Serum-stable and selective backbone-N-methylated cyclic peptides that inhibit prokaryotic glycolytic mutases

Abstract

N-methylated amino acids are privileged residues of naturally occurring peptides critical to bioactivity. However, de novo discovery from ribosome display is limited by poor incorporation of N-methylated amino acids (N-MeAA) into the nascent peptide chain attributed to a poor EF-Tu affinity for the N-methyl-aminoacyl-tRNA. By reconfiguring the tRNA’s T-stem region to compensate and tune the EF-Tu affinity we conducted RaPID (Random nonstandard Peptides Integrated Discovery) display of a macrocyclic peptide (MCP) library containing six different N-MeAAs. We have here devised a “pool-and-split” enrichment strategy using the RaPID display and identified N-methylated MCPs against three species of prokaryotic metal ion-dependent phosphoglycerate mutases. The enriched MCPs reached 57% N-methylation with up to three consecutively incorporated N-MeAAs, rivalling natural products. Potent nanomolar inhibitors ranging in ortholog-selectivity, strongly mediated by N-methylation, were identified. Co-crystal structures reveal an architecturally related Ce-2 ipglycermide active site metal ion-coordinating Cys lariat MCP, functionally dependent on two cis N-MeAAs with broadened iPGM species selectivity over the original nematode-selective MCPs. Furthermore, isolation of a novel metal ion-independent S. aureus iPGM inhibitor utilizing a phosphoglycerate mimetic mechanism illustrates the diversity of possible chemotypes encoded by the N-MeAA MCP library.

Graphic Abstract

Introduction

N-methylation of peptide backbones can drastically improve a bioactive peptide by improving its weakest characteristics, e.g., metabolic stability and membrane-permeability^1–5. While many powerful de novo discovery tools for peptide ligands have been developed⁶, few de novo discovered N-methylated ligands have been reported. This is caused by the poor ribosomal incorporation of N-methylated amino acids (N-MeAA) into the nascent peptide chain⁷.

Though low in efficiency, N-MeAA incorporation is possible, for example, ubiquitin ligase inhibiting macrocyclic peptides have been previously discovered by utilizing a randomized library containing four N-MeAA and 11 proteinogenic amino acids (pAA)⁸. These peptides contained N-MeAAs upwards to 36% (4/11 residues in the random region), as well as two consecutively incorporated N-MeAAs with dissociation constants (K_D) in the sub-nanomolar to nanomolar range. This is significant, as subsequent N-methylation of previously discovered peptide ligands often lowers or abolishes the target affinity of the peptide^9–11. By increasing the number of N-MeAAs in the translation system, we would improve the probability and extent of peptide N-methylation. However, the poor ribosomal incorporation efficiency of N-MeAAs can cause the library to become untranslatable if too many are included.

The cause of suboptimal ribosomal incorporation of N-MeAAs is the weakened affinity of N-MeAA-tRNA to elongation factor thermal unstable (EF-Tu) compared to natural pAA-tRNAs due to the steric clashes caused by the N-methyl residue with the tight binding pocket in which the pAA amine conventionally resides^{12, 13}. The affinity between pAA-tRNA and EF-Tu is mediated by the acylated residue as well as the tRNA T-stem domain. In pAA-tRNAs, the T-stem domain compensates the pAA-mediated EF-Tu affinity, resulting in near equal affinity for all pAA-tRNAs¹⁴. By preparing a series of tRNAs with increasingly strong binding T-stem domains and pairing them to N-MeAAs, a similar affinity balance could be achieved, significantly improving the ribosomal incorporation of N-MeAAs and allowing for the in vitro translation of a cyclic peptide containing nine different N-MeAAs¹⁵. Here, we applied this methodology to generate a randomized cyclic peptide library containing six N-MeAAs and 16 pAAs (Figure 1a). Compared to our previous library⁸, this library increases N-MeAAs from four to six and total available AAs from 15 to 22.

Figure 1. tRNA tuning approach of N-MeAA incorporation using “pool-and-split” RaPID selection paradigm.

(a) N-methyl amide amino acid codons. (b) RaPID selection where rounds 1–3 utilize the pooled procaryotic iPGMs, whereas rounds 4 onward continue the selection process using individual iPGMs and the enriched library. (i) Puromycin ligation, in vitro translation, and reverse transcription. (ii) Negative selection of bead binders. (iii) Affinity selection with pooled or individual iPGMs. (iv) Transcription of enriched DNA sequences.

We used target-based affinity selection system, referred to as the Random non-standard Peptide Integrated Discovery (RaPID) system, to interrogate and enrich members of the N-methylated cyclic peptide library through mRNA display^16,17 against three prokaryotic orthologs of the co-factor independent phosphoglycerate mutase (iPGM) enzymes derived from Staphylococcus aureus (S. aureus), Helicobacter pylori (H. pylori), and Mycoplasma orale (M. orale). iPGM catalyzes the interconversion between 2- and 3-phosphoglycerate in a spring-loaded open-closed-open catalytic cycle and is an essential factor in both glycolysis and gluconeogenesis pathways^{18, 19}.

Humans and other higher mammals use cofactor-dependent phosphoglycerate mutase (dPGM) a homodimer comprised of 23 kDa subunits that requires 2,3-biphosphoglycerate as a cofactor. Whereas iPGM is a monomeric 56 kDa bi-domain enzyme having a divalent metal ion dependence²⁰. Therefore, these structural and catalytic differences between the dPGM and iPGM isozymes may permit selective targeting of an iPGM-dependent pathogen’s critical glycolytic pathway without affecting host dPGM-dependent glucose metabolism. However, iPGM appears to be poorly druggable, which has been attributed to the structural dynamics of its catalytic mechanism²¹.

In a previous RaPID screen on iPGM from Caenorhabditis elegans (C. elegans) a new class of high-affinity peptide was uncovered that we have termed Ce-2 ipglycermide, an 8-amino acid macrocycle with an extra-cyclic 6-amino acid sequence terminating in Cys14 (Table 1), which is cradled at the interdomain crevice of the open enzyme conformation^{22, 23}. The Ce-2 ipglycermide chemotype broadly inhibits iPGMs from parasitic nematodes but exhibits no significant activity on other iPGMs (Table 1). However, many clinically relevant human pathogens utilize iPGM as their glycolytic phosphoglycerate mutase^24–26. Given the significant sequence similarity among iPGM orthologs and the demonstrated high affinity of ipglycermides across nematode iPGMs, we set out to examine the potential for cross-species selectivity among prokaryotic iPGMs. As such we can expand the infectious disease indications for iPGM inhibitors, and draw on a “one target, many diseases” concept.

Table 1.

