Domain Fusion of Two Oxygenases Affords Organophosphonate Degradation in Pathogenic Fungi

Abstract

Proteins of the HD-domain superfamily employ a conserved histidine-aspartate (HD) dyad to coordinate diverse metallocofactors. While most known HD-domain proteins are phosphohydrolases, new additions to this superfamily have emerged such as oxygenases and lyases, expanding their functional repertoire. To date, three HD-domain oxygenases have been identified, all of which employ a mixed-valent Fe^IIFe^III cofactor to activate their substrates and utilize molecular oxygen to afford cleavage of C−C or C−P bonds via a diferric superoxo intermediate. Phylogenetic analysis reveals an uncharacterized multidomain protein in the pathogenic soil fungus Fonsecaea multimorphosa, herein designated PhoF. PhoF consists of an N-terminal Fe^II/α-ketoglutarate-dependent domain resembling that of PhnY and a C-terminal HD-domain like that of PhnZ. PhnY and PhnZ are part of an organophosphonate degradation pathway in which PhnY hydroxylates 2-aminoethylphosphonic acid, and PhnZ cleaves the C−P bond of the hydroxylated product yielding phosphate and glycine. Employing electron paramagnetic resonance and Mössbauer spectroscopies in tandem with activity assays, we determined that PhoF carries out the O₂-dependent degradation of two aminophosphonates, demonstrating an expanded catalytic efficiency with respect to the individual, but mechanistically coupled PhnY and PhnZ. Our results recognize PhoF as a new example of an HD-domain oxygenase and show that domain fusion of an organophosphonate degradation pathway may be a strategy for disease-causing fungi to acquire increased functional versatility, potentially important for their survival.

Metalloproteins of the HD-domain superfamily share a conserved core sequence motif (H···HD···D) capable of coordinating most transition metal ions.¹ Over 90% of characterized HD-domain proteins are hydrolases; however, an expanded functional repertoire has been recognized in recent years to include oxygenases (myo-inositol oxygenase (MIOX), PhnZ, and TmpB) and lyases (Ddi2).²⁻⁵ This diversification appears to be linked to the chemical nature of the metallocofactor and differences in the coordinating amino acids.¹ Nevertheless, the current diminutive list of characterized oxygenases prevents unambiguous establishment of features responsible for chemical variation, hindering functional prediction of uncharacterized members.

All known HD-domain oxygenases bind a diiron cofactor via an extended H···HD···H···H···D motif in which the iron ions are coordinated in an octahedral geometry and bridged by the aspartate of the HD dyad. Using molecular oxygen and a mixed-valent (Fe^IIFe^III) state that is unique to this class of nonheme diiron enzymes, they activate their substrates via a diferric superoxo intermediate that initiates the 4 e⁻ oxidative cleavage of C−C or C−P bonds.^3,4 MIOX is the founding member of the HD-domain oxygenase subfamily and plays a vital role in inositol metabolism by cleaving the C6−C1 bond of myo-inositol, a key factor in insulin signaling.² Two more oxygenases, PhnZ and TmpB, were later discovered to degrade organophosphonates by cleaving C−P bonds.^3,4

Organophosphonates have both medicinal and agricultural applications due to their antimicrobial and antifungal properties.^6–8 They serve as a phosphorus source for several bacterial and fungal species in phosphate-depleted environments such as marine ecosystems and some soil conditions.^9,10 In conjunction with the Fe^II/α-ketoglutarate (α-KG) oxygenase PhnY, a member of the phytanoyl-CoA dioxygenase (PhyH) family, PhnZ catabolizes the most abundant naturally occurring organophosphonate, 2-aminoethylphosphonate (2AEP), via the third recognized pathway of C−P bond degradation in bacteria.^3,11 In this oxidative pathway, 2AEP is first hydroxylated by PhnY forming (R)-2-amino-1-hydroxyethylphosphonate (OHAEP), followed by cleavage of the C−P bond by PhnZ.^3,11,12 A similar two-enzyme pathway composed of the PhnY and PhnZ homologues TmpA and TmpB has also been recognized in the degradation of the 2AEP analog, 2-(trimethylammonio)ethylphosphonate (TMAEP).⁴

Discovery of the PhnZ and TmpB oxygenases expands the functional repertoire of the HD-domain superfamily and suggests that other metalloenzymes may perform the O_2-dependent cleavage of chemically stable C−P (or more generally C−X) bonds. To identify HD-domain candidates with potential organophosphonate oxygenase activity and shed light into their evolutionary relationships, we carried out phylogenetic analyses and explored operon organization of the identified gene clusters. The retrieved sequences are from both bacterial and eukaryotic organisms and conserve the six residues necessary for coordinating a dinuclear cofactor. The genes encoding for these proteins are colocalized with genes predicted to encode for α-KG-dependent oxygenases such as PhnY and TmpA or proteins related to organophosphonate degradation or transport (Figure 1A).¹³

Figure 1.

