Besides the housekeeping AcpP proteins (CC_1677 in C. crescentus and ECD_01090 in E. coli BL21(DE3)), both bacteria possess a structural gene for putative ACPs (CC_1163 and ECD_02853, respectively), located just upstream of their respective spt genes. These four structural acp genes were cloned and overexpressed in E. coli. Separation of cell-free protein extracts by 20% native polyacrylamide gel electrophoresis (PAGE) and subsequent staining with Coomassie blue (Figure 2A) indicates that ACPs and putative ACPs were readily expressed from all cloned genes (lanes 2–8) and showed characteristic migration behavior. In parallel, in vivo labeling experiments with radioactive β-alanine, the biosynthetic precursor of 4'-PPT, were performed with the same strains and after separating cell-free extracts by PAGE, radioactive bands were visualized (Figure 2B). For 4 known ACPs (ECD_01090 = AcpP_Ec, lane 3; NodF, lane 6; AcpXL, lane 7; RkpF, lane 8), radiolabeled bands with similar relative mobility as the overexpressed proteins were detected. Also, the putative ACPs (ECD_02853 = AcpR_Ec, lane 2; CC_1677 = AcpP_Cc, lane 4; CC_1163 = AcpR_Cc, lane 5) displayed radiolabeled bands that had similar relative mobility as the overexpressed protein bands (Figure 2), suggesting that in addition to the housekeeping AcpP_Cc from C. crescentus, the specialized predicted ACPs AcpR_Ec from E. coli and AcpR_Cc from C. crescentus also contained β-alanine, the building block of the 4'-PPT prosthetic group characteristic for ACPs. The autoradiogram also reveals that the housekeeping AcpP_Ec from E. coli, which is present in all samples, including the E. coli strain harboring the empty vector pET9a, migrates close to the front and incorporated β-alanine (Figure 2B). In an attempt to obtain a better separation of the distinct ACPs, the same samples analyzed for Figure 1 were also subjected to a SDS-Tricine PAGE (Supplementary Figure 1). Although the housekeeping AcpP_Ec from E. coli is known to migrate much slower than expected from its molecular weight in SDS-containing gel systems (Rock and Cronan, 1979) and although it is well separated from other ACPs in an SDS-Tricine PAGE (Supplementary Figure 1), all other ACPs migrated similarly, to each other and according to molecular weights of about 10 kDa (Supplementary Figure 1).

To obtain acylated versions of ACPs from C. crescentus and E. coli, the four distinct ACPs AcpP_Cc (CC_1677), AcpR_Cc (CC_1163), AcpP_Ec (ECD_01090), and AcpR_Ec (ECD_02853) were expressed in E. coli BL21(DE3) × pLysS and purified (Supplementary Figure 2). Purified ACP preparations contained mixtures of apo- and holo-forms and treatment with the tool of a holo-ACP synthase, as previously described (Ramos-Vega et al., 2009), converted these mixtures into the holo-forms of the respective ACPs (Supplementary Figure 3). Acyl-ACP synthetase (Aas) from Vibrio harveyi (Aas_Vh) is known for its broad substrate tolerance for many fatty acids and distinct ACPs (Beld et al., 2014). The distinct holo-ACPs were incubated with ATP, fatty acids of various chain lengths and homogenous Aas_Vh (Supplementary Figure 4A) and at the end of the assay, potentially acylated versions of the ACPs were separated in a conformationally sensitive PAGE system and visualized by Coomassie blue staining (Supplementary Figure 5). The housekeeping AcpP_Ec from E. coli (ECD_01090) maintains a more compact structure when acylated with fatty acids and so migrates faster when analyzed by conformationally sensitive PAGE (Beld et al., 2014). When unacylated, AcpP_Ec has a less compact structure under the electrophoretic conditions employed and migrates slowly. Acylation with fatty acids of increasing chain lengths causes more compact structures of the acylated AcpP and more rapid migration (Supplementary Figure 5A). Similarly, acylated forms of the housekeeping AcpP from C. crescentus (CC_1677) migrate more rapidly in conformationally sensitive PAGE suggesting a more compact form (Supplementary Figure 5B). In the case of specialized AcpR_Cc, potentially acylated forms of AcpR_Cc might migrate slightly faster than the apo-form and slightly slower than the holo-form (Supplementary Figure 5C). However, as no major change in migration was observed, there is no evidence that AcpR_Cc can be acylated by Aas_Vh. The potentially different forms of specialized AcpR_Ec all migrate in a similar way (Supplementary Figure 5D) and therefore this type of analysis does not permit us to evaluate whether acylation of AcpR_Cc or AcpR_Ec occurred when Aas_Vh was employed as acylating enzyme.

To explore which thiol substrates could be acylated by which acylating enzymes, homogeneous preparations of Aas from V. harveyi (Aas_Vh) (Supplementary Figure 4A), predicted ACS AasR_Cc (CC_1165) from C. crescentus (Supplementary Figure 4B), and acyl-CoA synthetase from Sinorhizobium meliloti (FadD_Sm) (Supplementary Figure 4C), were obtained. Each of the homogeneous enzymes was incubated with different potential thiol substrates (CoA, housekeeping AcpP_Cc, or specialized AcpR_Cc), ATP, and [³H]palmitate and at different time points, aliquots were analyzed for radioactivity linked to protein, in the cases of AcpP_Cc and AcpR_Cc, or to CoA. From the time course for acyl-thioester formation, it is evident that putative ACS AasR_Cc from C. crescentus caused an increase in the amount of acyl-AcpR_Cc whereas no acyl-thioesters were formed when AcpP_Cc or CoA were used as substrates (Figure 3A). When different potential thiol substrates were incubated with ATP, [³H]palmitate and Aas_Vh, an increase in the amount of acyl-AcpP_Cc was detected, whereas only very minor increases might have occurred when AcpR_Cc or CoA were used as substrates (Figure 3B). Using FadD_Sm in the acylation assay resulted in an increase in the amount of acyl-CoA whereas no acyl-thioesters were formed when AcpP_Cc or AcpR_Cc were used as substrates (Figure 3C). These results clearly show that acyl-CoA synthetase FadD is able to form acyl-CoA but cannot acylate the housekeeping AcpP or specialized ACP AcpR from C. crescentus, that the acyl-ACP synthetase Aas from V. harveyi is able to acylate the housekeeping AcpP from C. crescentus, and that the putative ACS is a specialized acyl-ACP synthetase (AasR) that selectively acylates the specialized AcpR_Cc but not CoA or the housekeeping AcpP_Cc.