Sequence and activity of prokaryotic iPGM-derived N-methyl amide macrocyclic

MCP ID	MCP Sequence	pAC50 apparent ±sd
		Nematode						Bacterium
		C. elegans		B. malayi		W. bancrofti		M. orale		M. pneumoniae		E. coli		s. aureus		H. pylori
Ce-2	AcyDYPGDYCYLYGTC	8.61	0.03	8.57	0.06	8.63	0 09	na	-	na	-	6.50	0.10	na	-	5.44	002
Ce-2act																6 78	0 15
Ce-2d	AcyDYPGDYCYLY	8.54	0.05	7.35	0 08	7.33	0 12	na	-	na	-	na	-	na	-	na	-
Hp-L2	AcYMeY(OMe)RMeFDMeSMeYQSMeSMeY(OMe)VMeYMeNlePC	5.19	0.13	5.21	0.25	5.25	0.53	5.73	005	5.84	nd	5.51	0.03	na	-	5 22	0 10
Hp-D3	AcyDSYAMeY(OMe)MeYMeFNGMeYNHYWC	7.81	0.07	7.61	0.06	7.09	0.03	5.69	008	5.92	0.03	6.93	0.10	na	-	6.01	006
Hp-D3 ∆Me	AcyDSYA_Y(OMe)_Y_FNG_YNHYWC	na	-	na	-	na	-	na	-	<5	-	na		na	-	na	-
Hp-D4	AcyYWQMeYMeYEYGYMeYKC	4.87	0.04	5.44	0.06	6.33	0.06	5.92	004	4.85	020	5.20	0.18	na	-	5.84	0.10
Mo-L1	AcYMeYMeFYDMeFMeYGNMeYSYRVPC	6.03	0.03	6.09	0.03	6.51	0.06	8.03	0.18	7.80	007	5.36	0.10	na	-	5.23	0.17
Mo-L1 ∆Me	AcY_Y_FYD_F_YGN_YSYRVPC	na	-	na	-	4.78	nd	4.75	0.14	4.75	nd	na	-	na	-	na	-
Mo-L2	AcYPDTNWWYMeYDEWHC	na	-	na		5.84	026	8.61	0.20	8.73	0.14	na	-	na	-	na	-
Mo-L2 ∆Me	AcYPDTNWWY_YDEWHC	na	-	na	-	4.48	nd	5.48	0.27	5.42	0.15	na		na	-	na	-
Mo-D1	AcySGDWWYMeYDEWMeSVAMeFMeGC	na	-	4.88	nd	na	-	8.21	028	7.61	007	na	-	na	-	na	-
Mo-D1 ∆Me	AcySGDWWY_YDEW_SVA_F_GC	na	-	na	-	4.25	nd	4.93	0.12	4.59	021	na	-	na	-	na	-
Mo-D2	AcyWMeYDMeSMeYMeY(OMe)YPGTAQMeYTC	4.46	0.56	4.46	0.15	na	-	7.99	0.20	7.17	0.09	na	-	na	-	na	-
Mo-D2 ∆Me	AcyW_YD_S_Y_Y(OMe)YPGTAQ_YTC	na	-	na	-	4.57	nd	4.72	0.04	4.65	0.05	na	-	na	-	na	-
Sa-L2	AcYMeY(OMe)NDSGWMeYNTMeFC	5.56	004	4.08	0.01	4.34	nd	4.34	003	4.37	0.08	5.16	0.12	6.76	0.06	na	-
Sa-D2	AcyRYEMeYYKMeYECPKCC	8.35	0.10	8.74	0.05	8.57	0.13	8.55	0.19	8.65	0.30	7.21	0.10	na	-	6.40	021
Sa-D2 ∆Me	AcyRYE_YYK_YECPKC	na	-	na	-	4.69	nd	4.61	0.12	4.71	0.08	na	-	na	-	na	-
Sa-D2 C13S	AcyRYEMeYYKMeYECPKS	4.96	0.09	4.83	0.03	5.45	0.11	5.26	0.56	5.08	0.06	na	-	na	-	na	-
Sa-D2 ∆PKC	AcyRYEMeYYKMeYEC	na	-	na	-	5.26	0.61	<5	nd	5.29	0.44	na		na	-	na	-
Sa-D2 LgMC	AcyRYEMeYYKMeYECPKC	5.40	0.08	5.77	0.14	5.77	0.13	5.97	0.08	5.85	0.21	4.73	0.07	na	-	4.48	0.08
Sa-D2 LgMC ∆Me	AcyRYE_YYK_YECPKC	na	-	na	-	na	-	na	-	na	-	na	-	na	-	na	-
Sa-D3	AcyQVTVWWAMeSPWEDC	na	-	na		4.92	0.57	8.58	0.17	8.32	0.15	na		7.42	0.09	na	-
Sa-D3 ∆Me	AcyQVTVWWA_SPWEDC	na	-	na	-	na	-	8.10	0.15	6.79	042	na	-	7.40	0.04	na	-

Notes: pAC₅₀ in bold indicated potent nM activity and pAC₅₀ in italics indicates the activation of H.p. iPGM by Ce-2 from a biphasic response. Macrocycle is formed between Ac and cystine in blue, N-methyl amino acids are indicated in red. All MCPs were inactive on the protozoan, L. Mexicana. ‘na’ indicates no measurable inhibitory activity. The error, reported as s.d., represent determinations from N≥3 experiments. Error was not determined, nd, when partial curves fits were only obtained from one of the experimental replicates due to a low MCP inhibitory activity.

To investigate the diversity of ipglycermide chemotypes and expand iPGM selectivity with RaPID technology we employed a pooled target selection strategy, illustrated in Figure 1b, with S. aureus, H. pylori and M. orale iPGMs to identify cross-prokaryotic ortholog and ortholog specific ipglycermides. By incorporating a translation system containing six N-MeAAs, we aim to discover novel N-methyl amide containing iPGM macrocyclic peptide (MCP) inhibitors embodying natural product-like characteristics.

Results and discussion

Library construction

Two cyclic peptide libraries were constructed from a semi-randomized mRNA library following the translation codon table as shown in Figure 1a. In one of the libraries, the initiation AUG codon was reprogrammed to encode for ClAc-L-Tyr, whereas ClAc-D-Tyr was used in the other library. Following AUG, the libraries contain 7 to 15 randomized NNK codons followed by a fixed CGSGSGS linker and an amber stop codon. Ligation of the mRNA to a puromycin linker ensures the encoding mRNA is displayed on the peptide’s C-terminus. Following in vitro translation, the peptide spontaneously undergoes intramolecular cyclization through nucleophilic attack by a cysteine thiolate with the N-terminal ClAc group.

