A photoaffinity probe that targets folate-binding proteins

Abstract

Folate and related derivatives are essential small molecules required for survival. Of significant interest is the biological role and necessity of folate in the crosstalk between commensal organisms and their respective hosts, including the tremendously complex human distal gut microbiome. Here, we designed a folate-based probe consisting of a photo-crosslinker to detect and quantitate folate-binding proteins from proteomic samples. We demonstrate the selectivity of our probe for the well-established human folate-binding protein dihydrofolate reductase and show no promiscuous labeling occurs with human caspase-3 or bovine serum albumin, which served as negative controls. Affinity-based enrichment of folate-binding proteins from an E. coli lysate in combination with mass spectrometry proteomics verified the ability of our probe to isolate low-abundance folate-dependent proteins. We envision that our probe will serve as a tool to elucidate the roles of commensal microbial folate-binding proteins in health and microbiome-related diseases.

Graphical Abstract

Folate and related derivatives are essential nutrients for most organisms, as these molecules are key participants in the biosynthesis of nucleobases, proteins, and other cofactors.^{1, 2} Humans cannot manufacture folate de novo and rely on uptake of this small molecule from exogenous sources, such as diet and folate-synthesizing commensal bacteria localized to the large intestine (e.g., distal gut microbiome).³ Such species include lactic acid bacteria and members of the Bifidobacterium genus that contain the necessary genes required for the de novo synthesis of folate. The importance of folate synthesis to bacteria is best represented by a recent study on E. coli that showed disruption of folate biosynthesis significantly increased susceptibility of E. coli to antibiotics.⁴ While a select group of bacteria make folate, the majority of commensal organisms rely on folate scavenging similarly to their human host and suggests that those species with optimal folate-uptake properties may have a competitive advantage.³ As such, fluctuations in luminal concentrations of folate in the large intestine due to diet, de novo bacterial production and/or administration of folate-related drugs (e.g., methotrexate) will likely influence the taxonomic composition of human gut microbiota, interactions between the host and commensal organisms, as well as the overall health of the host.^{3, 5-7} To date, few studies have examined the biological relevance of folate and the impact of folate-dependent mechanisms to gut health and disease.

We ultimately aim to identify microbes from the distal gut and quantitate their respective folate-processing and folate-dependent proteins. Our overall approach relies on LC–MS/MS shotgun proteomics to identify proteins of interest; however, the species- and protein-level complexity of the gut microbiome renders it highly impervious to deep proteome profiling efforts without a pre-enrichment step.^{8, 9} Toward the goal of identifying gut microbes and microbial proteins that process folate and folate-related drugs (e.g., methotrexate), we focus our efforts on the design and evaluation of an affinity-based protein profiling probe for the enrichment of folate-binding proteins from complex proteomic mixtures, such as the human gut microbiota. The anatomy of a desirable probe requires three components: an affinity group (e.g., folate), an enrichment tag (e.g., biotin), and a covalent crosslinker (Fig. 1).

Fig. 1.

Structure of folate probe and a general schematic for the enrichment of probe-binding proteins.

Several molecular tools currently enable the visualization or covalent labeling of folate-binding proteins. The Numasawa and Dan groups have reported folate-linked fluorescent dyes for the visualization of folate receptors overexpressed on the membrane of tumors.^{10, 11} Those probes incorporated fluorescent dyes at the γ-carboxyl group of folate. Other groups have reported probes based on folate that can bind and covalently attach to dihydrofolate reductase (DHFR).^12-14 These groups utilized folate analogues and capitalized on the glutamic acid moiety of folate as an anchor point for the addition of glyoxal, sulfonyl fluoride, or aryl azide (nitrene-generating group) crosslinkers. However, the glutamic acid moiety on folate is typically required to bind most proteins, as demonstrated by co-complex crystal structures.^15-17 The two glutamate carboxylates provide hydrogen bond donors and acceptors to protein active sites and binding pockets and are critical for protein affinity.¹⁸ Of note, folate is polyglutamated upon cellular entry to yield a highly negatively charged molecule that is readily retained within cells. While the glutamate moiety extends outside the folate binding pocket with respect to the human folate receptor and DHFR,^{19, 20} this binding mode may not be generalizable to other organisms, especially members of the microbiota.²¹ Thus, we posited that a crosslinker might be better incorporated at a position that would not adversely affect potentially important binding interactions. In addition, it is widely established that the folate pterin group is reduced in metabolic processes, such that the folate can accept a one-carbon unit at the N-10 position and undergo a series of oxidative/reductive transformations, known as one-carbon metabolism.²² We therefore designed a probe with a crosslinker and enrichment tag attached on the phenyl ring to minimize any adverse effect upon binding to proteins (Scheme 1).

