A universal pocket in fatty acyl-AMP ligases ensures redirection of fatty acid pool away from coenzyme A-based activation

Fatty acyl-AMP ligases (FAALs) channelize fatty acids towards biosynthesis of virulent lipids in mycobacteria and other pharmaceutically or ecologically important polyketides and lipopeptides in other microbes. They do so by bypassing the ubiquitous coenzyme A-dependent activation and rely on the acyl carrier protein-tethered 4′-phosphopantetheine (holo-ACP). The molecular basis of how FAALs strictly reject chemically identical and abundant acceptors like coenzyme A (CoA) and accept holo-ACP unlike other members of the ANL superfamily remains elusive. We show that FAALs have plugged the promiscuous canonical CoA-binding pockets and utilize highly selective alternative binding sites. These alternative pockets can distinguish adenosine 3′,5′-bisphosphate-containing CoA from holo-ACP and thus FAALs can distinguish between CoA and holo-ACP. These exclusive features helped identify the omnipresence of FAAL-like proteins and their emergence in plants, fungi, and animals with unconventional domain organizations. The universal distribution of FAALs suggests that they are parallelly evolved with FACLs for ensuring a CoA-independent activation and redirection of fatty acids towards lipidic metabolites.


25
The ANL superfamily includes enzymes such as the Acyl/Aryl-CoA ligases (ACS or FACLs), Adenylation 26 domains (A-domains) and Luciferases along with the recently identified Fatty acyl-AMP ligases (FAALs). 27 These enzymes are involved in the production of both primary metabolites such as acyl-CoA and 28 secondary metabolites such as antibiotics 1 , complex lipids 2 , cyclic peptides 3 and lipopeptides 4,5 . Basic 29 metabolic pathways such as β-oxidation, membrane biogenesis, post-translational modifications etc., 30 use primary metabolites such as acyl-CoA. The secondary metabolites such as complex lipids that 31 function as virulent molecules in Mycobacteria and bioactive molecules in several microbes that help 32 tide over unfavourable conditions and establish themselves in their niches. Such diverse metabolites 33 are produced by the members of the ANL superfamily through a two-step catalytic mechanism. It 34 begins with the activation of carboxylate-moiety of substrates such as fatty acids or amino acids by 35 ATP hydrolysis and finally transferring it to an acceptor such as CoA or holo-ACP. Multiple structural 36 and biochemical studies show that members of the superfamily such as FACLs and A-domains employ 37 a common pocket for the chemically identical CoA and the 4'-PPant moieties attached to the holo-38 ACP, respectively for the final transfer. It was later demonstrated that the A-domains can cross-react 39 with CoA to form aminoacyl-CoA 6 , which points to the liabilities of utilizing a common pocket for 40 binding chemically identical moieties. Infidelity towards the final acceptor has now been noted in 41 different classes of ANL superfamily members where Luciferases are shown to catalyse fatty acyl-CoA 42 formation 7 and FACLs producing bioluminescence with molecular oxygen 8 . While fatty acid/amino 43 acid substrate promiscuity is well studied and exploited in combinatorial biosynthesis of bioactive 44 molecules, the origin and basis of acceptor promiscuity is relatively less understood. 45 FAALs are atypical enzyme systems of the ANL superfamily as they completely lack acceptor 46 promiscuity, where they transfer the activated fatty acyl-AMP to the 4'-PPant of holo-ACP 9 but not 47

