Deciphering Evolutionary Trajectories of Lactate Dehydrogenases Provides New Insights into Allostery

Abstract

Lactate dehydrogenase (LDH, EC.1.1.127) is an important enzyme engaged in the anaerobic metabolism of cells, catalyzing the conversion of pyruvate to lactate and NADH to NAD⁺. LDH is a relevant enzyme to investigate structure–function relationships. The present work provides the missing link in our understanding of the evolution of LDHs. This allows to explain (i) the various evolutionary origins of LDHs in eukaryotic cells and their further diversification and (ii) subtle phenotypic modifications with respect to their regulation capacity. We identified a group of cyanobacterial LDHs displaying eukaryotic-like LDH sequence features. The biochemical and structural characterization of Cyanobacterium aponinum LDH, taken as representative, unexpectedly revealed that it displays homotropic and heterotropic activation, typical of an allosteric enzyme, whereas it harbors a long N-terminal extension, a structural feature considered responsible for the lack of allosteric capacity in eukaryotic LDHs. Its crystallographic structure was solved in 2 different configurations typical of the R-active and T-inactive states encountered in allosteric LDHs. Structural comparisons coupled with our evolutionary analyses helped to identify 2 amino acid positions that could have had a major role in the attenuation and extinction of the allosteric activation in eukaryotic LDHs rather than the presence of the N-terminal extension. We tested this hypothesis by site-directed mutagenesis. The resulting C. aponinum LDH mutants displayed reduced allosteric capacity mimicking those encountered in plants and human LDHs. This study provides a new evolutionary scenario of LDHs that unifies descriptions of regulatory properties with structural and mutational patterns of these important enzymes.

Introduction

Lactate dehydrogenases (LDHs, EC.1.1.127) are critical enzymes involved in the anaerobic metabolism of cells. They are mainly present in eukaryotes and bacteria, while they are rare in archaea. LDHs from different species were purified and characterized (Brochier-Armanet and Madern 2021, and references therein). They operate in the last step of glycolysis by catalyzing the reversible chemical transformation of pyruvate into lactate with NADH as coenzyme (Everse and Kaplan 1973; Holbrook et al. 1975; Fersht 1985). From a functional point of view, LDHs allow the regeneration of NAD⁺ and thus sustain the glucose catabolism of cells under conditions of limited oxygen concentration (Everse and Kaplan 1973; Holbrook et al. 1975; Fersht 1985). The catalytic mechanism of LDHs has been extensively studied. When the competent catalytic state is reached, LDHs catalyze the direct transfer of a hydride ion from the pro-R face of NADH to the C2 carbon of pyruvate to produce lactate (Burgner and Ray 1984; Clarke et al. 1986; Clarke et al. 1988; van Beek et al. 1997; McClendon et al. 2005; Deng et al. 2011; Callender and Dyer 2015; Egawa et al. 2019). This reaction is controlled by a rate-limiting step due to the closing of a mobile loop (MbL), which covers the catalytic vacuole (Clarke et al. 1985; Pineda et al. 2007). The MbL carries also a glutamine residue at position 102 (Q102) that plays a key role in substrate recognition and specificity (Wilks et al. 1988; Cendrin et al. 1993; Katava et al. 2020).

LDHs are homotetrameric enzymes with the 4 subunits related by 3 molecular 2-fold axes named P, Q, and R (Rossmann et al. 1973). Crystal structures show that tetrameric LDHs have 4 active sites. The active site of each subunit lies near the Q-axis interface and involves mainly H68, Q102, R109, D168, R171, T246, and I250 residues. To date, canonical eukaryotic and bacterial LDHs crystal structures present a very similar fold with the exception of a 25 amino acid N-terminal extension only presents in vertebrate's enzymes (Clarke et al. 1989, Piontek et al. 1990; Iwata, Kamata, et al. 1994; Iwata, Yoshida, and Ohta 1994; Auerbach et al. 1998; Read et al. 2001, Chaikuad et al. 2005; Coquelle et al. 2007; Swiderek et al. 2009; Ikehara et al. 2014; Matoba et al. 2014; Kolappan et al. 2015; Friberg et al. 2020, Iorio et al. 2021).