In order to confirm that acyl residues are covalently linked to the respective ACPs from C. crescentus, acylation assays with elevated enzyme concentrations of AasR_Cc, Aas_Vh, or FadD_Sm were employed to achieve near complete conversion of holo-ACPs to acyl-ACPs. Potentially acylated ACP preparations were separated by conformation-sensitive 5 M urea-PAGE. Gels containing samples obtained with non-radioactive palmitate were stained with Coomassie blue (Figure 4A). Potential acyl-ACP samples generated by using [³H]palmitate were separated by 5 M urea-PAGE and autoradiograms were obtained (Figure 4B).

In acylation assays involving non-radioactive palmitate, both ACPs AcpR_Cc and AcpP_Cc changed their migration in the conformation-sensitive urea-PAGE. Acyl-AcpR_Cc migrated slightly slower than the holo-form (between the apo- and the holo-forms) after being incubated with the AasR_Cc from C. crescentus (Figure 4A), which corresponds to the band that had incorporated [³H]palmitate in the radiolabel assay (Figure 4B). When holo-AcpR_Cc was treated with Aas from V. harveyi or with FadD from S. meliloti the slight shift in migration of AcpR_Cc did not occur (Figure 4A) nor was any radioactive palmitate attached to AcpR_Cc (Figure 4B). This result confirms that the putative ACS is really an acyl-ACP synthetase that can covalently link a fatty acyl residue to the specialized AcpR_Cc for sphingolipid biosynthesis in C. crescentus. Incubation of AcpP_Cc with Aas from V. harveyi led to the formation of a protein band that migrated as expected for an acyl-AcpP_Cc and much faster than the unacylated AcpP_Cc (Figure 4A). Also, the fast-migrating band had incorporated radiolabeled palmitate (Figure 4B), confirming that this substance is acyl-AcpP_Cc. However, when AcpP_Cc was treated with AasR_Cc from C. crescentus or with FadD from S. meliloti, the strong shift in migration of AcpP_Cc did not occur (Figure 4A) nor was any palmitate attached to AcpP_Cc in radiolabel assays (Figure 4B). These results show that Aas from V. harveyi is an efficient acylation tool for AcpP, but not for AcpR.

Mass spectrometric analyses of caulobacterial ACPs revealed an average mass of 9,020.3 ± 0.4 Da for AcpR (Figure 5A), which is close to the theoretical molecular weight of 9,022.4 Da calculated from the amino acid sequence for the apo-form of AcpR using the ExPASy – SIB Bioinformatics Resource Portal.¹ Similarly, an average mass of 8,599.1 ± 0.3 Da was determined for the purified AcpP preparation (Figure 5B). However, the theoretical molecular weight calculated from the complete amino acid sequence amounted to 8,732.4 Da. In proteins where the second amino acid residue is a serine, the N-terminal methionine residue is usually removed by methionine aminopeptidase (Frottin et al., 2006) to generate mature proteins. This post-translational methionine removal is frequently found for housekeeping AcpPs (Platt et al., 1990). An apo-AcpP that lacks the N-terminal methionine (mass 131.2 Da) would have a theoretical mass of 8,601.7 Da, which is in close agreement with the determined average mass of 8,599.1 ± 0.3 Da (Figure 5B). Treatment of apo-ACPs with holo-ACP synthase AcpS leads to species with an average mass of 9,360.8 ± 0.8 Da in the case of AcpR (Figure 5C) or of 8,939.8 ± 0.9 in the case of AcpP (Figure 5D), corresponding to mass increases of 340.5 or 340.7, respectively, when compared to the apo-forms, and as would be expected for the addition of a 4'-PPT group (mass 339). Treatment of holo-AcpR with AasR in the presence of palmitate and ATP led to a species with an average mass of 9,601.2 ± 1.3 Da (Figure 5E) and the observed mass increase of 240.4 is in agreement with the attachment of a palmitoyl residue (+238) to the 4'-PPT arm of AcpR. When holo-AcpP was treated with Aas_Vh in the presence of palmitate and ATP a species with an average mass of 9,180.3 ± 1.3 Da was formed (Figure 5F) and the observed mass increase of 240.5 is in agreement with an attachment of a palmitoyl residue to the 4'-PPT arm of AcpP. However, in this latter case, conversion of holo-AcpP to palmitoyl-AcpP was not complete (Figure 5F), probably due to the fact that the acylation reaction was performed in the absence of Triton X-100 to avoid serious contaminations of the mass spectrometer. Theoretical and measured masses as well as the relative errors are given for the distinct caulobacterial ACP versions (Supplementary Table 1). In summary, only acyl-ACP synthetase from V. harveyi is able to acylate the housekeeping AcpP and only AasR can efficiently acylate the specialized AcpR from C. crescentus.

In order to measure AasR activity in a reproducible way, we investigated the influence of some components on the enzyme assay. The effect of the non-ionic detergent Triton X-100 concentration on the standard AasR assay was studied (Supplementary Figure 6), demonstrating that under these conditions there was a strong detergent dependence for the AasR activity with a maximal activity at 1 mM (0.06%; w/v) Triton X-100. At higher concentrations, Triton X-100 quickly became inhibitory, indicating that substrate dilution kinetics was followed (Eaton and Dennis, 1976).