Identification of an alternative 4'-PPant-binding pocket in FAALs 129
The The entrance of the distinct tunnel is on the N-terminal side of the FSH, while the canonical pocket to 143 accommodate CoA is on the C-terminal side of the FSH. The approach towards the active site in both 144 cases is not on the same plane, but they coincide near the active site near the β-alanine of the 4'-145 PPant. The longest length along this pocket is aligned at ~25° to the canonical CoA-binding pocket. The 146 canonical pocket is mainly formed after the rotation of the C-terminal domain in the thioesterification 147 state (T-state) bringing the A8-motif near the active site. The space generated between the A8-motif 148 (from the C-terminal domain) and the subdomain-B of the N-terminal domain constitute the canonical 149 CoA-binding pocket. In contrast, the newly identified pocket is the space between the loops in 150 subdomain-A and the FSH region of subdomain-B of FAALs. These loop regions are highly variable in length but rich in prolines (occasionally threonine or serine) some of which are conserved in FAALs 152 such as P226 and P107 in EcFAAL (Figure-3b), while the FSH has a unique secondary structure 153 characteristic of FAALs. The structurally analogous sites in FACLs show a high degree of variability with 154 occasional prolines but the frequency of prolines at the indicated positions is poor and often replaced 155 by Asn/Leu (Supplementary Figure-4). The ability to consistently identify the tunnel with regions 156 enriched in prolines around the opening led to the proposition that this alternative pocket is perhaps 157 a universal attribute of FAALs that has evolved to accommodate the 4'-PPant arm. 158

The alternative pocket in FAALs is functional and accepts a 4'-PPant-tethered ACP 159
The ability of the alternative pocket to accommodate 4'-PPant arm tethered to ACP was tested using 160 structure-guided mutagenesis. Bulkier residues (Phe/Arg) were introduced at the entrance of the 161 tunnel that could potentially block the accessibility of the pocket for the incoming 4'-PPant arm. The 162 biochemical analysis of these mutants requires an assay system to monitor the transfer of acyl-AMP 163 to the 4'-PPant arm of holo-ACP. Typically, such acyl-transfers are assessed using SDS-PAGE 9,29 or 164 conformationally sensitive Urea-PAGE (CS-PAGE) 30 . The ACPs that accept the acyl-chain from FAALs 165 presented multiple complications such as poor conversion from apo-ACP to holo-ACP (Supplementary 166 Figure-5a) and lack of separation on a CS-PAGE (Supplementary Figure-5b). Therefore, the CS-PAGE 167 assay was modified for enhanced detection of the FAAL-dependent acyl-transfer on holo-ACP for the 168 first time using the radio-labelled fatty acids. We tested the efficacy of the modified radio-CS-PAGE 169 assay to probe three pairs of FAAL-ACP systems from diverse organisms (Figure-4a), viz FAAL-ACP pairs 170 from E. coli (EcFAAL-EcACP), Myxococcus xanthus (MxFAAL-MxACP) and Ralstonia solanacearum 171 (RsFAAL-RsACP). The appearance of bright bands on the radio-CS-PAGE indicates that the 172 radiolabelled fatty acid tethered to ACP, which is absent when apo-ACP is used, or ATP is omitted in 173 the reaction. Thus, the assay system can enable the simultaneous probing of multiple FAAL-ACP pairs 174 along with their mutations and facilitate similar studies in the future.
The modified radio-CS-PAGE was then used to test if mutating prolines (occasionally a threonine) 176 guarding the entrance of the pocket to bulkier residues can cause abrogation of the acyl-transfer 177 activity ( Figure-4c). It was found that a single point mutation, T83F or T83R, T252F or T252R and P107F  178 or P107R in EcFAAL can almost abrogate the acyl-transfer reaction on holo-EcACP (Figure-4c). It should 179 be noted that these mutations did not affect the adenylation ability of the proteins (Figure-4d). The 180 mutations in other FAAL-ACP pairs, MxFAAL-MxACP and RsFAAL-RsACP also resulted in abrogation of 181 the acyl-transfer ability on their respective cognate holo-ACPs. An additional FAAL-PKS pair of 182 MsFAAL32-MsPKS131-1042 was also mutated and probed biochemically using the traditional radio-SDS-183 PAGE. The mutations of residues guarding the alternative pocket to bulkier residues in the MsFAAL32-184 MsPKS131-1042 system also resulted in diminished acyl-transfer ability. These results indicate that 185 blocking the entrance of the alternative pocket prevents the entry of the incoming 4'-PPant arm 186 tethered to ACP. Therefore, in FAALs, the identified alternative pocket is fully functional and distinct 187 from the non-functional canonical pocket. Such a unique pocket facilitates the entry of the 4'-PPant 188 arm of the ACP to approach the active site and catalyse the acyl-transfer reaction, a feature absent in 189 the other members of the superfamily. 190 A universal mechanism for rejection of highly abundant CoA in the alternative pocket 191 The primary attribute of a functional alternative pocket in FAALs is to discriminate and reject CoA from 192 the chemically identical 4'-PPant of holo-ACP. Therefore, the pocket should allow the 4'-PPant entry 193 into the tunnel but not the additional "head group", adenosine 3',5'-bisphosphate moiety of CoA.  FACLs. Therefore, we propose that FAALs may not have descended from FACLs and they rather share 265 a parallel evolutionary history. The ancestral ANL fold could have diverged as non-promiscuous FAAL-266 like members and promiscuous nonFAAL-like members simultaneously in the last universal common 267 ancestor ( Figure-6b). 268