Phylogenetic and biochemical studies have revealed that LDHs belong to a large superfamily of 2-hydroxy acid dehydrogenases that includes also malate dehydrogenases (MDHs) (Madern 2002; Madern et al. 2004; Boucher et al. 2014). While LDHs use pyruvate as substrate, MDHs convert oxaloacetate (OAA) into malate. Most recent studies indicate that the capacity to convert pyruvate into lactate emerged from MDHs several times independently (Madern 2002; Madern et al. 2004; Boucher et al. 2014; Steindel et al. 2016; Brochier-Armanet and Madern 2021): 1 event led to the large group of LDHs found in bacteria and in most eukaryotes that are characterized by the presence of a conserved glutamine at position 102 (referred thereafter as to canonical LDH), while 4 events led independently to the emergence of LDH in Plasmodium (Eucarya and Apicomplexa), Cryptosporidium (Eucarya and Apicomplexa), Trichomonas vaginalis (Eucarya and Parabasalia), and Selenomonas ruminantium (Bacteria and Firmicutes) (see (Brochier-Armanet and Madern 2021, and references therein). Conversely, a single case of LDH toward MDH activity has been recently documented in Planctopirus limnophila (Bacteria and Planctomycetes) (Brochier-Armanet and Madern 2021).

In vertebrates, tetrameric LDHs are encoded by 3 homologous genes, which are expressed in different tissues: LDH-A [encoding a muscle type-specific (M) isozyme], LDH-B [encoding a heart type-specific (H) isozyme], and LDH-C [encoding a testis-specific (C) isozyme] (Goto et al. 2016). The phylogenetic relationships between vertebrate LDHs and more generally between metazoan LDHs are not well resolved. However, early studies suggest that they derive from bacterial sequences (Tsoi and Li 1994; Tsuji et al. 1994; Stock et al. 1997). Henceforward, the number of available sequences has considerably grown allowing to reinvestigate the evolutionary relationships between eukaryotes LDHs, including those from non-metazoan eukaryotes (e.g. plants and fungi), and bacterial LDHs.

Numerous studies have shown that LDHs are relevant model enzymes to study allosteric regulation (Garvie 1980; Brochier-Armanet and Madern 2021, and references therein). In fact, most of the characterized bacterial LDHs are typical allosteric enzymes, for which the allosteric effector is fructose 1,6-bisphopshate (FBP). In contrast, a single case of non-allosteric bacterial LDH, which is constitutively activated without FBP, has been identified and documented in Lactobacillus pentosus (Uchikoba et al. 2002). In the absence of FBP, bacterial allosteric LDHs exhibit a sigmoid pyruvate saturation profiles with a complete lack of activity at physiological low concentration of pyruvate (Arai et al. 2002; Iorio et al. 2021). This corresponds to a typical homotropic activation phenomenon. When the enzymatic reaction proceeds in presence of FBP, the activity profile turns hyperbolic, demonstrating a heterotropic allosteric activation (Schroeder et al. 1988; Arai et al. 2002; Feldman-Salit et al. 2013; Taguchi 2017). Crystal structures have revealed that residues R173, H188, and Y190 [referred to as the FBP-binding site (FBP-BS) signature sequence thereafter] located at the P-axis-related interface participate in the binding of FBP in allosteric LDHs (Iwata and Ohta 1993; Iwata, Kamata, et al. 1994; Iwata, Yoshida, and Ohta 1994; Coquelle et al. 2007).

Allosteric transition of bacterial LDHs fits well the ensemble model of allostery (Motlagh et al. 2014; Nussinov et al. 2014; Guo and Zhou 2016), as well as the classical concerted Monod–Wyman–Changeux (MWC) model, in which the T-inactive and R-active states of enzymes coexist in a pre-equilibrium independently of allosteric effectors (Monod et al. 1965).