Also, the effect of the AcpR substrate concentration on AasR activity was studied (Figure 6). Clearly, the initial velocity V₀ for the AasR-catalyzed reaction does not follow a Michaelis-Menten-like hyperbolic dependence on AcpR substrate concentration. In contrast, AasR exhibited a sigmoidal AcpR activation curve, with a measured maximum velocity (V_max = 12 pmol/min) at 40 μM of AcpR. The constant, where half of V_max is obtained (K_0.5 = 24 μM), indicates a high affinity of AasR for its AcpR substrate. A Hill coefficient of n_H = 1.98 indicates positive cooperativity for AcpR binding to the AasR. Surprisingly, higher concentrations of AcpR (>40 μM) inhibit the AasR-catalyzed reaction (Figure 6).

The lipid profiles of E. coli strains expressing individual genes or combinations of genes required for 3-oxo-sphinganine synthesis were analyzed by TLC. If the combination of AcpR_Cc and AasR_Cc or individual SPTs (SPT_Cc, SPT_Ec, or SPT_Sw) were expressed, the lipid profile of those samples was similar to the profile obtained from an E. coli strain harboring empty vectors pET9a and pCDFDuet-1 (Figure 7A). We had previously shown that the combined expression of AcpR_Cc, AasR_Cc and SPT_Cc in E. coli led to the formation of a compound that migrated like 3-oxo-sphinganine in TLC analyses and high resolution mass spectrometry analysis of such lipid extracts indicated the presence of a compound with m/z = 300.28953 expected for C18-3-oxo-sphinganine (Olea-Ozuna et al., 2021). We now demonstrate that the combined expression of AcpR_Cc/AasR_Cc, together with any of the three SPTs (SPT_Cc, SPT_Ec, or SPT_Sw) studied, led to abundant formation of compounds that migrated like the 3-oxo-sphinganine standard (Figure 7A). Also, in the cases of all three SPTs, compounds that migrated like sphinganine or ceramide were formed, possibly by intrinsic enzymes of E. coli BL21(DE3) (Supplementary Figure 7). The gene ECD_02852 might encode a keto-sphinganine reductase and convert 3-oxo-sphinganine to sphinganine and the gene product of ECD_02850 might be an N-acyltransferase that converts sphinganine to (dihydro)ceramide. Surprisingly, all newly formed compounds were more intense when the previously characterized SPT from S. wittichii (Raman et al., 2010) was employed (Figure 7A). These results suggest that the acylated AcpR_Cc serves as an excellent substrate for each of the three SPTs investigated.

To further clarify the requirement of ACPs for 3-oxo-sphinganine formation, housekeeping and specialized ACPs from C. crescentus and E. coli BL21, respectively, were coexpressed with acyl-ACP synthetase (AasR_Cc) and serine palmitoyltransferase (SPT_Cc) from C. crescentus. 3-oxo-sphinganine formation was observed when AcpR_Cc was coexpressed (Supplementary Figure 8A), but not upon coexpression of any of the housekeeping ACPs (AcpP_Cc or AcpP_Ec). When AcpR_Ec from E. coli was coexpressed, a compound migrating like 3-oxo-sphinganine was also formed, but to a lesser extent than upon expression of AcpR_Cc (Supplementary Figure 8A). Therefore, AasR_Cc can acylate both, its native AcpR_Cc and to a lesser extent AcpR_Ec from E. coli. Therefore, in this tripartite system (AasR_Cc AcpR_Cc SPT_Cc) the specific AcpR_Cc from C. crescentus is required for efficient 3-oxo-sphinganine formation. Also, when we studied coexpression of distinct acyl-CoA/acyl-ACP synthetases (AasR_Cc Aas_Vh FadD_Sm) with the specialized ACP (AcpR_Cc) and serine palmitoyltransferase (SPT_Cc) from C. crescentus, 3-oxo-sphinganine was only formed when the specific acyl-ACP synthetase (AasR_Cc) from C. crescentus was coexpressed (Supplementary Figure 8B), but not upon coexpression of an acyl-ACP synthetase (Aas_Vh) from V. harveyi or coexpression of an acyl-CoA synthetase (FadD_Sm) from S. meliloti.

Finally, upon coexpression of cognate acyl-ACP synthetase (AasR_Ec), serine palmitoyltransferase (SPT_Ec), and specialized ACP (AcpR_Ec) from E. coli, formation of 3-oxo-sphinganine was observed (Figure 7B). This did not occur when AcpR_Cc and SPT_Cc from C. crescentus were coexpressed with AasR_Ec from E. coli. However, when AcpR_Ec and SPT_Ec from E. coli were coexpressed with AasR_Cc from C. crescentus a compound migrating like 3-oxo-sphinganine was formed. These data suggest that AasR_Cc from C. crescentus can acylate its native AcpR_Cc and to a lesser extent AcpR_Ec from E. coli. In contrast, AasR_Ec from E. coli can acylate AcpR_Ec from E. coli, but not AcpR_Cc from C. crescentus. Therefore, the specific interaction between AasR and AcpR from the same organism seems crucial for acyl-AcpR and efficient 3-oxo-sphinganine formation.

To determine which potential thioester substrates (acyl-CoA, acyl-AcpR_Cc or acyl-AcpP_Cc) were preferred by each SPT, enzymatic assays were performed similar to those described by Ikushiro et al. (2001). Cell-free extracts of E. coli BL21(DE3) × pLysS, in which SPT from C. crescentus, E. coli BL21(DE3) or S. wittichii (Raman et al., 2010) had been overexpressed, were used to study 3-oxo-sphinganine formation in the presence of L-[¹⁴C]serine and one of the potential thioester substrates.