Phylogenenetic analysis of putative diiron HD-domain oxygenases and structural analogies between PhoF and PhnY/Z. (A) Unrooted maximum likelihood phylogenetic tree and genomic comparison of HD-domain operons. The phylogenetic tree contains 370 proteins that are putatively associated with organophosphonate degradation operons. MIOX is used as an outgroup and the horizontal line corresponds to the number of substitutions per site. The tree was computed with the IQ-tree software¹⁴ and includes bootstrap values representing node confidence. Genomic neighborhoods were generated via the EFI-Genome Neighborhood Tool.¹³ Clade I (purple) contains proteins encoded in tandem with proteins homologous to γbb hydroxylases and includes TmpB (dark purple). Clade II (yellow) collects bacterial proteins encoded in operons related to organophosphonate degradation. The small subclade within Clade II (brown) includes sequences similar to the PhnZ1 sequence from Gimesia maris.¹⁵ The genomic structure of this subclade resembles that of the large Clade II, yet neighboring genes have not been characterized. Clade III (pink) includes PhnZ and proteins for which the respective genes are genomically colocalized with members of the PhyH family, like PhnY. The dark pink subclade comprises fungal proteins encoded by a gene fusion of PhnY-like and PhnZ-like genes. Red branches indicate sequences that have been characterized here or in other studies.^3,4,15 (B) Sequence alignment between PhnY, PhnZ, and PhoF. Conserved residues between PhoF and PhnY are highlighted in green, while conserved residues between PhoF and PhnZ are highlighted in pink. (C) Predicted secondary structure overlap of PhoF domains with PhnY and PhnZ. PhoF (blue) and PhnY (green) structures were predicted using AlphaFold2.¹⁶ The predicted PhoF structure was overlaid with the predicted PhnY structure and PhnZ crystal structure (PDB 4MLM).

The phylogenetic tree segregates into three major clades. Clade I is composed of bacterial proteins encoded alongside putative γ-butyrobetaine (γbb) hydroxylases annotated to function in the metabolic shikimate pathway (Figure 1A).⁸ TmpB belongs to a subgroup of this clade and is encoded in tandem with TmpA, an α-KG-dependent oxygenase originally misannotated as γbb hydroxylase but later demonstrated to hydroxylate TMAEP.⁴ Clade II contains proteins encoded by genes that are part of a operon cluster annotated to function in the bacterial C−P lyase pathway, suggesting a potential role in phosphonate degradation (Figure 1A).¹⁷ A subgroup of this clade contains sequences that include PhnZ1 from Gimesia (G.) maris (GmPhnZ1),¹⁵ which has high sequence homology to PhnZ and colocalizes with a PhnY-like encoding gene. While the overall genomic structure resembles that of the C−P lyase pathway, G. maris does not have a recognized phn operon (Figure S1).¹⁵

Clade III includes proteins with high sequence homology to PhnZ, the genes for which colocalize with genes encoding for proteins belonging to the PhyH protein family. A separate branch of this clade consists of proteins with sequence lengths significantly longer than those typical for HD-domain oxygenases (Figure 1A). Interestingly, the genomic structure of these proteins includes an N-terminal region reminiscent of PhyH members, followed by an HD-domain region, suggesting that these proteins are chimeric fusions of PhnY and PhnZ (Figure 1A). While the tree, except for the MIOX-containing outgroup, is dominated by bacterial sequences, the fusion proteins specifically occur in soil fungi implicated in human disease (Table S2).^18,19 While there is currently no characterized fungal organophosphonate degradation pathway, some fungi can use organophosphonates, including the herbicide glyphosate, for growth in nutrient depleted environments.^10,20 Considering the wide use of organophosphonate compounds as fungicides, an understanding of the mechanisms by which pathogenic fungi utilize organophosphonates for survival and acquire resistance to herbicides is of growing importance.²⁰⁻²²