Canonical eukaryotic LDHs differ from their bacterial homologs by several aspects. While all the bacterial LDHs occur as homotetramers, vertebrate LDH may assemble in different ways. In particular, LDH-M and LDH-H can form homotetrameric or heterotetrameric assemblies, whereas LDH-C can form exclusively homotetramers (Sakai et al. 1987; Read et al. 2001). M and H forms display a marked difference in their respective affinity for pyruvate and their sensitivity to inhibition by high concentration of this substrate (Dawson et al. 1964). The formation of M/H LDH heterotetramers is thought to play a role in the development of vertebrate embryo (Cahn et al. 1962; Goto et al. 2016). Enzymatic properties of the various forms of vertebrate LDHs indicate an absence of cooperativity between subunits (Pesce et al. 1967; Everse and Kaplan 1973; LeVan and Goldberg 1991; Holland et al. 1997). Compared to bacteria, structures from vertebrates LDHs have revealed that the N-terminal extension created additional interactions between subunits (Kolappan et al. 2015 and references therein), and the deletion of this extension in Homo sapiens LDH (H. sapi LDH) led to a strongly unstable enzyme (Zheng et al. 2004 ). This structural feature was suggested to be responsible for the absence of allosteric capacity, allowing vertebrate LDHs to continuously exist in an R-active state (Kolappan et al. 2015). However, some studies have demonstrated that such an assumption is no longer valid. In fact, using biophysical and molecular dynamic simulations, it has been shown that the LDH-M from the rabbit displays some reminiscent allosteric properties (Katava et al. 2017). More recently, it has been shown that rabbit LDH-M and human LDH-M undergo allosteric transition under mildly acidic conditions (Iacovino et al. 2022; Pasti et al. 2022).

Regarding plants, LDH biochemical studies are rare. However, they display a different behavior compared to vertebrate LDHs. By showing that their enzymatic properties do not follow the Michaelis–Menten kinetic, they are considered as homotropically activated enzymes (Betsche 1981; Tihanyi et al. 1989; O'Carra and Mulcahy 1996; Sugiyama and Taniguchi 1997). Sequence comparison indicates that plant LDHs also exhibit a long N-terminal extension (Hondred and Hanson 1990) without structural data to date. Data concerning fungi are even rarer. The LDH from Phycomyces blakesleeanus (P. blak LDH) was shown to harbor both the homotropic and heterotropic allosteric activation as bacterial LDH (De Arriaga et al. 1982; Soler et al. 1982). Primary sequence inspection of fungi sequences indicates that the N-terminal extension is absent. As in the case of plants, no structural data are available regarding fungi LDH.

Here, we investigate the links between canonical LDHs from eukaryotes and their bacterial counterparts using an integrated approach. An in-depth phylogenetic analysis reveals that in eukaryotes, LDHs have been acquired via 2 independent horizontal gene transfers (HGTs) from bacteria: vertebrates and plants LDHs have been acquired from cyanobacteria, while fungi sequences have been acquired from another bacterial donor. The biochemical and structural characterization of the Cyanobacterium aponinum LDH (C. apon, Cyanobacteria) and the kinetic characterization of the Rosa chinensis LDH (R. chin, plants) reveal that C. apon LDH displays an unexpected mix of bacterial and eukaryotic LDH properties, while the R. chin LDH is representative of a strict homotropically activated enzyme. In addition, we characterized a set of C. apon LDH mutants designed to specifically alter the allosteric behavior of the resulting enzymes.

This work reveals the scenario by which allosteric regulation capacity in LDHs evolved over the bacterial and eukaryotic domains.