CONCLUSIONS 269
The study uncovers the mechanistic basis of how FAALs strictly reject CoA, which is highly abundant 270 and almost chemically identical to their actual substrate, holo-ACP. The rejection mechanism relies on 271 a discriminatory 4'-PPant-accepting pocket while avoiding the promiscuous canonical CoA-binding 272 pocket that is rendered non-functional. It has been achieved through bulky hydrophobic residues in 273 the pocket and a unique secondary structural element, FSH at the entrance. The discriminatory 4'-274 PPant accepting pocket in FAALs, on the other hand, has a unique architecture that negatively selects 275 adenosine 3', 5'-bisphosphate moiety and also lacks Arg/Lys residues for positive selection.
Interestingly, these rejection criteria are not only conserved in bacteria but also in all forms of life 277 (excluding archaea). The unique remodelling of pockets to ensure acceptor discrimination probably 278 allowed the evolutionary recruitment of FAAL-like domains in metabolic crossroads for redirecting the 279 fate of molecules to specific pathways. Such a wide-spread conservation of FAAL-like proteins puts 280 the evolutionary origin of FAALs parallel with FACLs in the last universal common ancestor and not as 281 a subset of FACLs. 282 The scaffold of the ANL superfamily of enzymes is known to be promiscuous not only for the substrates 283 they act on 33 but also the final acceptor of the acyl adenylates. Surprisingly, billions of years of 284 evolution has not prevailed upon the substrate-promiscuity problem as well as the acceptor-285 promiscuity problem. The probability of acceptor-promiscuity influencing the erroneous product 286 formation is high as some of the acceptors such as CoA and pantetheine are the most abundant 287 molecules in the cell. The problem is greatly amplified as it can redirect the fate of metabolites from 288 one pathway to another such as primary metabolism to secondary metabolism. Our analysis shows 289 that none of the solved CoA-bound structures of ANL superfamily members, even the protomers of 290 the same crystal (e.g., SeFACL), exhibit conformity in the binding mode of 4'-PPant or the adenosine 291 3',5'-bisphosphate moiety 15,34 . The variability in the 4'-PPant binding is also true for the A-domain:ACP 292 complexes 17,35,36 . Hence, it is evident that a defined CoA-binding pocket is lacking in ANL superfamily 293 members, which is commensurate with the failure to identify the pocket in FACLs using various pocket-294 search algorithms. Therefore, arbitrary access of the active site without any selection determinants is 295 likely to be the root cause of final acceptor promiscuity in the case of these enzymes. 296 It is not clear if these observed spectrum of latent activities in these enzymes' design is an evolutionary 297 relic or has any physiological relevance in a specific cellular context. The persistence of acceptor 298 promiscuity can only have two explanations: viz., the cross-reaction products are beneficial, or pocket 299 modification is not possible without compromising the basic function. In this context, FAALs are 300 surprisingly high-fidelity enzymes representing the extreme end of the promiscuity spectrum, offering no cross-reaction with CoA as acceptors of the acyl adenylates [9][10][11]37 . A functional and discriminatory 302 4'-PPant binding in FAALs have established them as an alternative enzymatic bridge between FA 303 synthesis, exogenous fatty acids import and PKS/NRPS machinery. FAALs act as a loading module in 304 PKS/NRPS-mediated biosynthesis of diverse bioactive natural products with fatty acids such as 305 mycosubtilin, daptomycin, micacocidin, ralsomycin, olefin, tambjamine, ambruticin, puwainphycins, 306 jamaicamide, columbamides to name a few. The relevance of the lack of acceptor promiscuity in FAALs 307 has been demonstrated in Mycobacteria as being responsible for dictating the fate of free-fatty acids 308 in producing virulent lipids 10 . 309 The presence of several elements of the CoA-rejection mechanism and their close identity to FAALs allow 336 us to put forth the hypothesis that eukaryotic FAAL-like domains were recruited for their high acceptor 337 fidelity property. It is also possible that these divergences have some functional relevance in the 338 context of eukaryotic metabolism, which may represent an additional member in the spectrum of 339 biochemical activities represented in the ANL superfamily. 340 The current work identifying FAAL-like enzymology using a universal CoA rejection mechanism may 341 form the platform for further studies to delineate why they have been recruited in fungal and animal 342 systems. The study provides new structural and sequence attributes to confirm the identity of FAALs, 343 many of which remains misannotated and uncharacterized. The study opens new avenues in 344 combinatorial engineering of PKS/NRPS by using FAALs as a unique module to load fatty acids or 345 engineer them to load unique molecules with exceptional fidelity. Thus, FAALs can be exploited to 346 produce novel bioactive molecules by virtue of their unique acceptor-fidelity property.