Pool-and-split RaPID selection

Unlike the conventional single target RaPID selection scheme, we opted to use a new multi-target “pool-and-split” selection approach where several orthologous targets are initially mixed to discover both ortholog-selective and non-selective prokaryotic iPGM inhibitors (Figure 1b). The bead immobilization efficiency of the three iPGM orthologs was individually determined to be 10 pm/μL bead slurry for S. aureus iPGM, 20 pm/μL for H. pylori iPGM and 30 pm/μL for M. orale iPGM (Supplementary Figure 1). Using these immobilization densities, the immobilized iPGMs were mixed as a 1:1:1 ratio. During the first three cycles of RaPID, the peptide libraries were selected with this iPGM mixture. On the fourth round, the pre-enriched libraries were split into three and individually screened against either S. aureus, H. pylori, or M. orale iPGM until enrichment of the library was observed (Supplementary Figure 2). The individual MCPs are designated by the final iPGM selection enzyme and initiator amino acid stereochemistry, D or L defining the library (see ID column in Table 1).

The efficacy of the new T-stem tuned translation system is demonstrated by the sequences of the enriched peptides. Poor incorporation of N-MeAAs would have caused highly methylated peptides to be eliminated in favor of more efficiently translated candidates bearing less methylation. Nevertheless, the peptides identified in the final selection were highly methylated with upwards of three consecutive N-MeAAs and 57.1% total methylation (8/14) in the random region (Table 1, peptide Hp-L2). Most peptides contain more N-MeAAs than the average 27.3% (6/22) of available residues, suggesting that the N-MeAAs are both well incorporated and generally beneficial to the binding affinity.

Functional characterization of ligands

The 10 MCPs were synthesized by solid phase peptide synthesis (SPPS; Supplementary Figure 3) and functionally characterized with a luminescent coupled-enzyme ATP-quantifying assay against a panel of iPGM orthologs²³. The panel contained representative iPGM species orthologs derived from the nematodes, C. elegans, Brugia malayi (B. malayi) and Wuchereria bancrofti (W. bancrofti), prokaryotic forms from Escherichia coli (E. coli.), Mycoplasma pneumoniae (M. pneumoniae), S. aureus, H. pylori, M. orale, and the phylogenetically more distant trypanosome, Leishmania mexicana (L. mexicana). Overall, the MCPs display a broad range of inhibitory potencies and selectivity. The MCP sequence and apparent IC₅₀ values derived from concentration-response curves (CRCs) on each iPGM ortholog are shown in Table 1 and Supplemental Figure 4a, respectively.

Several observations were made from the functional activity profiling, the most salient of which were potent nanomolar activity (see bold entries in Table 1) of M. orale and S. aureus-selected MCPs (e.g., Mo-L2, -D1, -D2 and Sa-D2, -D3) on M. orale iPGM, whereas on S. aureus iPGM these MCPs were inactive or >10-fold less potent. Notwithstanding, Sa-D3 is remarkably potent (45 nM) on S. aureus iPGM. The potency and selectivity distribution of MCPs varied depending on the iPGM ortholog used in the final selection (Figure 1b). For H. pylori iPGM, MCPs Hp-L2 and -D3 were obtained ranging between 7 μM−16 nM across prokaryotic and nematode iPGMs, showing moderate potency on H. pylori iPGM (1–6 μM). For the four M. orale iPGM selected MCPs studied, all but one (Mo-L1) displayed high M. orale iPGM selectivity and very high inhibitory potency, with Mo-L2, -D1 and -D2 having IC₅₀ values between ~2 and 11 nM. MCPs undergoing the final enrichment step with S. aureus iPGM resulted in an intermediate distribution of iPGM selectivity. While the most potent S. aureus iPGM MCPs observed in this study were obtained from the S. aureus iPGM enrichment (Sa-D3), this enrichment step yielded the highest potency panortholog iPGM MCP, As well as the only Ce-2 ipglycermide lariat-like MCP, Sa-D2, having an IC₅₀ range between 2–400 nM across 7 of the 9 iPGMs profiled was isolated in this selection.

Essential nature of methylation on binding activity and serum stability

To investigate the effect amide N-methylation has on the binding of the discovered peptides, amide (∆Me) analogs of eight MCPs were prepared and characterized with the same activity-based enzyme assays, the results of which are summarized in Table 1. In most of the assayed peptides, loss of N-methylation leads to a near complete loss of inhibitory activity (Figure 2a–c) and reduction in serum stability (Figure 2d–f).

Figure 2. Comparative functional and human serum stability characterization of select Sa-D and Mo-L/D synthetic cyclic peptides and ∆N-Me analogs.

(a–c) MCP CRCs on representative iPGM orthologs from bacteria, S. aureus (Sa), E. coli (Ec), H. pylori (Hp), M. orale (Mo) and M. pneumonia (Mp) and nematode, C. elegans (Ce), B. malayi (Bm) and W. bancrofti (Wb). Top plots represent N-Me containing MCPs and bottom plots correspond to ∆N-Me analogs. Data are representative of N≥3 independent experiments. (d–f). MCP stability time course where solid symbols represent N-methyl amide-containing MCPs and open symbols the corresponding amide analogs, where solid dashed lines represent analog pairs. MCPs pairs containing an internal standard peptide were incubated with human serum at 37°C as described in materials and methods. At 0, 1, 3, 9, 24, and 72 h, the human serum was reduced with 10 mM TCEP and relative intensity of each peptide to the standard peptide was determined by LC/MS. Relative intensity at 0 h was defined as 100%. The experimental data were fitted by non-linear regression one phase decay curve using GraphPad Prism 9.2.0.

As illustrated in Figure 2e, N-methylation can significantly improve a peptide’s serum stability but does not do so universally. In fact, in one example, peptide Mo-L1, we found that N-methylation lowered the peptide’s serum stability (Figure 2f). Here, the structural shift induced by the N-methyl groups may result in a peptide topology where the endopeptidase cleavage site is more exposed than it would be in an all-pAA peptide. Nevertheless, in half of the peptide pairs analyzed we found considerable improvement of serum stability mediated by the presence of N-methylation. Remarkably, it appeared that the peptides Mo-L2 and Mo-D2 are completely resistant to degradation by serum-borne metabolic enzymes.

This data suggest N-methylation appears to affect the MCP conformation so significantly that retention of the combinatorial effect on activity and metabolic stability after conversion between proteogenic amide and N-methyl amide backbone residues is exceptionally unlikely. These results, therefore, demonstrate the potential advantage of incorporating N-MeAAs during the discovery phase of peptide ligands, rather than during the medicinal chemistry phase of lead optimization to have maximal impact of N-methylation on MCP activity and metabolic stability.