To date, we have identified 41 fungal fusion proteins. We selected the protein, herein designated PhoF, from the soil fungus Fonsecaea (F.) multimorphosa, which is implicated in skin diseases such as chromoblastomycosis.^19,23 Because gene fusions are an important means of protein evolution, we sought to determine the functional range and potential evolutionary advantages of this fused architecture. The N-terminal domain of PhoF (residues 1−260) shares 54% similarity to PhnY and contains the conserved histidine-carboxylate-histidine (i.e., His-Asp-His) motif essential for coordination of a mononuclear Fe site (Figure 1B).²⁴ Employing α-KG and molecular oxygen as cosubstrates, PhnY affords hydroxylation of the phosphonate α carbon via a common intermediate of these hydroxylases, i.e., an Fe^IV-oxo species.²⁵ The C-terminal HD-domain of PhoF (residues 280−485) shares 53% similarity to PhnZ and conserves the six residues for coordination of a diiron site (Figure 1B). In contrast to known nonheme oxygenases and oxidases, HD-domain oxygenases react with O₂ from the Fe^IIFe^III manifold to afford the 4 e⁻ oxidation of their substrates. This mechanism involves O₂ binding to produce an Fe^IIIFe^III-superoxo complex, the pendant oxygen of which initiates hydrogen atom abstraction leading to cleavage of the C−X bond.²⁶

The N- and C-terminal domains of the AlphaFold predicted structure of PhoF exhibit significant overlap with the AlphaFold predicted PhnY structure and the PhnZ crystal structure, respectively (Figure 1C).^16,27 The HD-domain residues overlay well with those of PhnZ (Figure 2A), suggesting that PhoF similarly binds a diiron site. PhoF copurifies with 2.2 ± 0.4 equiv of Fe and has a purple color, indicating the presence of a diiron-tyrosinate complex like that observed in PhnZ and purple acid phosphatases.²⁷⁻²⁹ The 4.2-K/78-mT Mössbauer spectrum exhibits a quadrupole doublet with an isomer shift (δ) of 0.49 mm/s and a quadrupole splitting (ΔE_Q) of 0.78 mm/s, characteristic of an antiferromagnetically (AF)-coupled Fe^IIIFe^III cluster (S = 0) (Figure 2B). There are no detectable mononuclear Fe^III (S = 5/2) forms present (lack of any paramagnetic baseline), indicating that >95% of the Fe in the sample is assembled in a diiron form and that the PhnY-like domain of PhoF is mostly (if not completely) vacant. Substitution of histidine-117 of the His-Asp-His motif results in a 4.2-K/78-mT Mössbauer spectrum identical to that of WT, confirming successful assembly of the diiron cofactor (Figure 2B).

Figure 2.

Spectroscopic characterization of PhoF. (A) Structural comparison between the PhnZ crystal structure and the AlphaFold PhoF predicted structure. The PhoF residues (blue) align well with those of PhnZ (PDB 4MLM) (pink). (B) Low temperature Mössbauer spectra of the aerobically isolated WT and H117A PhoF. Spectra were recorded at 4.2 K in the presence of a magnetic field (78 mT) applied parallel to the direction of the γ-beam. Experimental spectra are shown as vertical bars, fits are shown as solid lines. (C) EPR spectra of WT and variant PhoF at 10 K. The spectra of ascorbate-reduced WT (top) and H117A (middle) show features characteristic of an Fe^IIFe^III center, while D335A exhibits no signals. (D) Mössbauer spectra of WT PhoF under different redox conditions. Spectra of aerobically isolated (top), dithionite reduced (middle), and ascorbate reduced (bottom) PhoF were recorded at 80 K in the absence of a magnetic field. The ascorbate reduced spectrum consists of four overlapping quadrupole doublets corresponding 10% Fe^IIIFe^III (purple), 7% Fe^IIFe^II (orange), and 83% Fe^IIFe^III (green) states. The Fe^II and Fe^III sites of the mixed-valent cofactor have equal integrated intensities and are shown in light and dark green.

Reduction of PhoF with sodium ascorbate yields an electron paramagnetic resonance (EPR) spectrum with an almost axial signal with g_av < 2, characteristic of an AF-coupled mixed-valent Fe^IIFe^III state (S = 1/2) (Figure 2C).^3,4,30,31 This signal is identical to that of PhnZ with principal g-values of 1.93 and 1.81.³ The EPR spectrum of the ascorbate-reduced H117A variant is essentially indistinguishable from that of the WT PhoF, demonstrating that this substitution does not influence the diiron site electronic properties (Figure 2C). In contrast, D335A copurifies with 0.37 ± 0.2 equiv of Fe and yields an EPR-silent spectrum, confirming lack of dimetal site assembly. H117A and D335A variants therefore represent PhoF constructs with cofactors similar to PhnZ and PhnY, respectively.