The in vitro formation of a minor compound migrating like 3-oxo-sphinganine in TLC was observed when using the SPTs from C. crescentus, E. coli BL21(DE3) or S. wittichii, and palmitoyl-CoA as thioester substrate (Figure 8). However, no formation of 3-oxo-sphinganine was observed when using cell-free E. coli extracts harboring the empty plasmid pET9a. When palmitoyl-AcpR_Cc was employed as thioester substrate, several lipidic compounds, including a major compound migrating like 3-oxo-sphinganine were formed by SPTs from C. crescentus, E. coli BL21(DE3), or S. wittichii but none in the absence of any SPT (EV) (Figure 8). Presently, we do not know the chemical nature of the other compounds formed in an SPT-dependent way. However, as 3-oxo-sphinganine possesses both functional groups, a primary amine and an oxo group, required for abiotic, chemical imine (Schiff base) formation, it is likely that 3-oxo-sphinganine undergoes condensations by Schiff reactions with itself and other compounds carrying oxo- or primary amine groups. In fact, the fastest migrating compound in Figure 8 is also formed from a 3-oxo-sphinganine standard when incubated with buffer only. When palmitoyl-AcpP was used as thioester substrate, only minor amounts of a compound that migrated like 3-oxo-sphinganine were formed by all three SPTs, suggesting that palmitoyl-AcpP_Cc is not a good substrate for any of these three rhodobacterial SPTs (Figure 8). Therefore, for all three SPTs from the Rhodobacteria assayed [C. crescentus, E. coli BL21(DE3), and S. wittichii], palmitoyl-AcpR_Cc was much preferred as thioester substrate over the acylated AcpP_Cc or acyl-CoA derivatives.

Kinetic studies with purified SPTs from S. paucimobilis and S. wittichii suggest that the affinities for the substrates serine (K_m 1 mM) and palmitoyl-CoA (K_m 30 μM) are similar for both enzymes (Raman et al., 2010). However, turnover numbers and catalytic efficiencies of the S. paucimobilis SPT were found to be at least one order of magnitude higher than those for the S. wittichii SPT (Raman et al., 2010). Based on these results, the authors suggested that a specialized acyl-AcpR might be a better substrate for S. wittichii SPT than acyl-CoA and that this might not be the case for the well-characterized S. paucimobilis SPT. Our results (Figure 8) show that even the acyl-AcpR from C. crescentus is a better substrate for SPT from S. wittichii than acyl-CoA. However, we cannot confirm the assumption that S. paucimobilis SPT is not accompanied by a cognate AcpR (Raman et al., 2010) encoded in its genome. Our data base analysis of available S. wittichii and S. paucimobilis genomes (Supplementary Table 2) suggests that essentially all of them possess structural genes for AcpR and AasR and that therefore an AasR/AcpR-dependent SPT-catalyzed reaction occurs in the Sphingomonas genus.

In this work, we demonstrated that the specialized AcpR for sphingolipid biosynthesis from C. crescentus was the preferred thioester substrate (acyl-AcpR) for the SPTs from C. crescentus CB15, E. coli BL21(DE3), and S. wittichii to form the sphingolipid precursor, 3-oxo-sphinganine (Figure 7). SPTs from these bacteria were also able to use acyl-CoA or the housekeeping acyl-AcpP from C. crescentus as thioester substrate, however, to a much lesser extent. Therefore, although all three SPTs from Rhodobacteria showed a relaxed thioester substrate specificity, they clearly preferred the specialized ACP AcpR for initiating sphingolipid biosynthesis.

In eukaryotes the first committed step of sphingolipid biosynthesis is catalyzed by SPT and consists in the decarboxylative condensation of serine with an acyl-CoA, resulting in the formation of 3-oxo-sphinganine. Although palmitoyl-CoA can serve as a substrate for SPT from C. crescentus (Stankeviciute et al., 2022), we have now demonstrated for C. crescentus that a fatty acyl residue is linked to the specialized ACP AcpR by the specialized acyl-ACP synthetase AasR, and that subsequently acyl-AcpR is used as the preferred acyl-thioester substrate in the SPT-catalyzed reaction (Figure 1). Also, our in vivo experiments indicate that AcpR/AasR pairs from the same organism together with SPT are required for efficient 3-oxo-sphinganine formation (Figure 7 and Supplementary Figure 8). Therefore, the biological significance of the AcpR/AasR proteins might consist of sequestering acyl pools and specifically directing them toward sphingolipid synthesis. This intricate formation of 3-oxo-sphinganine seems to be common in members of the Rhodobacteria as their structural gene for SPT is preceded by a structural gene for the specialized AcpR, both forming an operon (Figure 9 and Supplementary Table 2). Also, a structural gene for the specialized acyl-ACP synthetase (aasR) is found in genomes of other sphingolipid-producing Rhodobacteria (Figure 9 and Supplementary Table 2). Whereas in most Rhodobacteria aasR is located in the close neighborhood of the acpR-spt operon, in the Sphingomonadaceae (Sphingomonas, Zymomonas) aasR is encountered in distant parts of the genome (Figure 9 and Supplementary Table 2). It is thought that also in bacteria keto-sphinganine reductase would covert oxo-sphinganine to sphinganine which would subsequently be acylated by an acyl-CoA N-acyltransferase during dihydroceramide biosynthesis (Olea-Ozuna et al., 2021). Recently, an alternative pathway for ceramide biosynthesis in C. crescentus was proposed in which N-acylation would precede reduction (Stankeviciute et al., 2022). Good homologs are encountered for the presumptive keto-sphinganine reductase (epi) and acyl-CoA N-acyltransferase (aca) (Olea-Ozuna et al., 2021) as well in other SPT-containing Rhodobacteria (Figure 9 and Supplementary Table 2). These five structural genes for dihydroceramide synthesis in Rhodobacteria are clustered together only in the genomes of C. crescentus and E. coli BL21 (Figure 9 and Supplementary Table 2).

