1. Introduction
Aurantiochytrium limacinum is an obligate heterotrophic marine unicellular protist belonging to the Thraustochytrid family. Thraustochytrids are present almost everywhere in the oceans, from tropical to polar areas, and from the surface down to 2000 m below sea level [
1,
2,
3,
4]. Because they are obligate heterotrophs, Thraustochytrids are more abundant in habitats containing decaying biological material, such as superficial sediment layers, mangroves, or river effluents [
3]. In the past decade, Thraustochytrids have attracted biotechnological interest because they naturally accumulate high levels of triacylglycerols (TAGs). TAGs represent a valuable source of fatty acids (FAs) for either human and animal health or green chemistry purposes [
5,
6]. The peculiarity of Thraustochytrids compared with many other microalgae is their strikingly high content of very long chain polyunsaturated fatty acids (VLCPUFA), mainly ω3-docosahexaenoic acid (DHA, 22:6) [
7,
8,
9,
10,
11]. DHA is synthesized at extremely low levels in animals and is therefore considered as a ‘conditionally essential’ FA [
12], which means it must be obtained from the diet. In humans, DHA accumulates in the brain and is required for the good visual and neural development in infants [
13]. Presently, the most widely and naturally available dietary source of ω3-VLCPUFAs is fish oil. However, overexploitation of fish stocks and their contamination by toxic substances such as heavy metals impose us to find alternative and more sustainable sources [
14]. For the reasons depicted above, Thraustochytrids are emerging models for fundamental and applied research.
FA synthesis in Thraustochytrids is more complex than in many other marine protists because it involves two different pathways that operate independently [
6]. The first pathway requires a type I Fatty Acid Synthase (FAS) system similar to that found in animals and produces FA chain length of 16 carbons (16C). The second pathway involves a Polyketide Synthase-like (PKS-like) machinery, or PUFA synthase, to produce VLCPUFAs of 20C and 22C [
15,
16,
17,
18]. Many efforts and many different strategies are currently being undertaken to elucidate how the two pathways are regulated and coordinated, and how the lipid production could be improved in Thraustochytrids. Indeed, as already reviewed [
5,
6], the literature is abundant about the impact of culture conditions, or about attempts of metabolic engineering based on genetic modifications. To date, most of the molecular approaches attempted to improve the VLCPUFA production by targeting anabolic pathways such as those involved in FA synthesis, including the FAS system [
19,
20] and the PUFA synthase complex [
21,
22,
23], or those involved in glycerolipid synthesis [
24,
25]. FA catabolism is also a potential target to enhance the lipid production. Genomic and qPCR analyses of
Aurantiochytrium limacinum and
Hondaea fermentalgiana [
11,
26] indicate that FA oxidation in Thraustochytrids occurs in both mitochondria and peroxisomes. FA catabolism first requires detaching the FAs from glycerolipids (TAGs), transporting them into peroxisomes, and activating them in the form of acyl-CoAs to allow their oxidation by the β-oxidation cycle. The products of the peroxisomal β-oxidation are then shuttled to mitochondria for a complete oxidation into CO
2 and H
2O [
26,
27,
28]. To our knowledge, there is, so far, only one publication exploring the catabolic side of FA metabolism in
Aurantiochytrium. The authors showed that disrupting the genes coding for acyl-CoA oxidases resulted in a higher FA productivity [
29].
Genetic transformation by biolistic (i.e., using particle bombardment) offers the advantage to deliver any form of RNA, DNA, or protein in almost any type of cell, including those with thick walls or silica walls such as diatoms [
30]. This relatively simple-to-use method is widely used to transform plants [
31], which prompted specific efforts to estimate the genetic instability and unintended consequences resulting from such approaches. Indeed, biolistic violently integrates DNA, in a rather random way. Although some transgenic events appear as simple insertions with no other evident genome damage, others show serious genome damages including chromosome breakages and large deletions [
32] or transpositional activation [
33]. Thus, a successful biolistic transformation also relies on a combination of various and complex mechanisms of DNA repair. A corollary is that the process for biolistic transformation can be also used for the production of some random mutations.
