Molecular mechanisms in fungal fatty acid synthase (FAS) assembly

The fungal fatty acid synthase (fFAS) multienzyme is a barrel-shaped 2.6 MDa complex comprising six times eight catalytic domains. Upon barrel-formation, up to several hundred kDa large polypeptides intertwine to bury about 170,000 Å2 of protein surface. Functional, regulatory and structural data as well as evolutionary aspects of fFAS have been elucidated during the last decades. Notwithstanding a profound knowledge of this protein family, the biogenesis of the elaborate structure remained elusive. Remarkably, experimental data have recently demonstrated that fFAS self-assembles without the assistance of specific factors. Considering the infinitesimal probability that the barrel-shaped complex forms simply by domains approaching in the correct orientation, we were interested in understanding the sequence of events that have to orchestrate fFAS assembly. Here, we show that fFAS attains its quaternary structure along a pathway of successive domain-domain interactions, which is strongly related to the evolutionary development of this protein family. The knowledge on fFAS assembly may pave the way towards antifungal therapy, and further develops fFAS as biofactory in technological applications.


47
Fatty acid synthases (FAS) have been structurally studied during the last years, and a deep 48 understanding about the molecular foundations of de novo fatty acid (FA) synthesis has been 49 achieved 1-3 (Figure S1A and B). The architecture of fungal FAS (fFAS) was elucidated for the 50 proteins from Saccharomyces cerevisiae (baker's yeast) 4-6 and the thermophilic fungus Thermomyces 51 lanuginosus 7 , revealing an elaborate 2.6 MDa large α 6 β 6 barrel-shaped complex that encapsulates 52 fungal de novo FA synthesis in its interior ( Figure 1A). The functional domains are embedded in a 53 scaffolding matrix of multimerization and expansion elements. Acyl carrier protein (ACP) domains,

54
shuttling substrates and intermediates inside the reaction chamber, achieve compartmentalized 55 synthesis 4,8 (Figure 1B and C). The concept of metabolic crowding makes fFAS a highly efficient 56 catalytic machinery, running synthesis at micromolar virtual concentrations of active sites and 57 substrates 9 . The outstanding efficacy in fungal FA synthesis is documented by (engineered) oleagenic 58 yeast that can grow to lipid cellular contents of up to 90% 10 . fFAS have also raised interest as 59 biofactories in microbial production of value-added compounds from saturated carbon chains [11][12][13] .

60
Facing the complexity of the fFAS structure, we recently started the project of deciphering its assembly 61 mechanism. We were interested in two aspects. First, based on the observation that fFAS can be 62 recombinantly expressed in E. coli 14,15 , it can be posited that specific assembly factors are not 63 required for fFAS biogenesis. Autonomous self-assembly of fFAS may essentially be envisioned by 64 distributing the complexity of the assembly process onto a sequence of domain-domain interactions 65 that are formed one after another. We aimed to explore this sequence of events and to analyze 66 whether it can be correlated to the evolutionary development of fFAS, since it has been suggested that 67 assembly pathways generally reflect protein evolution 16 . Second, we sought to evaluate whether the 68 knowledge on fFAS assembly may be exploited for inhibiting de novo fungal FA synthesis in selective 69 antifungal therapy [17][18][19] , as well as for designing fFAS based biofactories 20 .

70
Our studies of the S. cerevisiae FAS assembly were greatly aided by engineering fFAS on the basis of 71 the available atomic resolution models [4][5][6][7]21 . Wildtype and several engineered S. cerevisiae FAS 72 constructs were used for complementing a FAS-deficient yeast strain. Full-length and truncated S.

73
cerevisiae FAS constructs were further recombinantly expressed in Escherichia coli. These tools in 74 hand, we were able to address fFAS assembly in a "forward-approach", which means that instead of 75 often-performed dissociation based ("reverse") approaches, we generated information based on halted 76 assembly states and truncated structures. Here, we present a multitude of data suggesting that S.

101
The fFAS family is topologically heterogeneous on gene level: Genome sequence analysis has 102 characterized fFAS as a heterogeneous family comprising different gene-topological variants ( Figure   103 2A). As most evident gene-topological variation, fFAS are either encoded by single genes or by two 104 genes. Two-gene-encoded fFAS appear to originate from a single-gene encoded precursor split into 105 two at various fission sites that are generally located within domains 3,22 . In S. cerevisiae and T.

106
lanuginosus FAS, both representing the Ascomycota-type fFAS, the C-terminus of the β-chain and the 107 N-terminus of the α-chain intertwine to form the MPT domain (Figure 2Bi

119
Abbreviations used as in Figure 1. Four fFAS variants differing in fission sites as well as in the distribution of 120 6 insertion elements are given (missing insertions in dark grey

131
In a first experiment, we analyzed the assembly of above described fFAS variants by constructing interaction may also be seen as event happening prior to the actual assembly (as the specific process 149 of barrel formation) that captures all variants to assemble via a single assembly pathway; in line with 150 the conception of the high evolutionary conservation of assembly pathways in protein families 26 . We 151 term this assembly step "pseudo-single chain formation" in the following.

152
To evaluate the impact of insertion elements on fFAS assembly, we further engineered fFAS

161
SEC and TSA data show compromised stability of the proteins with deleted insertion elements,

268
Post-translational modification occurs within a dimeric sub-structure: In a stepwise deconstruction 269 approach, we dissected fFAS into domains and multi-domain constructs, which we then analyzed in

279
The β-chain was not proteolytically stable as a separate protein in E. coli, which impaired in vitro 280 assembly experiments.

282
We dissected the α-chain from its C-terminus, and initially probed the role of the C-terminal PPT as a

309
Data collected on the truncated α-chain constructs imply that the phosphopantetheinylation active 310 species is dimeric, organized by the KS dimer as the prominent structural unit. It can further be 311 concluded that the sequence ACP-KR-KS-PPT bears the information for forming the 312 phosphopantetheinylation competent complex, but not for forming the D3 symmetric α 6 -wheel 313 structures. Since the α-chain constructs run into aggregation, but are nevertheless 314 phosphopantetheinylated, it seems that the phosphopantetheinylation status of fFAS is not proofread 12 during assembly. For confirming this result, we analyzed the phosphopantetheinylation-deficient 316 S180A S. cerevisiae FAS in our assembly assay (see Figure 3C). The mutated construct was unable 317 to restore de novo FA synthetic activity in the complementation assay, but indeed assembled to the 318 α 6 β 6 complex, supporting an assembly process that does not supervise post-translational

393
As shown for S. cerevisiae FAS, fFAS assembly is robust and tolerates gene fusion and alternative 394 gene fission (see Figure S3). Our data suggest that the key players driving assembly are mainly the 395 catalytic domains that successively interact during assembly. The insertion elements stabilize the final 396 barrel-shaped structure, and seem to be of minor significance for assembly except evolutionary 397 ancient motifs as e.g. DM3 22 . fFAS is the most efficient de novo fatty acid (FA) synthesizing protein 9 .

398
This property makes fFAS an attractive object in the endeavor to achieve microbial production of FA

413
hampers the assembly to fully active intact protein, since interfering in pseudo-single chain formation.

414
Such strategies can now be avoided when following the here presented guidelines 43 . 536 537