All bacteria possess an essential housekeeping ACP, which is expressed constitutively from the acpP gene (Rawlings and Cronan, 1992). Phylogenetic analysis of amino acid sequences from housekeeping (AcpP) and specialized acyl carrier proteins showed that they cluster separately (Figure 10). Housekeeping AcpPs from α- (C. crescentus, S. wittichii, S. paucimobilis, Z. mobilis, G. oxydans, and S. meliloti), β- (N. eutropha), γ- [E. coli BL21(DE3)], and δ- (S. cellulosum, M. xanthus, and S. aurantiaca) Proteobacteria group together. Also, specialized ACPs for sphingolipid biosynthesis (AcpRs) cluster together but clearly form a distinct subgroup of ACPs.

Notably, besides the housekeeping AcpP, distant AcpR homologs are also encoded in genomes of the δ-proteobacteria Sorangium cellulosum, Myxococcus xanthus, and Stigmatella aurantiaca. Other specialized ACPs with other functions, i.e., NodF or AcpXL from S. meliloti, group separately from both the AcpP and AcpR clusters. It is remarkable that the housekeeping AcpPs cluster more closely together than the specialized AcpRs (Figure 10). The tertiary structure of ACPs seems to be conserved consisting of a 3- or 4-helix bundle comprised of three major α-helices (helices I, II, and IV) and a short α-helical segment (helix III) connecting helices II and IV (Byers and Gong, 2007). For interaction with most enzymes, helix II of AcpP seems to play a dominant role and has been termed “recognition helix” (Zhang et al., 2003). It is noteworthy that most of the acidic amino acyl residues of the C-terminal part of helix II, which are essential for the interaction of AcpP with enzymes, are conserved in AcpRs as well (Supplementary Figure 9).

Phylogenetic analysis of selected fatty acyl AMP ligases clarifies that at least three subgroup clusters exist (Figure 11) consisting of FadD-like acyl-CoA synthetases, acyl-AcpP synthetases employing the housekeeping AcpP as thiol substrate, and AasR-like acyl-AcpR synthetases required for sphingolipid biosynthesis in Rhodobacteria. Although there is experimental evidence for acyl-CoA synthetase activity for the annotated ORFs in S. meliloti (Soto et al., 2002) and E. coli, FadD candidates for G. oxydans and N. eutropha are assigned based on similarity to the E. coli and S. meliloti FadD (Supplementary Table 3). C. crescentus can grow on oleate as sole carbon source, possesses acyl-CoA synthetase activity and degrades oleate by β-oxidation (O’Connell et al., 1986). P-BLAST searches with FadD from E. coli or S. meliloti identify 11 candidates for acyl-CoA synthetases/fatty-acid CoA ligases in the C. crescentus genome and the two best candidates for acyl-CoA synthetase, CC_0966 and CC_1321 (Supplementary Table 3), are included in Figure 11. However, if free fatty acids are activated to their CoA derivatives, one might expect that they are catabolized by β-oxidation. In contrast, if free fatty acids are converted to acyl-AcpR derivatives, also at the energetic cost of 2 ATP molecules (Figure 1), they are redirected to sphingolipid biosynthesis.

Although the biological significance of acyl-ACP synthetases is not always clear yet, acyl-ACP synthetase of E. coli together with 1-acylglycerol phosphate acyltransferase PlsC is a key player in reacylating lysophospholipids, for example in the lyso-phosphatidylethanolamine cycle (Sahonero-Canavesi et al., 2019). Similarly, a major function of acyl-ACP synthetases in the cyanobacteria Synechocystis and Synechococcus seems to consist in the recycling of endogenous fatty acids (Kaczmarzyk and Fulda, 2010). In contrast, the intracellular parasite Chlamydia trachomatis (Yao et al., 2015), pathogenic Neisseria (Yao et al., 2016), or the Bacteroidete Alistipes (Radka et al., 2020) from the human gut microbiome use their acyl-ACP synthetases to scavenge host fatty acids in order to incorporate them into their membrane lipids. Acyl-ACP synthetases acylating the housekeeping AcpP form at least two subgroups, one containing Aas from E. coli and another containing Aas from V. harveyi, which are distinct from acyl-CoA synthetases. In addition, the acyl-ACP synthetase of Neisseria clusters together with acyl-CoA synthetases, suggesting that acyl-ACP synthetases acylating housekeeping AcpPs might have evolved on several occasions from acyl-CoA synthetases.

Members of a third cluster involving possible specialized Aas for sphingolipid biosynthesis in Rhodobacteria (AasR), i.e., from the α- (C. crescentus, S. wittichii, S. paucimobilis, G. oxydans, and Z. mobilis), β- (N. eutropha), and γ- [E. coli BL21(DE3)] Proteobacteria group together, although they seem to form two subgroups (Figure 11). In one subgroup, AasRs from S. wittichii, S. paucimobilis, and Z. mobilis cluster closely together. In the other subgroup, AasRs from C. crescentus, G. oxydans, N. eutropha, and E. coli BL21(DE3) are more distant. Specialized AcpRs from these same bacteria cluster in a similar fashion, forming these same two subgroups (Figure 10). The observed clustering profiles of AcpRs and AasRs are complementary and might indicate a high specificity of these acyl-AcpR synthetases to acylate the specialized AcpRs for sphingolipid biosynthesis in these Rhodobacteria. Our own studies (Figure 7) confirm that AcpRs are acylated exclusively or at least preferentially by the AasR from the same organism. Phylogenetic trees of the Proteobacteria suggest that early during their diversification, the δ-Proteobacteria separated from the others (Madigan et al., 2018). In this context it is remarkable that in addition to homologs of FadD_Ec and Aas_Vh, the δ-proteobacteria M. xanthus, S. aurantiaca, and S. cellulosum possess good and distinct homologs of AasR_Cc (Supplementary Table 4) encoded in their enormous genomes and that these homologs branch off at the base of the AasR-like cluster (Figure 11). Also, M. xanthus, S. aurantiaca, and S. cellulosum have genes coding for AcpR homologs in addition to the structural genes for housekeeping AcpPs (Supplementary Table 4). Notably, acpR and aasR homologs form an operon in the δ-proteobacteria M. xanthus, S. aurantiaca, and S. cellulosum but are situated in different genomic contexts as they are in the Rhodobacteria.