The high temperature 80-K/0-T Mössbauer spectrum of WT PhoF consists of a single quadrupole doublet with δ = 0.48 mm/s and ΔE_Q = 0.73 mm/s (Figure 2D, top). The slight differences in the Mössbauer parameters are a result of their well-known temperature dependence due to second-order effects. Reduction with sodium dithionite yields a spectrum dominated by a single quadrupole doublet with δ = 1.25 mm/s and ΔE_Q = 3.20 mm/s, demonstrating complete reduction to the Fe^IIFe^II state (Figure 2D, middle). Reaction with the milder reducing agent sodium ascorbate results in an 80-K/0-T Mössbauer spectrum with overlapping quadrupole doublets corresponding to all three accessible redox states of the diiron cluster (Figure 2D, bottom). The mixed-valent Fe^IIFe^III state accumulates to 83% and can be fit with two quadrupole doublets, one for the Fe^II site with δ = 1.27 mm/s and ΔE_Q = 2.49 mm/s and the other for the Fe^III site with δ = 0.48 mm/s and ΔE_Q = 1.06 mm/s (Figure 2D, bottom). The remaining spectral intensity is attributed to the nonreduced Fe^IIIFe^III state (10%) and the fully reduced Fe^IIFe^II state (7%). The high extent of accumulation of the mixed-valent Fe^IIFe^III form is typical for HD-domain oxygenases, which commonly accumulates to ∼60%, because this is the active cofactor form.^3,4,11

The high sequence homology between the two PhoF domains and the PhnY/Z enzymes suggests that PhoF may also catalyze the O₂-dependent, two-step degradation of 2AEP (Figure 3A). In fact, some of the fungal organisms that encode PhoF, such as Trichoderma and Aspergillus (Table S2), are able to utilize 2AEP as a sole phosphorus source for growth, yet the mechanism by which these organophosphonates are utilized has not been established.¹⁰ We monitored degradation of 2AEP to inorganic phosphate in the absence and presence of O₂ by ³¹P nuclear magnetic resonance (NMR) to demonstrate whether PhoF is functionally similar to the PhnY/Z system. When PhoF is incubated with ascorbate, α-KG, Fe^II, and 2AEP in an anaerobic chamber for 2 h, no inorganic phosphate is detected (Figure 3B). However, in the presence of O₂, complete conversion of 2AEP to phosphate is observed, demonstrating that (i) PhoF can act on 2AEP similarly to the PhnY/Z system and (ii) the reaction is O₂-dependent, establishing PhoF as a new HD-domain oxygenase.

Figure 3.

Comparison of PhoF activity to the PhnY/Z system. (A) Reaction scheme of the PhnY/Z system. (B) ³¹P NMR spectra of PhoF reacted with 2AEP in the presence and absence of oxygen. (C) ³¹P NMR spectra of H117A and D335A variants reacted with 2AEP and OHAEP. (D) Steady-state kinetics of PhoF (blue) and PhnY/Z (red) with 2AEP. (E) Steady-state kinetics of PhoF (blue) and PhnZ (red) with OHAEP. Error bars represent standard error.

We carried out the same experiment with the H117A and D335A PhoF variants to demonstrate activity of the respective domains and confirm that these constructs are functional mimics of the individual PhnY and PhnZ enzymes. H117A cannot coordinate Fe^II in the PhnY-like domain and should behave like PhnZ, while D335A lacks the HD-domain and should behave like PhnY. Reaction of D335A with 2AEP in the presence of O₂ results in accumulation of the hydroxylated product without any detectable formation of phosphate. When D335A is incubated with OHAEP no reaction occurs (Figure 3C), confirming that the N-terminal domain carries out hydroxylation of the phosphonate α carbon. Reaction of H117A with 2AEP under the same conditions yields no product (Figure 3C), yet reaction with OHAEP leads to complete conversion to free phosphate (Figure 3C). PhoF therefore follows the paradigm of the PhnY/Z system and degrades 2AEP in a stepwise reaction, in which each step can be separated in the two PhoF variants.