Materials and Methods

Phylogenetic Analyses

A recent in-depth analysis of 16,052 reference proteomes from UniProt (https://www.uniprot.org/proteomes/) resolved the relationships between MDH and LDH (Brochier-Armanet and Madern 2021). Starting from this study, we retrieved the 484 canonical LDH sequences identified by Brochier-Armanet and Madern (2021) in a representative and nonredundant sampling of 2,272 proteomes (266 eukaryotes, 269 archaea, and 1,737 bacterial proteomes). We also included the proteome and the LDH sequence of the plant R. chin that is part of the 16,052 reference proteomes but not to be part of the 2,272 proteomes retained by Brochier-Armanet and Madern (2021). So, in total, we considered 2,373 proteomes. We also included 4 bacterial MDH used as outgroup: the sequence from the alphaproteobacterium Rhodospirillum centenum ATCC 51521 (uniprot_id B6IYP5), Chloroflexi Chloroflexus aurantiacus ATCC 29366 (uniprot_id P80040), the Chlorobi Chlorobaculum tepidum ATCC 49652 (uniprot_id P80039), and the Bacteroidetes Salinibacter ruber DSM 13855 (uniprot_id Q2S289). The 489 sequences have been aligned with MAFFT v7.453 (Katoh and Standley 2013) using the accurate L-INS-i option. The resulting alignment has been trimmed using BMGE v1.2 (Criscuolo and Gribaldo 2010) with the BLOSUM30 substitution matrix. A maximum likelihood (ML) phylogenetic tree has been inferred with IQ-TREE v1.6.12 (Nguyen et al. 2015). The LG + G4 model was identified by ModelFinder (Kalyaanamoorthy et al. 2017) as the most suitable for the tree reconstruction. Branch supports have been estimated using the ultrafast bootstrap procedure implemented in IQ-TREE (1,000 replicates).

Tree Drawing, Residue Mapping, and Heatmap

Tree figures were drawn with iTOL v5 (Letunic and Bork 2021).

Protein Expression and Purification

The C. apon, R. chin LDHs genes and various mutants were purchased (GENECUST) and cloned in a pET 20a plasmid for overexpression. A 6 histidine extension was encoded at the C-terminal part of the resulting constructs. Escherichia coli BL21 (DE3) strain transformed with the LDH plasmid was grown in Luria Bertani (LB) medium with ampicillin (100 μg/mL) at 37 °C until OD (600 nm) of 0.5. After a cold shock (4 °C for 5 h), Isopropyl β-D-1-thiogalactopyranoside (IPRG) was added to a final concentration of 0.2 mM to induce expression, and the culture was incubated at 20 °C overnight. Cells were harvested by centrifugation at 5,000 × g for 30 min at 4 °C. The pellet was suspended in 40 mL of buffer 1 (50 mM Tris–HCl pH 7.0 and 50 mM NaCl). Prior to cell disruption, 5 µg/mL of DNAse (Roche), 10 mM MgCl₂, protease inhibitors (Roche), and lysozyme (Roche) were added. The preparation cooled at 4 °C was disrupted by sonication (Branson). The crude extract was then centrifugated at 13,000 × g for 30 min at 4 °C, filtered and applied on Nickel affinity column (HiTrap HP 5 mL—GE healthcare), and equilibrated in buffer 1. The column was washed with 15 mL of 50 mM Tris–HCl pH 7, 200 mM of NaCl, and 15 mL of 20 mM imidazole. The protein was eluted with buffer 1 complemented with 300 mM imidazole. The fraction was diluted 3 times before being applied to a cofactor affinity column (Blue Sepharose 5 mL, GE Healthcare) equilibrated with buffer 1. A gradient of NaCl (50 mM to 2.5 M) was applied to elute the protein. Fractions containing the purified protein were then concentrated in buffer 2 (50 mM bis-Tris propane pH 6.1 and 50 mM NaCl) before a final size exclusion chromatography (SEC) step (S200 10/300 increase equilibrated with buffer 2). The pure active fractions in buffer 2 were pooled and concentrated to 20 mg/mL and stored at 4 °C. We found buffer 2 was more efficient than buffer 1 for long-term stability storage at 4 °C.