The bacterial strains and plasmids used in this work and their relevant characteristics are listed in Table 1. E. coli strains were grown at 30°C either in Luria-Bertani (LB) broth (Sambrook et al., 2001), or in M9 minimal medium (Miller, 1972) for β-[³H]alanine labeling assays. C. crescentus CB15 strain was cultivated at 30°C in PYE (peptone-yeast extract) complex medium (0.2% peptone, 0.1% yeast extract, 1 mM MgSO₄ and 0.5 mM CaCl₂) (Hottes et al., 2004). S. wittichii RW1 strain was grown at 30°C in nutrient medium (0.5% peptone and 0.3% meat extract) (Halden et al., 2005). Antibiotics were added to media, when required, at the following final concentrations (μg/ml): carbenicillin 100, chloramphenicol 20, kanamycin 50, and spectinomycin 100, for E. coli strains.

Recombinant DNA techniques were carried out using standard procedures (Sambrook et al., 2001). Commercial sequencing of amplified genes by Eurofins Medigenomix (Martinsried, Germany) corroborated correct DNA sequences. DNA regions containing presumptive sphingolipid biosynthesis genes and derived protein sequences were analyzed using the National Center for Biotechnology Information (NCBI) BLAST network center (Altschul et al., 1997).

Genes of interest were amplified by PCR from genomic DNA of C. crescentus CB15, E. coli BL21(DE3), or S. wittichii RW1, using specific oligonucleotides (Supplementary Table 5) and introducing suitable restriction sites for cloning. After restriction with the respective enzymes, PCR-amplified DNA fragments were cloned into a pET9a, pET16b, pET17b, pET28a, pBAD24, or a pCDFDuet-1 vector as detailed in Supplementary Table 6.

SPT-encoding plasmids pJPG05, pJPG06 and pJPG07 (Supplementary Table 6) were each digested with NdeI and BamHI, and the spt-encoding fragments were recloned into pET9a, which had been digested with NdeI and BamHI, to yield the pET9a derivatives pJPG08, pJPG01 and pJPG02, respectively. Recloning of the CC_1165 (aasR)-encoding NdeI/BamHI fragment from pDG04 (Supplementary Table 6) into a NdeI/BamHI-restricted pET16b vector yielded plasmid pDG10. The gene encoding acyl-ACP synthetase (aasS) from V. harveyi (LA59_RS23465) was synthesized by DNA2.0 from ATUM² and subsequently recloned as NdeI/BamHI fragment into NdeI/BamHI-digested pET16b, to yield plasmid pJPG10. Recloning of the CC_1165 (aasR)-encoding NcoI/BamHI fragment from pDG10 (Table 1) into a NcoI/BamHI-restricted pCDFDuet-1 (MCS-1) vector yielded plasmid pJPG13. Recloning of the CC_1163 (acpR)-encoding NdeI/BamHI fragment from pDG01 into a NdeI/BglII-restricted pJPG13 (MCS-2) vector yielded plasmid pJPG14. Recloning of the CC_1163/CC_1162 (acpR/spt)-encoding NdeI/BamHI fragment from pRJ03 (Olea-Ozuna et al., 2021) into a NdeI/BglII-restricted pCDFDuet-1 (MCS-2) vector yielded plasmid pJPG16. Recloning of the smc02162 (fadD)-encoding NdeI/HindIII fragment from pECH6 (Pech-Canul et al., 2020) into a NdeI/HindIII-restricted pET28a vector yielded plasmid pECH3.

Proteins were analyzed employing different PAGE systems (native PAGE, SDS-PAGE, urea-PAGE, or SDS-Tricine-PAGE) according to necessity. In vitro conversions from apo- to holo-ACPs, using the holo-ACP synthase (AcpS) from S. meliloti 1021 were analyzed by 20% native PAGE (Jackowski and Rock, 1983). In vivo β-[³H]alanine incorporation into the 4'-PPT prosthetic group of ACPs was analyzed by 20% native PAGE as well as by SDS-Tricine-PAGE (Schägger and von Jagow, 1987) containing 12% polyacrylamide. This latter system is frequently used for separating peptides and small proteins such as ACPs. Conformation-sensitive urea-PAGE (Post-Beittenmiller et al., 1991), containing 18.7% polyacrylamide, was employed to analyze the in vitro acylation of ACPs with fatty acids of different length, using AasR (CC_1165) from C. crescentus, and Aas from V. harveyi. Different concentrations of urea (up to 5 M) in both the separation and stacking gel were used to separate acyl-ACPs. Furthermore, overexpression and purification of His-tagged proteins was analyzed by denaturing SDS-PAGE (Schägger et al., 1988) containing 12% polyacrylamide.

In vivo labeling of ACPs with β-[³H]alanine, the biosynthetic precursor of 4'-phosphopantetheine, was carried out essentially as described previously using the β-alanine auxotrophic strain OG7001 (Epple et al., 1998; López-Lara and Geiger, 2000).

Strains of E. coli BL21(DE3) × pLysS harboring plasmid pDG01 (with specialized acpR CC_1163), pPEG01 (with housekeeping acpP CC_1677), pJPG12 (with specialized acpR ECD_02853), or pTB5079 (with housekeeping acpP ECD_01090 of E. coli) were grown at 30°C in 1 L of LB medium and at an optical density at 600 nm (OD₆₀₀) of 0.4, isopropyl-β-D-thiogalactoside (IPTG) was added to a final concentration of 0.1 mM to induce protein expression. After induction for 4 h, cells were collected by centrifugation at 7,500 × g for 30 min at 4°C, cell pellets were resuspended in 50 ml of buffer A (50 mM Tris/HCl, 0.1 M KCl, pH 6.8) and stored at −20°C. After thawing, the cells were broken by passing cell suspensions twice through a French pressure cell at 20,000 pounds per square inch (psi), the remaining unbroken cells and debris were removed by centrifugation at 7,000 × g for 15 min at 4°C and supernatants were used as cell-free extracts (CE).