Sa-D2, a lariat macrocyclic structure, shares a Ce-2 inhibitory mechanism

Of the analyzed MCPs, Sa-D2 stands out displaying potent, cross-ortholog affinity and inhibitory activity, structurally resembling the terminal Cys lariat structure of Ce-2 Ipglycermide but having a larger macrocycle and a shorter extra cyclic extension (Figure 3a and Table 1). Based on this, Sa-D2 may utilize the same inhibition mode as Ce-2; a thiolate-metal ion coordinated interaction. Unlike Ce-2, however, Sa-D2 appears to either interact with a more conserved iPGM binding site, and/or requires fewer contacts allowing the MCP to strongly inhibit both nematode and prokaryotic iPGM orthologs rather than displaying the potent nematode iPGM selectivity of Ce-2 Ipglycermide.

Figure 3. Functional, binding kinetics and thermal stability characterization of Sa-D2 and –D3 cyclic peptides and analogs.

(a) CRCs for activity of Sa-D2 on representative iPGM orthologs from the protozoan L. mexicana (Lm); bacteria S. aureus (Sa), E. coli (Ec), H. pylori (Hp) and M. orale (Mo), and nematode C. elegans (Ce) and W. bancrofti (Wb). (b, c) Sensorgrams for Sa-D2 (0.5 – 31 nM) or Sa-D2 (8 – 500 nM) binding kinetics derived from a C. elegans or S. aureus iPGM, respectively, conjugated sensor chip surface. (d) First derivative plots of the fluorescence emission as a function of temperature for either Sa-D2 (dashed line, T_m=48.7°C) or Sa-D3 (blue line, T_m=62.4°C) binding to S. aureus iPGM vs DMSO control (black line, T_m=45.7°C). (e, f) Sensorgrams for Sa-D3 (156 nM - 10 μM; K_D=790 nM, t_½ =27s) and Sa-D3 ∆N-Me (39 nM - 10 μM; K_D=280 nM, t_½ =25s) binding kinetics and saturation binding curve (inset) for S. aureus iPGM. The experimental SPR data (black) were globally fit to a 1:1 binding model (blue lines) using BIAevaluation, as described in Experimental procedures, to determine kinetic rate constants. Corresponding insets show the signal observed at equilibrium, R_eq, plotted as a function of the [MCP], fit to a hyperbolic, single-site binding equation. Data are representative of N≥3 independent experiments.

The Sa-D2 − C. elegans iPGM binding kinetics derived K_D = 1.8 nM and t_½= 15 min (Figure 3b), are within the same order of magnitude observed for Ce-2 binding to the C. elegans iPGM, B. malayi ortholog (K_D = 0.5 nM, t_½= 52 min), despite C. elegans iPGM sharing 71% and 40% identity with B. malayi and S. aureus iPGM, respectively^{22, 23}. This suggests that cross-species ortholog inhibitors, such as Sa-D2 are more likely to be derived from pool-and-split selection approaches. Interestingly, though demonstrating inhibitory activity on M. orale and H. pylori iPGM, Sa-D2 does not inhibit S. aureus iPGM despite appearing to bind strongly based on SPR analysis (Figure 3a, c).

The observation that Sa-D2 can bind to S. aureus iPGM but not inhibit the functional enzyme activity explains how Sa-D2 was enriched by S. aureus iPGM in the final affinity selection rounds. The lack of functional activity is likely because enzymatically active S. aureus iPGM requires Mn²⁺, whereas the thiolate-coordinating complex is only achievable with a Zn²⁺ containing enzyme. Supporting this is the absence of a significant thermal shift in the melting temperature of active S. aureus iPGM by Sa-D2 (Figure 3d; ∆T_m=3°C). Whereas, the non-thiolate containing inhibitor Sa-D3, which functionally inhibits (Table 1) and displays measurable binding kinetics (K_D = 790 nM, t_½= 27 s; Figure 3e) clearly stabilized the melting temperature (∆T_m= 17°C) of the enzyme. Taken together, this suggests that Sa-D2 strongly binds to the minority of inactive Zn²⁺ coordinating S. aureus iPGM but lacks affinity to the active Mn²⁺ coordinating enzyme.

To further investigate binding requirements of the Sa-D2 structure, four additional Sa-D2 analogs were prepared and characterized on the iPGM enzyme panel. These MCPs include a Cys13Ser substitution (Sa-D2 C13S), a truncation of the extra cyclic sequence, Pro11Lys12Cys13 (Sa-D2 ∆PKC), and the larger 13 amino acid macrocyclic analog (Sa-D2 LgMC). The effects of each modification on inhibitory activity, shown in Table 1 and Supplementary Figure 4b, were either lost or significantly diminished. These observations reinforce the hypothesis that, like Ce-2 Ipglycermide, the second cysteine (Cys13) of Sa-D2 is also anchored into the iPGM metal-binding site and may have a similar, though more ortholog-ambiguous mode of action. However, despite either relocation of the free thiol to the interior of the MCP (i.e., Sa-D2 LgMC) or substitution for a non-nucleophilic residue (Sa-D2 C13S), both analogs retained some ability to either inhibit or only bind an iPGM (Table 1 and Supplementary Figure 4a). In contrast, the apparently minor modification of substituting the N-methyl amides at Tyr5 and Tyr8 for amides in either Sa-D2 (Sa-D2 ∆N-Me) or Sa-D2 LgMC (Sa-D2 LgMC ∆N-Me) results in total loss of the cyclic peptide activity (Supplementary Figure 4b), suggesting a key structural role required for Sa-D2 activity.

Sa-D2 • C. elegans iPGM co-crystal structure: limited contacts contribute to broad iPGM species selectivity

For a Sa-D2 • C. elegans iPGM complex we obtained crystals diffracting to 1.90 Å having clearly defined electron density maps permitting all atoms for the cyclic peptide to be modeled except for the sidechain of Arg2 (Supplementary Figure 5a). At the phosphotransferase domain, Sa-D2 forms hydrogen bond interaction with G145, R216 and D286 as depicted in Figure 4a.^* Tyr6 of Sa-D2 binds in the same basic region near R284 (Figure 4b), in proximity to the 3PG binding site, as the position 7 aromatic amino acid substitutions, Phe, His and Tyr, observed in the co-crystal structures of Ce-2 Y7F (PDB: 7KGN), Ce-2 Y7H C14NHOH (PDB: 7KNF) and Ce-2d (PDB: 5KGN), respectively^{22, 23}. Further, the sidechain carboxylate of Sa-D2 Glu9 approaches R284 from the opposite side as Tyr6 as shown in Figure 4b. Superposition of the C. elegans iPGM co-crystal structures bound to Ce-2 Y7F and Sa-D2 yielded an RMSD deviation of 2.30 Å between Cα atoms (291 residues). Although the overall domain structures are similar, the secondary structure elements are offset slightly, particularly in the transferase domain, though the proximity of Tyr6 in Sa-D2 to Phe7 in Ce-2 Y7F can clearly be observed (Figure 4c).

Figure 4. Ipglycermides Ce-2 Y7F and Sa-D2 have unique C. elegans iPGM Zn²⁺-coordinating binding modes.