Scheme. 1.

a) iodine monochloride (2.7 eq), DMF, rt, overnight; b) N-Boc-propargylamine (5.0 eq), CuI (0.1 eq), PdCl₂(PPh₃)₂ (0.1 eq), DMSO-TEA (10:3, v/v), 90 °C, 20 min; c) TFA-DMF (1:6, v/v), 75 °C, 2 h; d) 3-{3-[3-oxo-3-(pent-4-yn-1-ylaminopropyl]-3H-diazirin-3-yl}propanoic acid (1.9 eq), HATU (1.5 eq), HOAt (1.5 eq), DMF-DIPEA (10:1, v/v), rt, 5 min.

We first constructed a modular crosslinking enrichment tag utilizing a UV-active diazirine moiety as a crosslinker and a terminal-alkyne as a proxy for biotin via the copper-catalyzed azide alkyne cycloaddition (CuAAC click reaction) (Scheme 1).^23-28 The synthesis began with a direct iodination of the 3’ position of p-aminobenzoic acid moiety of folic acid, a monoglutamate form of folate.^{29, 30} N-Boc-propargylamine was then incorporated via Pd–Cu catalyzed Sonogashira coupling.²¹ After deprotection of the Boc group, the carboxyl group of the diazirine linker was pre-activated, then conjugated to the free amino group of the propargylamino linker.³¹

Exposure of folate to UV light, as well as increases in temperature, promote cleavage between the pterin and p-aminobenzamide moieties³² and occurs through several proposed degradation mechanisms. One such mechanism that initiates the cleavage upon UV irradiation, includes a free radical formation on folate followed by oxidation to yield the degraded products.³³ Considering the generally high reactivity of radical species, we anticipated that non-specific modifications of proteins within a complex mixture may take place by reactive free-radical species generated during the labeling process (e.g., diazirine reactivity).³⁴ Importantly, ascorbate quenches free-radical reactions and may have a protective effect against folate degradation in the presence of UV light or high temperature, as previously suggested (Fig. S1).^35-37 While the effects of ascorbate on folate radical production has not been published, we opted to pre-mix ascorbate with the folate probe and proteins before UV irradiation in a preventative effort to limit degradation of the folate probe.

We evaluated the ability of our folate probe to irreversibly label a folate-dependent protein. DHFR, a well-established folate-binding protein, was subjected to probe labeling, as well as non-folate binding proteins bovine serum albumin (BSA) and human caspase-3 (i.e., negative controls). DHFR, BSA, and caspase-3 were treated with 250 μM of the probe in the presence of 2.5 mM sodium ascorbate and irradiated for 30 min under 365 nm UV light. After irradiation, the mixture was treated with biotin-PEG₃-azide under CuAAC conditions (see Supporting Information). Western blot membranes were stained with IRDye 800CW-conjugated streptavidin to reveal strong and selective labeling of DHFR by the folate probe with no promiscuous labeling of BSA or caspase-3 (Fig. 2A and B). Importantly, UV-based adherence of the folate probe to DHFR in the absence of sodium ascorbate resulted in labeling of several high-molecular weight bands and suggested DHFR dimerizes (e.g., 50 kDa band) and/or aggregates (e.g., bands >250 kDa) in a folate/probe-dependent manner (supplement data 1, Fig. S2). As such, sodium ascorbate is a necessary addition to the photo-crosslinking protocol to prevent folate degradation and promiscuous labeling.

Fig. 2.

(A) Western blot (left) and corresponding Coomassie-stained SDS-PAGE gel (right) of 250 μg/mL purified DHFR or BSA, or 34 μg/mL human caspase-3 labeled with 250 μM probe. (B) Western blot of three different concentrations of DHFR labeled with 250 μM probe in the presence and absence of 250 μM folic acid or methotrexate. (C) Western blot (left) and Coomassie-stained SDS-PAGE gel (right) of an E. coli lysate labeled with 100 μM of the folate probe. All samples (A-C) were exposed to 30 min of UV irradiation in the presence of 2.5 mM sodium ascorbate prior to analysis.

We wanted to ascertain if the folate probe labels DHFR at or near the folate-binding site and performed labeling in the presence and absence of folic acid or the folate-based drug methotrexate. Western blots showed a decrease in DHFR labeling in the presence of folic acid at various concentrations of DHFR (300, 100, 33 μg/mL), while methotrexate completely inhibited the labeling even in the highest concentration of DHFR (Fig. 2B). Qualitatively, the relative levels of inhibition by folic acid and methotrexate, as measured by Western blot, correlate well with the reported K_d values against human DHFR, including 111 nM for folic acid and 10 nM for methotrexate.^{38, 39} Our results provide with confidence that our folate probe labels at the folate binding site on DHFR and is likely a biologically relevant small molecule.

Finally, we sought to determine if our folate probe enriched folate-binding proteins from a complex mixture. To this end, we treated lysates of overnight E. coli cultures with 100 μM of our folate probe. After biotin-PEG₃-azide CuAAC-based attachment as described, labeled proteins were subsequently enriched with streptavidin agarose resin and partitioned for Western blot imaging (Fig. 2C) and for LC-MS/MS proteomics.