In the present work, following a biolistic attempt to transform a strain of
Aurantiochytrium limacinum (CCAP 4062/1) [
11], we isolated by serendipity a clone that did not integrate any recombinant DNA but displayed two to three times more FAs than the original strain. Using metabolic, transcriptomic, and genomic analyses, we identified the biolistic side effects at the origin of this phenotype. The results pointed out a potential target indirectly involved in FA oxidation, which could significantly increase the lipid content in
A. limacinum when down regulated.
4. Discussion
After a biolistic treatment aimed to produce zeocin-resistant mutants, we unexpectedly selected a clone displaying a strong lipid phenotype. Metabolic analyses pointed out a significant lipid accumulation resulting from a deficit in FA catabolism. Surprisingly, the whole genome and transcriptome sequencing did not identify any recombinant DNA, indicating that the phenotype was not the result of genetic material transfer. Rather, it likely resulted from the biolistic method itself, the particle projections provoking some DNA damages and rearrangements in nuclei [
32]. Expression analyses of genes directly involved in lipid metabolism could not corroborate our observations either. In particular, the genes involved in the mobilization, transport, and oxidation of FAs were not significantly downregulated in the mutant, as one would have expected if they were to be responsible for the lower FA catabolism. However, a genomic search for structural variants identified a sizeable deletion in one of the two alleles coding for a putative peroxisomal adenylate transporter. Phylogenetic analyses and yeast complementation experiments demonstrated that the affected gene was indeed a peroxisomal adenylate carrier.
An impaired ability to import ATP into peroxisomes upholds the observed phenotype. Indeed, the catabolism of FAs in peroxisomes involves several steps. First, FAs must be transferred from the cytosolic compartment to the peroxisomes. Two independent pathways are likely involved to transport FAs across peroxisomal membranes. The first pathway transports acyl-CoA chains through an ATP Binding Cassette (ABC) transporter, such as Pxa1p/Pxa2p in yeast or CTS in plants [
51,
52]. Although acyl-CoA chains may enter yeast peroxisomes directly as CoA esters, it is also possible that the CoA moiety is released during the transfer, either into the cytosol or into the peroxisomal lumen [
48,
51,
53]. CoA release prevails in plants [
49,
52]. Once inside the peroxisomal lumen, the corresponding free fatty acids (FFAs) are activated as CoA-esters before entering the β-oxidation cycle. This reaction is catalyzed by a peroxisomal acyl-CoA synthetase (Faa2p in yeast). A second pathway involves an unidentified transporter able to import short to medium FFA chains [
45,
48,
52]. Once in the peroxisome, the short to medium FFAs are activated as acyl-CoAs. Peroxisomal acyl-CoA synthetases catalyze ATP-dependent reactions and require the import of ATP from the cytosol. The peroxisomal adenylate carrier [PNC in
Arabidopsis thaliana, ANT1 (YPR128c) in
Saccharomyces cerevisiae] belongs to the Mitochondrial Carrier Family and catalyzes the import of ATP into peroxisomes in a strict counter exchange with AMP or ADP [
46,
49]. In
Saccharomyces cerevisiae, which does not accumulate lipids, a mutant devoid of peroxisomal ATP carrier (
ΔScant1) displayed an impaired growth when C12:0 was the unique source of carbon, whereas normal growth was observed with longer chains (C18:0) [
45]. In the oleaginous yeast
Yarrowia lipolytica, a
ΔYlant1 displayed a 20% increase of total FAs, whereas a mutant
ΔYlant1ΔYlpxa1ΔYlpxa2 accumulated twice as many FAs as the WT [
48]. This indicates that both FFAs and acyl-CoA are likely to enter yeast peroxisomes. In
Arabidopsis thaliana, suppressing the ATP import into the peroxisomes strongly impaired the breakdown of storage oil. Consequently, the total level of FAs increased 10-fold in the seedlings, and oil bodies accumulated. Mutant plants were defective in seedling growth and development, unless in the presence of sucrose [
49]. This illustrates the preponderant role of the peroxisomal adenylate carrier in plant β-oxidation and suggests that mainly FFAs enter plant peroxisomes.