(a) Key protein-ligand interactions and Cys13 – Zn²⁺ coordination observed between Sa-D2 and C. elegans iPGM. N-Me amides are shown with hydrogens (white) for clarity. (b) Two-dimensional protein-cyclic peptide pharmacophore interactions. Phosphatase (apricot) and phosphotransferase (pastel blue) domain are indicated having H-bonding (black dashes), ionic (red dots), or π-cation (purple dots) interactions. (c) Superposition of C. elegans iPGM•Sa-D2 (dark blue/dim grey; PDB: 7TL7) and C. elegans iPGM•Ce-2 Y7F (ice blue/violet; PDB: 7KNG) structures fixed over the amino acid range of 27–95 and 339–538 (phosphatase domain). Zn²⁺ ions are sea blue. (d) Conformation of Sa-D2 as bound to C. elegans iPGM, with the two cis N-methyl amides at Tyr6 and Tyr8 indicated by arrows.

The terminal Cys13 residue sidechain interacts with the Zn²⁺ ion through the S-atom (2.32 Å). This Zn²⁺ ion coordination resembles that observed for C. elegans iPGM in complex with the Ce-2 ipglycermide analog, Ce-2 Y7F (PDB 7KNG) however, the orientation of the Sa-D2 is strikingly different. Notably, although the two peptides ultimately coordinate a Zn²⁺ ion via their C-terminal Cys residues, the orientations in the binding pocket are markedly different (Figure 4c and Supplementary Figure 5b). This may be due to the larger cyclic portion of Sa-D2 versus Ce-2 Y7F which have 10 and 8-membered macrocycles respectively.

Strikingly, we observe both the Tyr5 and Tyr8 N-methyl amides of Sa-D2 adopt the cis configuration as shown in Figure 4d. The amide carbonyl of Tyr8 is positioned to hydrogen bond with N85, a phosphatase domain residue conserved among the iPGMs, proximal to the Cys13 thiolate–Zn²⁺ coordinating interaction (Figure 4a). In the Ce-2 Y7F structure, the neighboring E87 engages with Thr13 potentially reinforcing the Cys14 thiolate–Zn²⁺ interaction. Interestingly, these conserved iPGM residues, N85 and E87 are located on either side of S86, the serine nucleophile at the center of the catalytic mechanism of phosphoglycerate isomerization.

Overall Sa-D2, not surprisingly, shares only key binding interactions with highly conserved residues and regions shared by prokaryotic and nematode iPGMs. Unlike the extensive interactions Ce-2 ipglycermides make with the Hinge 1 and 2 regions of nematode iPGMs, there are no residues from Sa-D2 that venture into this area of relatively low conservation between nematode and prokaryotic iPGMs (Figure 4c).

The Sa-D3 macrocycle features a 3PG mimetic carboxylate

Sa-D3 represents the majority of ipglycermides structures identified in this study, having exclusively a macrocyclic configuration without a free thiolate-bearing linear tail region. To explore the novel inhibitory mechanism of Sa-D3 we obtained the co-crystal structure of this ligand with S. aureus iPGM (Figure 5a). The electron density maps for Sa-D3 were clearly defined and all atoms for the macrocycle could be modeled (Supplementary Figure 5c). Along with the native diffraction data, a second data set was collected from another crystal at a wavelength of 1.3Å, which is on the low-energy side of the Zn²⁺ absorption edge. A strong anomalous signal was observed at the metal-binding sites (Supplementary Figure 5d) suggesting occupancy by Mn²⁺ ions, in agreement with previous literature^{19, 20}. The Mn²⁺ ions interact with D397, H401, H456 and three water molecules at site 1 and D12, S62 (the active-site serine nucleophile equivalent to S86 in C. elegans iPGM), D438 and H439 at site 2 (Supplementary Figure 5d).

Figure 5. Sa-D3 ipglycermide partially mimic 3PG binding to S. aureus iPGM.

(a) Conformation of Sa-D3 as bound to S. aureus iPGM, showing the trans N-methyl amide at _MeSer9 (shown with N-Me hydrogens (white) for clarity) and Asp13 interaction with R185, R191, R257 and R260 of the 3-phosphoglycerate (3PG) binding site. (b) Two-dimensional protein-cyclic peptide pharmacophore interactions. Phosphotransferase (pastel blue) domain and Hinge regions (celadon) are indicated having H-bonding (black dashes) or ionic (red dots) interactions. An edge-face aromatic F377 interaction from the phosphatase domain (apricot) is shown by purple dots. (c) Electrostatic surface representation of S. aureus iPGM • Sa-D3 complex where the cyclic peptide (grey) surface is rendered in yellow and the trans N-methyl amide circled. (d) Same as panel c but rotated 90° about horizontal axis to show 3PG binding site. (e) Superposition of the S. aureus iPGM - Sa-D3 (white; PDB: 7TL8) with 3PG bound (PDB 4NWJ) S. aureus iPGM (lilac). The Sa-D3 cyclic peptide is rendered as grey, and 3PG as green/magenta cylinders (chemical structure shown in inset). Prepared with CCP4MG, version 2.10.10.

The cyclic peptide forms hydrogen bond interactions with the backbone carbonyl of Trp6 and Q77 Nε; the backbone carbonyl of Trp7 and D78 Oγ; the backbone carbonyl of N-Me Ser9 Oγ and R257 NH1; the backbone carbonyl of Glu12 and R257 NH1; the backbone carbonyl of Asp13 and R257 NH1; the backbone carbonyl of Asp13 and R191 NH1; the side chain of Asp13 OD2 and R191 NH1 and the side chain of Asp13 OD2 and R260 Nε (Figure 5a–b). The latter hydrogen bonds between Asp13 of the cyclic peptide and R191/R260 of S. aureus iPGM form strong interactions between residues of opposite charge.

We had observed that the single N-Me amide (Ser9) had a marginal contribution to the inhibitory potency and binding kinetics of Sa-D3 on M. orale and S. aureus iPGMs (Table 1 and Figure 3f). Notably, unlike the cis configurations observed for the N-methyl amides of Sa-D2, the Ser9 N-methyl amide adopts the lower energy trans configuration (Figure 5a).

Sa-D3 has a surface area of 1594.2 Ų and 924.1 Ų of this total area is buried within the S. aureus iPGM interdomain cleft. Residues D-Tyr1, Trp6 and Trp7 are snuggly fit within pockets of the binding site (Figure 5c). Similarly, Trp11 is positioned within a relatively neutral cleft and Asp13 occupies a positively charged pocket-forming the interactions noted above (Figure 5d). Interestingly, the side chain of Glu12 does not form interactions with the Arg residues in this region and is directed away from the iPGM surface presumably due to the proximity of D154 which repels the side chain.