Biochemical analysis of FAALs and FACLs:
The acyl-AMP and acyl-CoA formation by FAALs and FACLs 358 were performed by previously described methods 10 . All FACLs (MtFACL13, EcFACL, AfFACL), EcFAAL 359 and MsFAAL32 were used at 5 μM concentration while MxFAAL and RsFAAL were used at 7.5 μM. The 360 14 C-fatty acids allowed detection of the products on a phosphorimager (Amersham Typhoon FLA 9000), 361 which were then quantified by densitometry using Image Lab (Bio-Rad Laboratories Inc.). All 362 experiments were performed as triplicates. The percentage of acyl-AMP converted to acyl-CoA from 363 the total acyl-AMP formed is used to plot and compare the activity, along with the standard error 364 of the mean, of wild-type against the respective mutants. 365 Conversion of apo-ACP to holo-ACP: All the purified ACPs were converted to holo-ACP before being 366 flash-frozen using previously described protocols 44 . Briefly, after Ni-NTA purification, they were 367 buffer exchanged to the phosphopantetheinylation buffer (20 mM Tris pH 8.8, 10 mM MgCl2 and 10 368 mM dithiothreitol). ~250 μM of ACP was then incubated with 1.25 μM of a non-specific 369 phosphopantetheinyl transferase Sfp from B. subtilis in a 300 μl reaction mixture along with 10 mM 370 MgCl2 and 1 mM CoASH. The reaction mixture was incubated at 25°C for 12-16 hours and further 371 purified using size-exclusion chromatography at 4°C. The holo-ACP was flash-frozen in liquid nitrogen 372 and stored at -80°C until further use. The conserved serine on which the 4'-PPant moiety is added, 373 post-translationally, is mutated to alanine and the resulting protein is used as apo-ACP for all assays. 374 holo-MsPKS131-1042 on an 8% SDS-PAGE. All gels were then dried using a gel drier (Bio-Rad gel dryer 387 583) and the radiolabelled acyl-ACP was detected by using the phosphorimager (Amersham Typhoon 388 FLA 9000). 389

Loading activated acyl chains on holo-ACP by
Sequence and structural analysis: All sequences were identified and retrieved from the NCBI 390 sequence database using EcFAAL as template and BLAST search algorithm. A structure-based 391 sequence alignment was generated using msTALI 45

COMPETING INTERESTS 416
The authors declare that they have no competing interests. 417

DATA AND MATERIALS AVAILABILITY 418
All data needed to evaluate the conclusions in the paper are present in the paper and/or the 419 Supplementary Materials. Additional data related to this paper may be requested from the authors. Chem Biol 3, 923-936, doi:10.1016/s1074-5521(96)90181-7 (1996)      in pale green) with other representative members of the ANL superfamily reveals that FAALs have a 599 higher frequency of prolines (occasionally Thr/Ser) than other members of the superfamily.