Our Western blots clearly demonstrate selective labeling of a fraction of proteins within the E. coli lysate, as the majority of proteins observed in the Coomassie-stained gel were not labeled (Fig. 2C). Additionally, we subjected the E. coli lysate to folate probe labeling in the presence of methotrexate and found that methotrexate eliminated all folate probe labeling, suggesting that our probe is likely selectively labeling folate-binding proteins (supplement data 1, Fig. S3). Enriched samples were subjected to on-resin trypsin digestion to generate LC-MS/MS-ready peptides. Of note, those bands that suggested significant folate probe labeling disappeared from the stained western blot membrane after trypsin treatment and suggested successful digestion of the target proteins (supplement data 1, Fig. S3).

In a data-dependent LC-MS/MS analysis, protein abundances were quantified by a label-free methodology relying on integration of precursor (MS1) peak intensities.⁹ Unique peptide peak intensities were mapped to their parent proteins, summed within parent protein groups, and normalized (See supplement data 1). Enrichment ratios were calculated by dividing peak intensities of probe-enriched proteins by those of their counterparts from unenriched lysate (Table 1). 1126 proteins were detected from an unenriched sample of E. coli lysate. Our probe enriched 382 proteins. 380 out of which, were also detected in unenriched lysate (supplement data 2).

Table 1.

LC–MS/MS analysis of enriched protein from lysate of single cultured E. coli. treated with our probe. Fold-change of peak intensities shows the area under the precursor curve intensity of enriched proteins relative to the value from unenriched sample (background).

Cluster representative	Description	Fold change of peak intensities	kDa
GlyA	Serine hydroxymethyltransferase	14.4	45
MoeA	Molybdopterin molybdenumtransferase	20.7	44
MurC	UDP-N-acetylmuramate-L-alanine ligase	240	53
MurQ	N-acetylmuramic acid 6-phosphate etherase	911	31
YtfE	Iron-sulfur cluster repair protein	592	25

Sixty-eight of the 382 probe-enriched proteins had greater than 10-fold higher normalized intensity values than in the corresponding unenriched lysate. Notably, serine hydroxymethyltransferase (GlyA), a known folate-related protein, was enriched by 14.4-fold with probe treatment compared the unenriched lysate, which provides evidence for the efficacy of our probe. Interestingly, molybdopterin molybdenumtransferase (MoeA) is enriched with a 20.7-fold change (Table 1). MoeA is known to bind to adenylyl-molybdopterin, which contains a pterin group similar to folate. Further experimentation is needed to verify if MoeA also binds folate.

Proteins that have not been characterized as folate-binding proteins to our knowledge were also significantly enriched, as exemplified by the fold-change of peak intensities, including proteins involved in muramic acid metabolism, UDP-N-acetylmuramate--L-alanine ligase (MurC) and N-acetylmuramic acid 6-phosphate etherase (MurQ), and an iron-sulfur cluster repair protein (YtfE) (Table 1). Of note, another enzyme involved in muramate metabolism, folylpolyglutamate synthase (FPGS), is known to bind folate and FPGS is structurally homologous to MurC despite limited sequence homology (supplement data 1, Fig. S4). Superposition of x-ray structures from L. casei FPGS in complex with 5,10-methylene-6-hydrofolic acid (PDB 1jbw)⁴⁰ and H. influenzae MurC in complex with uridine diphosphate-2(N-acetylgulcosaminyl)butyric acid (PDB 1p31)⁴¹ show that both enzymes coordinate a Mg²⁺ ion via His and Lys side chains. The Mg²⁺ ion in the H. influenzae MurC structure binds and orients a carboxylate from the butyric acid analog. This region is highly conserved with FPGS and suggests that MurC may also interact with the glutamic acid moiety of our folate probe (supplement data 1, Fig. S4). While there is no obvious conservation of the available MurQ structures with folate-binding proteins, MurQ may be enriched via complexes to one or more folate-binding proteins. Similarly, YtfE is not an established folate binder; however, other iron-sulfur cluster repair proteins in E. coli, including YgfZ, are known to bind folate.⁴² Additional studies will determine if MurC, MurQ and YtfE are directly or indirectly enriched by our folate probe.

In summary, we designed and synthesized a photo-crosslinking probe bearing folic acid to enable the selective capture folate binding proteins. We showed that our probe selectively labels human DHFR with little promiscuous labeling of BSA or human caspase-3. Using LC-MS/MS protein profiling, we also showed that our probe is capable of enriching for naturally occurring folate binding proteins from an E. coli lysate. Of note, we did not enrich E. coli DHFR in our proteomic analyses and is likely attributable to the structural differences in comparison to human DHFR. Protein profiling efforts employing this folate probe, as well as future analogs capable of broader labeling will greatly assist in identifying the role of folate-binding proteins in health and microbiome-related diseases. This proof-of-concept study also provides with a groundwork for future utilization of folate analog-bearing probes (e.g., methotrexate probe) in proteomic work.