The overall S. aureus iPGM complex with the Sa-D3 peptide is quite similar to previously determined S. aureus iPGM structures. Notably, the superposition of S. aureus iPGM•Sa-D3 with apo S. aureus iPGM (PDB 4MY4¹⁹) and 3-phosphoglycerate bound S. aureus iPGM (PDB 4NWJ¹⁹) yielded RMSD deviations between Cα atoms of 0.99 Å (484 residues) and 0.97 Å (485 residues) respectively. The 3-phosphoglycerate molecule in 4NWJ occupies a position between Glu12 and Asp13 of the Sa-D3 peptide as observed in the superposition of co-crystal structures 7TL8 and 4NWJ shown in Figure 5e.

Conclusion

Tuning the affinity of N-MeAA-tRNA to EF-Tu via the T-stem domain (Figure 1a) enables the ribosomal synthesis of more exotic N-MeAA containing peptides than previously possible, allowing for screening and discovery of heavily N-methylated MCPs. Using this method we discovered a broad variety of N-methylated MCPs targeting prokaryotic iPGMs containing between one and eight N-MeAAs (Table 1). Interestingly, as determined with respective non-N-methylated MCP analogs, the N-methylation state of the peptide may have a range of effects on the MCP to act as an inhibitor of a given iPGM which generally aligns with its binding kinetics (Table 1 and Figure 2). In our limited evaluation of N-methylation state on metabolic stability we observed that in some cases N-methylation greatly prevented the serum protease degradation of the MCP, as seen with Mo-D1 and Mo-D2, or had minimal to no discernable effect in others, while in one case, Mo-L1 N-methylation hastened its degradation (Figure 2f).

In our previous studies we obtained nematode iPGM selective ipglycermide Ce-2 (Table 1) using C. elegans iPGM for the affinity selection²³. To increase the probability of isolating cross-iPGM ortholog selective MCPs we pooled three prokaryotic iPGMs to initially enrich the library, performing the final selections on the individual enzymes (Figure 1b). The approach successfully isolated several potent cross species inhibitors, as well as highly selective MCPs where potency is tightly tied to N-methylation state (e.g., see Mo-L/D series Table 1). Remarkably, while S. aureus iPGM captured Sa-D2, a potent nematode and M. orale iPGM inhibitor, this MCP has no significant inhibitory activity on S. aureus iPGM (Figure 3a and Table 1). We speculate that a fraction of S. aureus iPGM contains Zn²⁺ and represents the functionally inactive species that bound Sa-D2 resulting in the observable signal by SPR (Figure 3c). Whereas, under thermal shift conditions (e.g., [E]>>K_D), Sa-D3 stabilized the enzyme from thermal denaturation while Sa-D2 fails to do so (Figure 3d), further suggesting any Zn²⁺ species is a minor component.

Co-crystal structures of two iPGM•MCP complexes provide insight into the structural basis for novel binding modes, cross-ortholog selectivity, and define the stereochemistry around the N-methyl amide residues in Sa-D2 and Sa-D3. Unexpectedly, we observed that the two N-methyl amides required for Sa-D2 activity reside in the cis configuration (Figure 4d), while in Sa-D3 we observe a trans configuration for the single, non-essential N-methyl amide (Figure 5e). Sa-D2, having a cys-terminal lariat coordinates the active-site Zn²⁺ correspondingly to the previously reported Ce-2 ipglycermide, but with expanded cross-ortholog activity, likely resulting from the remainder of the macrocycle making fewer interactions and concentrated at non-conserved regions of the enzyme (Figure 4c). Sa-D3 represents a completely novel non-lariat macrocycle inhibitor that acts as a 3-phosphoglycerate mimetic stabilized by a hydrophobic interaction mediated by Tyr1 (Figure 5c). In this structure Asp13 forms ionic bonds with the basic phosphoglycerate-binding region formed by R191, R257 and R260 (Figure 5d). Interestingly, Glu12 may present a repulsive interaction with D154.

In summary, we demonstrated the de novo incorporation of N-methylation into high affinity MCPs and the impact these privileged residues have on target affinity and metabolic stability. Above all, the number of potent and structurally varied MCPs we discovered that target iPGMs with cross-ortholog selectively demonstrates that the isolated success of Ce-2 was not just a lucky coincidence. In fact, these results demonstrate that iPGM can no longer be considered as an undruggable target and pave the way for the potential development of an entirely new class of antimicrobials.

Materials and methods

In vitro translation system.

The Flexizymes dFx and eFx, as well as mRNAs were prepared as previously reported ²⁷. T-stem appropriate tRNAs were matched to their respective residues and prepared as previously reported ¹⁵. Cyanomethyl ester (CME) and 3,5-dinitrobenzyl ester (DBE) activated amino acids _MeY^(OMe)-CME, _MeF-CME, _MeS-DBE, _MeNle-DBE, _MeY-CME, _MeG-DBE, and ClAc-(L/D)-Y-CME were prepared as previously reported ^{7, 27, 28}. The cell free translation system used in this study was previously optimized ^{29, 30}, and adapted for the translation of N-methylated peptides ¹⁵.

Macrocyclic library design

The macrocyclic peptide library was constructed as a standard RaPID selection library with 7–5 randomized residues encoded by NNK codons and initiated with either ClAc-L-Tyr or ClAc-D-Tyr ³¹. The prepared DNA library was transcribed, ligated to puromycin, in vitro translated, and reverse transcribed prior to affinity selection against iPGMs.

Affinity selection and enrichment

Prior to the selections, immobilization rates of the three iPGM targets, S. aureus iPGM-His6, H. pylori iPGM-His10, and M. orale PGM-His10, to Dynabeads His-tag Isolation & Pulldown (Invitrogen) were determined to be 10 pm/μL for, 20 pm/μL, and 30 pm/μL respectively (Supplementary Figure 1).

The N-methylated mRNA-puromycin conjugated MCP libraries described above were synthesized by in vitro translation, reverse transcribed and subsequently desalted with Sephadex G-25 (Cytiva) in PBST. From this desalted solution, 0.5 μL was diluted 1000x as an input quantification sample. The libraries were cleared 6x with Dynabeads His-tag Isolation & Pulldown to remove His-tag binding peptides and His-tagged proteins of the FIT system from the solution. The three iPGM orthologs were immobilized onto Dynabeads at the previously established saturation points and mixed to have equal quantities of each orthologue in the final mix. The MCP library was incubated with 67 nM of each immobilized iPGM with 200 nM total iPGM for 30 min. The beads were washed, and cDNA of binding peptide was quantified by qPCR and amplified by PCR. The amplified cDNA was transcribed into mRNA and the method repeated using this new library. In round 2, following the desalting step of the MCP library, blocking buffer containing acetylated BSA and yeast RNA was added to limit nonspecific binding. Furthermore, cDNA of MCPs binding to the last set of negative beads was eluted and quantified by qPCR to monitor enrichment of bead binding MCPs.

