fold increase of campesterol, an end-point sterol. Accumulation of intermediates was higher in P. tricornutum cell lines expressing HMGR compared to those expressing NoSQE-mVenus (Figure 4a). This indicates that intermediates accumulated by MVA pathway manipulation (i.e., HMGR) do not necessarily increase the flux to brassicasterol, suggesting that sterol regulation is occurring at the conserved core point and at other points further down the metabolic pathway. In particular, this might suggest that in P. tricornutum the epoxidation of squalene might be involved in pathway flux modulation, as observed in mammals [33,34], and that, to complete the scenario suggested by the results obtained by expressing NoSQE, it is plausible that an additional pathway checkpoint exists at the level of the C22-desaturation (E.C 1.14.19.41, Phatr3_J51757) [14]. When treated with fluconazole and fenpropimorph, inhibitors targeting upstream of campesterol, the transcription of Phatr3_J51757 significantly increases [15]. This further supports the hypothesis that the last reaction in sterol synthesis could be a highly regulated point to maintain stable sterol levels in the cell. These results suggest that to observe changes in end-point sterol compounds, increasing the precursors pool is not enough; genetic manipulation should target other points further down in the metabolic pathway, such as committed steps in sterol synthesis. A future co-expression approach to increase end-point sterol compounds in P. tricornutum could involve simultaneous expression of enzymes in the conserved core (i.e., SQE, AltSQE, cycloartenol synthase) and enzymes further down such as sterol C-22 desaturase. Similar co-expression approaches for manipulation of sterol levels in diatoms has not been reported but has been used to increase triterpenoid production in other organisms [30,46,55,61].

Episome Construction and Transformation
All episomes used in this study were assembled using the uLoop assembly method [63]. Individual components for episome assembly (L0 parts) were built and domesticated using uLoop assembly syntax. Assembly reactions were performed using the respective uLoop assembly backbones for each level as described by Pollak et al. [63]. After domestication, each L0 part was confirmed by Sanger sequencing. Correct episome assemblies were confirmed by colony PCR and diagnostic restriction digestion. The source of each DNA part and primers used for domestication are listed in Table S1. All L0 parts used to assemble the plasmids used in this work have been deposited in Addgene (Table S1). Plasmid maps (Figures S1 and S2) and complete plasmid sequence are provided in Supplemental Material.

Diatom Transformation and Screening
Diatoms were transformed by bacterial conjugation [67]. To increase transformation efficiency, the transformation protocol for T. pseudonana was modified by increasing the starting bacterial density (OD 600 to 0.5) and the final incubation and recovery period for transformed diatom culture to 24 h prior to selection on plates containing nourseothricin. T. pseudonana and P. tricornutum colonies resistant to nourseothricin (50 µg mL −1 ) or blasticidin (10 µg mL −1 ), respectively, were inoculated in 96-multiwell plates containing 200 µL of L1 medium with 100 µg mL −1 of nourseothricin or 10 µg mL −1 blasticidin, depending on the diatom species, and subcultured every 5 days. Clonal lines from 96-well plates were screened by detecting mVenus fluorescence using a CytoFLEX S (Beckman Coulter) flow cytometer operated in plate mode. A total of 48 clones of each expression system were screened for T. pseudonana and 12 for P. tricornutum. A 488 nm laser was used for fluorescence excitation; mVenus fluorescence was detected using a 525/40 nm filter and chlorophyll fluorescence was detected using 690/50 nm filter. A total of 10,000 events were analyzed per sample. Three independent cell lines per construct with the highest median mVenus fluorescence readings were selected for full-scale experiments, including wild type, WT, and empty vector controls.

Experiments with Transgenic Diatom Cultures
Three replicates of each selected clone were inoculated in 5 mL of L1 medium (100 µg mL −1 nourseothricin or 10 µg mL −1 blasticidin) and grown for 3 days. Subsequently, cultures were upscaled to 50 mL L1 supplemented with the respective antibiotic for 5 days and these were used to inoculate cultures in L1 medium for sterol analysis experiments. Full-scale experiments were carried out in 200 mL flasks containing 120 mL of L1 medium and antibiotic under continuous light (150 µmol photons m −2 s −1 ) and constant shaking (95 rpm). Cell density and mVenus fluorescence were monitored daily by sampling 200 µL from each culture and transferring it to a 96-well plate for high-throughput flow cytometry analysis. Pulse amplitude modulated (PAM) fluorometry was used to estimate photosynthetic activity by comparing fluorescence yield of photosystem II, PSII, under ambient irradiance (F) and after application of a saturating pulse (Fm) [68]. After 48 h of growth, biomass was harvested by centrifuging at 4000 g for 10 min. Diatom pellets were washed with Milli-Q water to eliminate excess salt, freeze-dried to determine dry weight, and kept at −20 • C until sterol extraction.

Extraction and Analysis of Sterols by GC-MS
For sterol extraction, dry cell matter was heated in 1 mL of 10% KOH ethanolic solution at 90 • C for one hour. Sterols were extracted from cooled material in three volumes of 400 µL of hexane. An internal standard, 5a-cholestane, was added to each sample. Hexane fractions were dried under a gentle N 2 stream, and derivatized with 50 µL of 99% BSTFA + 1% TMCS (N,O-Bis(trimethylsilyl)trifluoroacetamide, Trimethylsilyl chloride) at 70 • C for one hour. The resulting extractions were resuspended in 50 µL of fresh hexane prior to GC-MS injection.
Gas chromatography/mass spectrometry (GC-MS) analysis was performed using an Agilent 7890 instrument equipped with a HP-5 capillary column (30 m; 0.25 mm inner diameter, film thickness 0.25 µm) coupled to an Agilent quadrupole MS (5975 N) instrument. The following settings were used: oven temperature initially set to 50 • C, with a gradient from 50 • C to 250 • C (15.0 • C min −1 ), and then from 250 • C to 310 • C (8 • C min −1 , hold 10 min); injector temperature = 250 • C; carrier gas helium flow = 0.9 mL min −1 . A split-less mode of injection was used, with a purge time of 1 min and an injection volume of 5 µL. Mass spectrometer operating conditions were as follows: ion source temperature 230 • C; quadrupole temperature 150 • C; accelerating voltage 200 eV higher than the manual tune; and ionization voltage 70 eV. Full scanning mode with a range from 50 to 650 Dalton was used.