For the purification of all ACPs, the respective CE was slowly stirred at 4°C and cold 2-propanol was added dropwise to a given CE up to a final concentration of 50% (v/v). After incubation for 60 min at 4°C, the precipitate was removed by centrifugation at 7,000 × g for 15 min at 4°C. The 2-propanol-containing supernatant was dialyzed overnight against 2 liter of buffer A. The dialysate was subjected to ion-exchange chromatography using DEAE-52 cellulose (Whatman). Each of the samples was applied to a 30 ml column, that had been equilibrated with two column volumes of buffer B (10 mM Bis-Tris/HCl, pH 6.0, 1 mM 3-cholamidopropyl dimethylammonio 1-propanesulfonate). Each column was washed with two column volumes of buffer B, and proteins were eluted with a linear gradient of NaCl (0.1–1 M) in buffer B in a total volume of 150 ml. Fractions (3 ml) were collected, and aliquots were analyzed in 20% native PAGE (Supplementary Figure 2). ACPs AcpP_Cc (CC_1677), AcpP_Ec (ECD_01090), and AcpR_Ec (ECD_02853) eluted in a range from 0.28 to 0.35 M NaCl, while the ACP AcpR_Cc (CC_1163) eluted in a range from 0.21 to 0.28 M NaCl. From each chromatographic separation, five ACP-containing fractions were combined and concentrated using Amicon Ultra-15 Centrifugal Filter Devices (Merck Millipore). Washing steps with 100 mM Tris/HCl, pH 8.0 in the centrifugal device ensured that the final concentration of NaCl was less than 20 mM. The respective ACPs were stored in a final volume of 1.5 ml at −20°C.

To convert purified apo-ACPs to functional holo-ACPs, AcpS_Sm obtained from E. coli containing plasmid pSBA01 was used as a tool for in in vitro reactions employing CoA as phosphopantetheine donor as previously described (Ramos-Vega et al., 2009).

Strains of E. coli BL21(DE3) × pLysS harboring plasmid pDG10 (with specialized aasR CC_1165 of Caulobacter), pJPG10 (with aas of V. harveyi), or pECH3 (with fadD of Sinorhizobium) were grown at 30°C in 25 ml of LB medium and at OD₆₀₀ of 0.4, IPTG was added to a final concentration of 0.1 mM. After induction for 4 h, cells were collected by centrifugation at 7,500 × g for 30 min at 4°C and cell pellets were resuspended in 2 ml of binding buffer (20 mM sodium phosphate, 0.5 M NaCl, 20 mM imidazole, pH 7.4). After passing cell suspensions twice through a French pressure cell at 20,000 psi, the remaining unbroken cells and debris were removed by centrifugation at 7,000 × g for 15 min at 4°C and supernatants were used as CEs.

The His-tagged proteins were purified using the HiTrap IMAC HP systems (GE Life Sciences) with column bed volumes of 1 ml, respectively. CEs were diluted 1:2 with binding buffer and applied to a chromatography column previously equilibrated with 10 ml of binding buffer. Then columns were washed with 10 ml of binding buffer. Stepwise elution was performed with portions of 5 ml of binding buffer containing increased concentrations of imidazole (100, 250, 500, 750 mM). Fractions were collected and analyzed in 12% SDS-PAGE (Supplementary Figure 4). Fractions containing the respective His-tagged protein (His₁₀-tagged putative AasR CC_1165, His₁₀-tagged Aas from V. harveyi (Aas_Vh) and His₆-tagged FadD from S. meliloti) were near homogeneous and were dialyzed against Aas buffer (20 mM Tris-HCl, pH 7.5, 10% glycerol, 1 mM EDTA, 0.1 mM DTT, and 0.002% Triton X-100), similarly, as described by Jiang et al. (2006). The protein content of homogeneous acyl-CoA and acyl-ACP synthetase preparations was determined using the method reported by Dulley and Grieve (1975).

The enzymatic assays for acyl-CoA and acyl-ACP synthetases described here are essentially modified versions of the assay reported by Jiang et al. (2006) for acylating ACPs with acyl-ACP synthetase from V. harveyi.

For each reaction, 10 nmol [9,10-³H]-palmitic acid (specific radioactivity: 743 mCi/mmol) in ethanolic solution and an equimolar amount of NaOH were combined with an aqueous solution containing 35 μg Triton X-100 in an Eppendorf tube and solvents were removed under vacuum. Aqueous Tris/HCl buffer, pH 8.0 was added, the tubes were vortexed three times for 30 s and sonicated for 30 min. In a total volume of 50 μl, the final assay contained 200 μM Na [³H]-palmitate (7.4 μCi), 1 mM Triton X-100, 100 mM Tris/HCl buffer, pH 8.0, 10 mM ATP, 10 mM MgSO₄, 5 mM dithiothreitol, 40 μM thiol substrate (ACPs or CoA) and different concentrations of each enzyme (30 nM for His₁₀-AasR_Cc (CC_1165), 100 nM for His₁₀-Aas_Vh, and 300 nM for His₆-FadD_Sm). The reactions were incubated for 0, 15, 30, 45 or 60 min at 37°C in a water bath. Holo-ACPs were quantified by detecting their single sulfhydryl group of 4'-phosphopantetheine using Ellman’s method (Ellman, 1959).