During round 4, the MCP libraries were split into 3 and overlaid onto 200 nM each of the three individual, immobilized iPGM orthologs separately. The six MCP libraries that resulted from this were further enriched until an exponential increase in recovery was observed by qPCR. This happened at round five for all libraries except the ClAc-D-Tyr initiated library enriching against S. aureus iPGM, which required an additional round. After enrichment was observed (Supplementary Figure 2), the cDNA of the libraries was analyzed by next-generation sequencing and enriched peptides were identified.

Chemical synthesis of N-methylated MCPs

Precyclized peptides were synthesized by standard Fmoc automated SPPS at a 25 μmol scale using Rink Amide resin ³². Standard couplings were conducted using HOBt/HBTU at 30°C for 40 minutes twice. Couplings to N-methylated amines were conducted using HOAt/HATU at 60°C for 40 minutes twice. Tyr(Me) was manually N-methylated via on-resin Mitsunobu coupling ³³. Peptides that contained multiple cysteine residues were prepared using Fmoc-Cys(Mmt)-OH for the cyclizing cysteine and Fmoc-Cys(Dpm)-OH for the others. Following SPPS, the N-termini were chloroacetylated using a 0.2 M solution of chloroacetoxy succinimide ClAc-NHS in DMF. The Mmt group of peptides containing multiple cysteine residues was deprotected with 5% TFA and 2.5% triisopropylsilane in DCM. These peptides were subsequently cyclized on resin by overnight agitation in 5% diisopropyl ethylamine in DMF.

Peptides were deprotected and cleaved off the resin with a 2.5% triisopropylsilane, 2.5% DoDT, 2.5% water in TFA cocktail for 3 h at RT. The resin was filtered, washed with TFA, and concentrated in vacuo followed by precipitation with ice-cold diethyl ether. The precipitated peptides were washed with diethyl ether and dried overnight. The dried peptides were dissolved in DMSO and cyclized by addition of triethylamine if they were not previously cyclized on resin. Peptides were finally purified by reverse-phase preparatory HPLC and identified by MALDI-TOF-MS and analytical UPLC (Supplementary Figure 3, Supplementary Table 3).

Serum stability assay

Human serum stability of the synthesized peptides was assayed similarly as previously described ³⁴. Both the N-methylated peptide and corresponding ΔN-Me analogue were mixed with internal control peptide in human serum to a final concentration of 10 μM. Peptides Ce-2 and Ce-2d were paired in a similar manner. At every time point (t= 0, 1, 3, 9, 24, 72h.) 13.5 μL of human serum was mixed with 1.5 μL 100mM TCEP to reduce the sample. Serum proteins were precipitated by addition of 35 μL methanol and incubation on ice for 5 min. The samples were centrifuged, and the supernatant was stored at −20°C until analysis. Immediately prior to analysis the samples were diluted with 9 volumes 1% TFA in water and centrifuged for 5 minutes prior to analysis by UPLC-ESI-MS. To control for serum activity, a canonical, linear peptide was assayed in parallel and observed to be fully degraded in approximately nine hours.

Preparation of iPGM enzyme constructs

W. bancrofti iPGM (10xHis C-terminal tag) EJW84318.1; M. orale iPGM (10xHis C-terminal tag, 1293A>T to avoid introduction of a premature stop codon during bacterial expression) NZ_LR214940.1; M. pneumoniae iPGM (10xHis C-terminal tag), CP010547.1; H. pylori iPGM, (10xHis C-terminal tag) CP001173.1; L. mexicana iPGM (10xHis C-terminal tag), AJ544274; and S aureus iPGM (6xHis C-terminal tag), BX571856.1 were cloned into pET21a(+) and transformed into BL21(DE3) for E. coli expression as described²³. For site-specific incorporation of biotin, the S. aureus iPGM was constructed to have the GLNDIFEAQKIEWHE sequence³⁵ located between the C-terminus and His tag as described in Supplementary Methods.

iPGM expression and purification

Expression and purification of the iPGMs (C. elegans, B. malayi, W. bancrofti, E. coli, S. aureus, M. orale, M. pneumonia, H. pylori and L. Mexicana) were carried out by following the previously described methods.^{22, 23} For the functional enzyme assays, 1 mL each of C. elegans, B. malayi, M. orale, and S. aureus iPGMs were supplemented with 10 μL of 10 mM ZnCl2 and 10 μl of 10 mM MnCl₂ then dialyzed for 16 hours against 1 L of 1X PGM buffer containing 20% glycerol using Slide-A-Lyzer^™ Dialysis Cassettes (Thermo Scientific, Cat# 66380). The dialyzed samples are stored at 4°C up to 8 weeks.

In vitro biotinylation

iPGM (C. elegans and S. aureus) biotinylation reactions were carried out at 100 μM scale in PBS buffer according to the procedure described in Fairhead and Howarth³⁵ as applied in Wiedmann et al.²² Free biotin and E. coli Bir-A were removed by loading the reaction mixture onto a HiLoad Sephadex 16/60 prepacked column and eluted at 0.5 mL/min with PGM storage buffer. Biotinylation efficiency was measured using the streptavidin conjugation assay.³⁵

PGM enzymatic assays and AC₅₀ determinations

Inhibition of phosphoglycerate mutase activity in the presence of macrocyclic peptides was measured as a luminescence end point output assay as previously described²³. Briefly, 4 μL enzymes were dispensed into white solid-bottom medium bind 1536-well plates (Greiner Bio-One) in a pH 8.0 assay buffer (30 mM Tris-HCl, 5 mM Mg2SO4 and 20 mM KCl) with the BioRaptor FRD (Beckman Coulter). Enzymes were run at the lowest concentration determined to give significant signal when run at an estimated Km substrate concentration (5 nM Caenorhabditis elegans and Brugia malayi iPGM, 40 nM Wuchereria bancrofti iPGM, 8 nM Mycoplasma orale and Mycoplasma pneumoniae iPGM, 50 nM Escherichia coli and Trypanosoma brucei iPGM, 10 nM Staphylococcus aureus iPGM, and 100 nM Helicobacter pylori iPGM), while H. pylori iPGM was run in the presence of 5 μM ZnCl2 and S. aureus and T. brucei iPGMs were run in the presence of 100 μM MnCl2. Assay buffer with no enzyme was added to column 1 of each respective plate as a control. Macrocyclic peptides were prepared at 5 – 20 mM in DMSO and added to enzymes at 23 nL per well transferred by a Pin tool (Wako Inc.) in a 16-pt, 1:3 titration in duplicate spanning a concentration range of 1.3 pM to 76.8 μM along with DMSO and a titration of Ce-2 and Ce-2d positive controls from top concentration 3.8 μM to 0.27 pM in columns 2–4 of each plate. Enzymes were incubated with peptide for >20 min at ambient temperature, protected from light. 2 μL of 3-phosphoglycerate in a coupled enzyme assay buffer (includes enolase and pyruvate kinase) was added to enzyme/peptide mixtures with BioRaptor FRD and reactions were incubated for 5 min at ambient temperature, protected from light for all enzymes except Hp iPGM that was incubated with substrate for 10 min at ambient temperature and E. coli iPGM where reactions were incubated at 37^oC for 15 min, followed by addition of 4 μL Kinase-Glo Plus reagent (Promega Corporation). Plates were incubated at ambient temperature for 10 min then measured by a ViewLux plate reader (PerkinElmer). See also Supplementary Protocol Table 1. CRCs were fit and AC₅₀ values calculated using nonlinear regression log(inhibitor) vs. response -- Variable slope (four parameters) in Prism (GraphPad Software).