Fluorescence Imaging
Live diatom transformants expressing mVenus were imaged without fixative with a confocal laser scanning microscope (Nikon A1 Plus, Tokyo, Japan) and photomultiplier tube (PMT) detector. The 488-nm and 637-nm lasers were used for mVenus and chlorophyll autofluorescence, respectively. Gains on the detector were kept constant between samples and controls. Images were acquired with 60 ×/1.4 objective oil immersion objective and processed using imaging software NIS-Elements Viewer 4.0 (Nikon, Japan).

Multiple Sequence Alignment and Phylogenetic Reconstruction
Diatom homologue sequences were retrieved either from the Marine Microbial Eukaryote Transcriptome Sequencing Project (MMETSP) [69,70] database or from GenBank protein database (Table S2) using BLASTp search with T. pseudonana HMGR as the query sequence. The species names and corresponding MMETSP ID numbers are listed in Table S2. HMGR from yeast, mammals, and plants were used as outgroups. Sequences from outgroups (Table S2) were obtained from GenBank protein database (complete sequence in Supplementary Material). Multiple sequence alignments of the full-length protein sequences were performed by MAFFT version 7 program with default parameters and alignments were manually edited by exclusion of ambiguously aligned regions. The maximum likelihood phylogenetic tree was constructed using MEGA 6 with partial deletion option. The reliability of obtained phylogenetic tree was tested using bootstrapping with 1000 replicates. Prediction of transmembrane helices in HMGR from diatoms was carried out using the TMHMM Server v. 2.0 with default parameters [70]. Conserved motifs in the selected sequences were identified by an InterProScan [71] search against all available member databases, including Pfam (protein families) and SUPERFAMILY (structural domains).

Statistical Analysis
All plots were generated using R: a language and environment for statistical computing. All experiments were conducted in triplicate. The analyses performed were Shapiro-Wilk to test normality, and non-parametric Kruskal-Wallis test and Pairwise Wilcoxon Rank Sum Tests to calculate pairwise comparisons between group levels with corrections for multiple testing. Differences between groups were considered significant at p < 0.05.

Conclusions
The results obtained in this study demonstrate the effectiveness of extra-chromosomal expression of key enzymes involved in sterol synthesis to influence levels of specific sterol compounds. We confirmed reported advantages of the use of extra-chromosomal episomes transformed via conjugation such as expression consistency among clones (Figures S4-S11) and no random genome integration [37]. Additionally, we demonstrated the convenience of modular assembling systems such as uLoop [63] to build versatile genetic constructs for a functional genetics study with multiple species.
Furthermore, we applied reproducible genetic transformation methods for extra-chromosomal and heterologous expression to provide important insights into the metabolic bottleneck and pathway-level regulation of sterol synthesis in diatoms. We obtained accumulation of sterol pathway intermediates by overexpression of HMGR-mVenus, indicating possible metabolic bottleneck(s) downstream of the MVA pathway that may limit flux into end-point sterols. Accumulation of pathway intermediates is interesting from a biotechnological perspective, as an increased intermediate pool could be used by heterologous pathways plugged into endogenous (tri)terpenoid synthesis, allowing production of other high-value terpenoids as geraniol [12]. Whilst significant accumulation of intermediates participating in sterol synthesis was observed in P. tricornutum transformants, T. pseudonana and P. tricornutum transformants did not appear to produce different levels of total sterols. It is presumed that several levels of regulation could be affecting the expression, localization, lifetime, and activities of introduced genes. Further research into the regulatory responses of diatoms to heterologous overexpression may provide further insights into these processes and improve strategies for more informed metabolic engineering approaches.
Supplementary Materials: The following are available online at http://www.mdpi.com/1424-8247/13/12/481/s1, Figure S1. Maps of plasmids used for P. tricornutum transformation, Figure S2. Maps of plasmids used for T. pseudonana transformation, Figure S3. Sequence alignment of HMGR protein (catalytic domain) from model organisms, Figure S4. mVenus fluorescence during full scale experiment in P. tricornutum transformants, Figure S5. mVenus fluorescence during full scale experiment in T. pseudonana transformants, Figure S6. Growth curves during full scale experiment for P. tricornutum transformants, Figure S7. Chlorophyll fluorescence during full scale experiment in P. tricornutum transformants, Figure S8. Maximum quantum yield for P. tricornutum transformants during full scale experiment, Figure S9. Growth curves during full scale experiment for T. pseudonana transformants, Figure S10. Chlorophyll fluorescence during full scale experiment in T. pseudonana transformants, Figure S11. Maximum quantum yield for T. pseudonana transformants during full scale experiment, Figure S12. Sterol levels in T. pseudonana transformants, Table S1. L0 parts for construction of episomes using uLoop assembly method, Table S2. Source and number of transmembrane domains of HMGR sequences used for phylogenetic and domain analysis, HMGR sequences for phylogenetic reconstruction.