When assaying for acyl-CoA synthetase activities, the reactions were stopped by adding 125 μl of isopropanol/n-heptane/1 M H₂SO₄ (40:10:1; v/v). Addition of 25 μl of distilled water and 125 μl n-heptane and thorough mixing led to the separation of the organic and aqueous phases. Subsequently, the organic phase (containing unbound [³H]palmitate) was removed and the aqueous phase (containing [³H]palmitoyl-CoA products) was washed 6 more times with 125 μl of n-heptane. Radioactivity associated with the aqueous phase ([³H]palmitoyl-CoA) was quantified in a scintillation counter.

When assaying for acyl-ACP synthetase activities, the reactions were stopped by freezing them rapidly at −80°C. Subsequently, defrosted samples were applied to Whatman 3MM cellulose filters, washed three times with methanol/chloroform/acetic acid (2:1:0.3; v/v) in a vacuum manifold. Filters were dried and filter-bound radioactivity was quantified in a scintillation counter, similarly, as described by Ray and Cronan (1976).

In order to study the effect of the AcpR substrate concentration on AasR activity, initial velocities for holo-AcpR_Cc (CC_1163) acylation were determined at different AcpR_Cc concentrations (2.5, 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, and 100 μM) employing His₁₀-AasR_Cc (CC_1165) and [³H]palmitate. Samples were incubated at 37°C in a water bath for 30 min, period for which product formation was linear for all AcpR concentrations employed. Subsequently, the initial velocities (V₀) were calculated measuring the slope (m) of the incorporation of [³H]palmitate into the product [³H]palmitoyl-AcpR_Cc and plotted against the concentration of the holo-AcpR (CC_1163) substrate (Figure 6).

Strains of E. coli BL21(DE3) × pLysS harboring plasmid pJPG08 (with spt CC_1162 of C. crescentus), pJPG01 (with spt ECD_02854 of E. coli BL21(DE3)), pJPG02 (with spt Swit_3900 of S. wittichii), or the empty pET9a vector were grown at 30°C in 25 ml of LB medium and at OD₆₀₀ of 0.4, IPTG was added to a final concentration of 0.1 mM. After induction for 4 h, cells were collected by centrifugation at 7,500 × g for 30 min at 4°C and cell pellets were resuspended in 2 ml of buffer containing 20 mM potassium phosphate (pH 7.4), 150 mM NaCl, 5 mM DTT (Raman et al., 2009). After passing cell suspensions twice through a French pressure cell at 20,000 psi, the remaining unbroken cells and debris were removed by centrifugation at 7,000 × g for 15 min at 4°C and supernatants were used as CEs.

Enzymatic assays for serine palmitoyltransferase were performed similarly, to those described by Ikushiro et al. (2001). Each assay contained in a final volume of 100 μl, 100 mM HEPES/NaOH (pH 7.5), 100 μM EDTA, 5 mM DTT, 10 μM pyridoxal phosphate, and 432 μM (0.2 μCi) of L-[¹⁴C(U)]serine (PerkinElmer; specific radioactivity: 46.4 mCi/mmol). Thioester substrates (palmitoyl-CoA, palmitoyl-AcpR_Cc or palmitoyl-AcpP_Cc were added to a final concentration of 10 μM. For each 100 μl assay, 10 μg protein of CEs from E. coli BL21(DE3) × pLysS carrying the empty pET9a vector, or SPT-expressing plasmids were employed respectively. Enzyme assays were incubated at 30°C for 4 h in water bath and the reactions were stopped by adding 100 μl of NH₄OH. For each sample, separation of the organic and aqueous phases was carried out by adding 375 μl of chloroform/methanol (2:1; v/v) (Bligh and Dyer, 1959). After centrifugation at 16,000 × g for 5 min, the organic phases were transferred to fresh tubes and dried (Raman et al., 2009). The dried lipids were dissolved in 20 μl of chloroform/methanol (2:1; v/v) and spotted onto a thin layer chromatography (TLC) plate (Silica Gel 60, Merck). Separation in one dimension (1D) was carried out with a mobile phase of chloroform/methanol/2N ammonium hydroxide (40:10:1; v/v). The developed chromatogram was exposed to Phosphor Screens (GE Healthcare), and the autoradiograms were visualized with an optical scanner (Typhoon FLA 9500 from GE Healthcare Life Sciences).

For all figures obtained using TLC (Figures 7, 8 and Supplementary Figure 8), relative mobilities of non-radioactive standard samples and radioactive experimental samples were compared after they had been separated on the same TLC plate. However, as the radioactive and non-radioactive samples needed to be detected by different methods, the aluminum-backed, developed silica plates were cut with scissors in order to visualize radioactive samples by autoradiography and non-radioactive samples by iodine staining, as indicated in the corresponding figure legends. Based on the reconstituted silica plates, the two image subparts can be clearly assembled. The subparts of the recomposed figures are surrounded by black squares.

Acylation reactions for mass spectrometric analyses were as previously described in “Enzymatic assays for acyl-CoA and acyl-ACP synthetases.” However, the acylations were performed in the absence of Triton X-100 and with 300 nM of each enzyme (AasR_Cc, or Aas_Vh).

ACP preparations were precipitated with acetone at −20°C and washed twice with cold acetone. After drying, ACP-containing samples were dissolved in water and of each sample 2.5 μg of protein were injected into a LCQ fleet ion trap mass spectrometer (Thermo Fisher Scientific, Waltham, MA, United States). Spray voltage was set to 5 kV, sheath gas flow rate at 20 units and the capillary temperature at 250°C.

Before injection into the mass spectrometer, acyl- (palmitoyl-) AcpP_Cc was subjected to chromatographic separation in an Ultimex 3000 HPLC system (Dionex, Sunnyvale, CA, United States) equipped with a Hypersil Gold C18 (100 × 2.1 mm) analytical column (Thermo Fisher Scientific).

Spectra visualization was performed using Xcalibur software (Thermo Fisher Scientific) and MZmine 2.38 (Pluskal et al., 2010). Multiple charge states were observed for each of the major species of each protein. The mass of each protein was calculated using ESIprot (Winkler, 2010).