Thermal Shift Assays

Ligand-stabilized iPGM melting temperatures were determined using the Thermofluor method.³⁶ Briefly, samples were tested in MicroAmp optical 384-well reaction plates (Applied Biosystems by Life Technologies) using a ViiA 7 real time PCR instrument (Applied Biosystems) and the fluorescent dye SYPRO Orange (Life Technologies). The final concentration of protein and dye were maintained at 5 μM and 10x respectively. The proteins were diluted in PBS (10 mM Na2HPO4, 1.8 mM KH2PO4, 137 mM NaCl and 2.7 mM KCl) and added to the 384 well plate, followed by compounds also diluted in PBS (final concentrations between 0 to 100 μM). Samples were incubated for 10 mins at ambient temperature and protected from light. Next, 4 μl of dye solution (initial concentration of 5000x diluted in PBS to 50x) was added to the each well using a multi-channel pipette. The plate was sealed with a MicroAmp optical adhesive film (Applied Biosystems), centrifuged for 1 min at 1000 RPM then loaded on to the real time PCR instrument for data collection after samples were initially cooled to 4 °C for 1 minute then ramping temperature from 4 to 95 °C at 1°/min. increments. The first derivative of the melting data was plotted in GraphPad Prism to determine Tm. A step-by-step detailed assay protocol is described in Supplementary Protocol Table 2.

SPR experiments

SPR experiments were performed using a Biacore S200 instrument at a flow rate between 20–40 μL/min, 25°C using 1X PBP-P as a running buffer as previously described.²² Briefly, biotinylated iPGMs were immobilized on a Biotin CAPture (CAP) chip. High performance kinetic experiments were performed by injecting a series of macrocyclic peptide concentrations (within 10-fold of K_D). Kinetic and affinity data was analyzed using Biacore S200 Evaluation Software. All sensorgrams were fitted, after background correction to a 1:1 binding model using the BIAevaluation software as previously described.²².

Crystallization and Data Collection

Purified apo C. elegans and S. aureus iPGM spanning residues M19−I539 and M26−H538, respectively, both harboring a C-terminal hexahistidine tag were concentrated to 12.0 mg/mL (0.2 mM) in 150 mM NaCl, 30 mM Tris pH 8.0, 1 mM TCEP. To prepare the respective Sa-D2 and Sa-D3 cyclic peptide complexes a 20 mM stock solution was prepared in DMSO, mixed in a 1:1.25 (protein:cyclic peptide) molar ratio and incubated on ice for 30 minutes prior to screening. All crystallization experiments were setup using an NT8 drop-setting robot (Formulatrix Inc.) and UVXPO MRC (Molecular Dimensions) sitting-drop vapor diffusion plates at 18 ^oC. 100 nL of protein-ligand mixture and 100 nL crystallization solution were dispensed and equilibrated against 50 μL of the latter.

For the C. elegans•Sa-D2 complex, crystals displaying a prismatic morphology were observed within approximately 7 days, from the Morpheus Screen HT³⁷ (Molecular Dimensions) condition H4, which contains the following mixtures: 37.5% (w/v) of precipitant mix 4: (25% (v/v) MPD; 25% (w/v) PEG 1000; 25% (w/v) PEG 3350), 100 mM of MB1 pH 6.5: (1.0 M imidazole, MES), 100 mM of MAA: (0.2 M DL-Glutamic acid monohydrate; 0.2M DL-Alanine; 0.2M Glycine; 0.2M DL-Lysine monohydrochloride; 0.2M DL-Serine). For the S. aureus•Sa-D3 complex, crystals displaying a plate morphology were observed within approximately 14 days from the Index HT Screen (Hampton Research) condition G7, containing 25% (w/v) PEG 3350, 100 mM Bis-Tris pH 6.5, 200 mM ammonium acetate. A cryoprotectant composed of 80% crystallant and 20% PEG 200 was layered onto the drop, samples were harvested and stored in liquid nitrogen for X-ray diffraction data collection.

Structure Solution and Refinement

X-ray diffraction data were collected at the Advanced Photon Source beamline 17-ID using a Dectris Eiger2 9M X detector for C. elegans•Sa-D2, and at the National Synchrotron Light Source II (NSLS-II) beamline 19-ID (NYX) using a Dectris Pilatus 6M pixel array detector for S. aureus•Sa-D3. Intensities were integrated using XDS^{38, 39} via Autoproc⁴⁰ and the Laue class analysis and data scaling were performed with Aimless⁴¹ which indicated that the highest probability Laue class was 2/m.

Structure solution was conducted by molecular replacement with Phaser⁴² using a previously determined isomorphous structure (PDB 7KNF²²) for C. elegans•Sa-D2, and the structure PDB 4MY4¹⁹ for S. aureus•Sa-D3, as the respective search models. For C. elegans•Sa-D2 the top solution was obtained in the space group P2₁ with 4 molecules in the asymmetric unit. For S. aureus•Sa-D3 the top solution was obtained in the space group P2₁ with a single molecule in the asymmetric unit, and the model was improved by automated model building Buccaneer⁴³.

For both, structure refinement and manual model building were conducted with Phenix⁴⁴ and Coot⁴⁵ respectively. Disordered side chains were truncated to the point for which electron density could be observed. Structure validation was conducted with Molprobity⁴⁶ and figures were prepared using the CCP4MG package⁴⁷. Superposition of iPGM-Ce structures was conducted using GESAMT⁴⁸ via the CCP4⁴⁹ interface over the phosphatase region (A/27–95 and A/339–538). Polder omit maps⁵⁰ were calculated with Phenix. Relevant crystallographic data are provided in Supplementary Table 1. Structures were further analyzed for binding interactions using Maestro (Schrödinger, LLC) by adding hydrogen atoms, followed by a rigid energy minimization using the OPLS4 force field⁵¹.