Gene fusion events can be advantageous as the derived multidomain protein is often more catalytically efficient due to reaction coupling.³²⁻³⁴ Thus, the chimera of phnY and phnZ genes in these fungal organisms may have been evolutionarily selected to improve activity toward 2AEP. When present in a 1:1 ratio, the PhnY/Z system turns over 2AEP with a k_obs of 0.91 ± 0.15 s⁻¹ and a K_M of 487 ± 140 μM (Figure 3D). PhoF degrades 2AEP to inorganic phosphate with a rate of 1.16 ± 0.17 s⁻¹ and a K_M of 197 ± 73 μM (Figure 3D). The ∼2.5-fold decrease in apparent K_M may be a result of an increase in the binding affinity of the hydroxylated substrate for the PhoF HD-domain and/or coupling of the reaction steps in which the hydroxylated intermediate is shuttled from the monoiron site directly to the diiron site, effectively increasing local substrate concentration. This is a common mechanism by which plants and fungi manage labile or toxic intermediates.³⁵ PhoF cleaves the C−P bond of OHAEP with a k_obs of 9 ± 2 s⁻¹ and a K_M of 43 ± 23 μM, while PhnZ performs the reaction with a k_obs of 15 ± 3 s⁻¹ and a K_M of 107 ± 29 μM. Because the overall catalytic rate of 2AEP to phosphate is much smaller than that of the second step of the reaction, 2AEP hydroxylation appears to be the rate-limiting step. This observation may explain the incomplete conversion of 2AEP to OHAEP by D335A (Figure 3C). PhoF is conservatively ∼3-fold more efficient with 2AEP than the dual PhnY/Z system, demonstrating that the fused architecture may provide a small fitness in catalytic efficiency.

Whereas the only recognized organophosphonate substrate of the PhnY/Z system is 2AEP, we sought to determine whether PhoF may exhibit enhanced substrate processivity. We screened a series of organophosphonates as potential substrates that vary in chain length, position or absence of the amino group, or overall structure (Figure 4A). PhoF can efficiently process 2AEP as well as 3-aminopropylphosphonate (3APP), but degradation was undetectable for aminomethylphosphonic acid (AMPA), 4-aminobutylphosphonate (4ABP), propyl-aminoethylphosphonate (propyl-AEP), TMAEP, and glyphosate (Figure 4B). We proceeded to interrogate ability of hydroxylated phosphonates to bind to the diiron site (Figure 4C). As expected, nonhydroxylated substrates are unable to bind to the HD-domain (Figure S2). Incubation with OHAEP, OHAPP, and OHTMAEP resulted in perturbations in the EPR spectra relative to the substrate-free enzyme, suggestive of coordination to the Fe^IIFe^III site and of a binding promiscuity. However, unlike the two-enzyme system in which PhnZ exhibits substrate promiscuity and PhnY is substrate specific, PhoF is unable to process OHTMAEP (Figure 4D), demonstrating that the HD-domain can also be capable of reaction selectivity.⁴

Figure 4.

Substrate screening for PhoF. (A) Phosphonates tested for activity with PhoF. (B) ³¹P NMR of PhoF reacted with various possible substrates. (C) CW EPR spectra of the ascorbate reduced WT PhoF (200 μM diiron) reacted with molar excess (2.5 mM) of the hydroxylated OHAEP, OHAPP, and OHTMAEP compounds recorded at 10 K. (D) Activity of PhoF and PhnZ toward hydroxylated substrates. Reactions were performed in triplicate with PhoF (blue) or PhnZ (red) and OHAEP (solid) or OHTMAEP (dotted). (E) Steady-state kinetics of PhoF (blue) and PhnZ (red) with 3APP as a substrate. Error bars represent standard error.

PhoF completely converts 3APP to phosphate with a catalytic rate of 0.57 ± 0.30 s⁻¹ and a K_M of 305 ± 263 μM, an activity not previously reported for the PhnY/Z system (Figure 4E). While the TmpA/B system degrades 3APP, it does so inefficiently.⁴ We find that the PhnY/Z system is also able to act on 3APP, but with a poor catalytic efficiency of 0.14 mM⁻¹ s⁻¹ (k_obs = 0.28 ± 0.16 s⁻¹) (Figure 4D), establishing 3APP as a unique second substrate for PhoF and a new substrate for oxygenases. Our findings show that the fungal fusion PhoF has a larger substrate repertoire with respect to the two-protein bacterial system.

The extended substrate capabilities of PhoF for 3APP evoke the possibility that these pathogenic soil fungi may have evolved to utilize other substrates, such as organophosphonate fungicides, not tested in this work. Our work identifies PhoF as a new example of an HD-domain oxygenase, which together with an α-KG-dependent monoiron hydroxylase domain utilizes an unprecedently stable Fe^IIFe^III cofactor to degrade 2AEP and 3APP. Our study places PhoF at the epicenter of a newly identified oxidative organophosphonate degradation pathway specific to disease-causing fungi, for which an expanded substrate pool of the PhoF fusion may provide functional fidelity relative to multiprotein pathways in bacteria. Furthermore, our phylogenetic analysis identifies other operon structures likely harboring new HD-domain oxygenases, some of which may act on organophosphonates, setting the stage for the discovery of new small molecule catabolic pathways by this enzyme subclass.