We did not succeed so far to stably transform
A. limacinum strain CCAP 4062/1, thus reverse genetics’ strategies could not be used to demonstrate that the lipid phenotype described here resulted solely from a peroxisomal deficiency in ATP import. However, a number of pieces of evidence pinpointed the peroxisomal adenylate carrier as a very likely candidate. First, one of the two alleles coding for AlANT1 was seriously damaged during the biolistic treatment. The mutant displayed a lower expression of
AlANT1, suggesting that the truncated gene was not or only minimally expressed. In addition, if translated, the truncated protein would lack two out of the three transmembrane domains, and would, therefore, most likely be inactive. Noteworthy, the truncated gene expressed in
S. cerevisiae did not produce any truncated protein (
Supplementary Figure S3), suggesting that such a protein was either not translated or rapidly degraded. Second, the lipid phenotype in LAS was associated with a lower FA catabolism activity. This phenotype is similar to those described for plant
pnc and yeast
ant1 mutants. In response to this metabolic event, several compensatory mechanisms could be observed. In particular, genes involved in TAG degradation and acyl-CoA production in the cytosol (DAG lipase, acyl-CoA synthetase/ligase), genes for peroxisomal β-oxidation (acyl-CoA synthetase/ligase, the bifunctional enzyme Ehhadh), and those for mitochondrial β-oxidation (carnitine palmitoyltransferase, acyl-CoA dehydrogenase) were upregulated. More surprisingly, the genes involved in FA synthesis (acetyl-CoA carboxylase, FAS, PUFA synthase) were also upregulated. We have no clear explanation for this, but it is possible that the decrease in FA catabolism is perceived as a deficit in FA availability. The upregulation of lipid synthesis genes could contribute, together with the decrease in AlANT1 activity, to the large FA accumulation. These potential metabolic events are summarized in
Figure 7.
Interestingly, the C16:0/C22:6 ratio fluctuated along the growth in both strains. In WT, C16:0 decreased faster than C22:6 during the rapid oil breakdown phase, whereas it increased faster than C22:6 in the oil accumulation phase. This could reflect a faster dynamics/turnover for C16:0 than for C22:6, with C16:0 being more rapidly mobilized/oxidized than 22:6, and more rapidly synthesized by the FAS system than C22:6 by the PUFA synthase complex. A higher proportion of C16:0 was also observed in LAS (35% of total FAs) compared with WT (20% of total FAs). In LAS, the genes encoding FAS and PUFA synthase were similarly upregulated (log2FC = 1.3 and 1.4–2, respectively), and it is difficult to estimate whether or not these regulations can support the observed change in the C16:0/C22:6 ratio. However, it is also possible that a sorting in the import of FAs into peroxisomes exists in A. limacinum, as it does in S. cerevisiae. If this is true, more C16:0 than C22:6 would enter peroxisomes as FFAs, and, thus, the β-oxidation of C16:0 in peroxisomes could be more ATP dependent than C22:6.
Biolistic transformation is a convenient method, well adapted to organisms having thick cell walls such as plants [
54,
55] or silica shells such as diatoms [
30]. However, as already reported [
32,
33,
56], it must be kept in mind that interpretation of a biolistic transformation needs caution. The results presented here appear relatively simple since no transgenic material was transferred into the genome of
A. limacinum. In this particular case, the biolistic experiment can be seen as a random mutagenesis experiment, possibly enhanced by the action of the selection agent, the zeocin, which induces double-strand breaks [
57]. In cases where the transfer of genetic material is effective, it may be difficult to determine, without sequencing the entire genome, what is due to transformation itself and what is due to collateral damages. The parallel study of independent, transformed lines is a pragmatic way to address this technical issue. Sequence breakage and reassembly are common events with biolistic transformation [
56,
58]. DNA repair involves mechanisms such as non-homologous end joining (NHEJ) or homology directed repair (HDR), eventually leading either to simple insertion with no other evidence of genome damage, chromothripsis-like rearrangements, or genomic deletions [
32]. It is also possible that the specific nature of
A. limacinum CCAP 4062/1 contributes to increase the probability to get genomic impairment events during biolistic treatments. Indeed, several cell types characterize the
A. limacinum life cycle, including mononucleated and multinucleated cells [
6,
35]. The presence of cells with several nuclei might increase the complexity of events resulting from the transformation process and the bombardment procedure.