SEC-MALLS

SEC combined with online detection by multiangle laser light scattering (MALLS) and refractometry (RI) was used to measure the absolute molecular mass of proteins in solution. The SEC run was performed using an ENrich SEC650 10 × 300 gel-filtration column (Biorad) equilibrated with buffer 1. Separation was performed at room temperature. Fifty microliters of the protein stock solution diluted at ∼5 mg/mL with buffer 1 was injected with a constant flow rate of 0.5 mL/min. Online MALLS detection was performed with a DAWN-HELEOS II detector (Wyatt Technology Corp.) using a laser emitting at 690 nm. Protein concentration was determined by measuring the differential refractive index online using an Optilab T-rEX detector (Wyatt Technology Corp.) with a refractive index increment dn/dc of 0.185 mL/g. Weight-averaged molecular weight (Mw) determination was done with the ASTRA6 software (Wyatt Technologies), and curve was represented with GraphPad Prism software.

Analytical Ultracentrifugation

Ultracentrifugation experiments were conducted in an XLI analytical ultracentrifuge (Beckman, Palo Alto, CA, USA) using an ANTi-50 rotor, using double channel Epon centerpieces (Beckman, Palo Alto, CA. USA) of 12-mm optical path length equipped with sapphire windows, with the reference channel being typically filled with the solvent of the sample. Acquisitions were done at 20 °C and at 42,000 rpm (130,000 × g), overnight, using absorbance (280 nm) and interference detection. Data processing and analysis were done using the program SEDFIT and GUSSI using standard equations and protocols (Le Roy et al. 2015).

Standard Enzymatic Assays

LDH activity was assessed by measuring the initial rates of PYR reduction (NADH oxidation) at 340 nm in a thermostated spectrophotometer from JASCO. The standard assay mixture contained 50 mM Tris–HCl pH 7.0, 50 mM NaCl, 0.3 mM NADH, and various concentration of substrate in a final volume of 0.6 mL. The reaction was initiated by addition of the enzyme. LDH assays were carried out at 35 °C. One unit of LDH activity corresponds to the amount of enzyme that catalyzes the oxidation of 1 µmol of NADH per min. The data were analyzed using GraphPad Prism V6 using the Michaelis–Menten or allosteric sigmoidal option. The substrate saturation profiles were normalized by the maximal velocity values obtained for each enzyme. With homotropically activated LDH, the maximal value considered was in the presence of FBP effector.

Crystallization of C. apon LDH

Initial crystallization screening was performed at HTX-lab (EMBL, Grenoble, https://htxlab.embl.org) as sitting drops in 6 standard screens (The Classics Suite and The Pegs Ions from Qiagen, The JCSG+ and the PACT from Molecular Dimensions, Wizard I & II from Rigaku and Salt-Grid derived from Hampton) at 20 °C in Crystal Direct plates. Drops were inspected and scored at different time points for hits over 35 d. Apo and Holo C. apon LDHs were crystallized in the presence of Crystallophore (TbXo4, https://crystallophore.fr/).

For Apo C. apon LDH crystallization, a TbXo4/C. apon LDH mixture (prepared by dissolving TbXo4 powder with C. apon LDH at 15 mg/mL for a final TbXo4 concentration of 10 mM) was prepared 2 h before setting up the crystallization experiment (100 nL of protein sample and then 100 nL of crystallization solution). Crystals were manually reproduced in 24-well plates (Molecular Dimensions) with the condition 17% Peg MME 550, 0.1 M bicine pH 9.0, and 0.1 M NaCl. Hanging drops were set up by adding 1.5 µL of TbXo4/C. apon LDH mixture plus 1.5 µL of reservoir solution. Crystals appeared within a few days at 20 °C. Prior to data collection, crystals were cryoprotected with a higher Peg MME 550 concentration (22%) and flash frozen in liquid nitrogen.

The C. apon LDH ternary complex was prepared by mixing oxamate (to substitute the LDH substrate), FBP, NADH, and C. apon LDH at a final concentration of 2 mM for the ligands and 12 mg/mL for the enzyme. The drops were set up using the Xo4-standard protocol as implemented at HTX-lab (addition of 100 nL of protein sample and then 100 nL of TbXo4 at 10 mM in 10 mM sodium bicarbonate and then 100 nL of crystallization solution). Crystals were obtained in Crystal Direct plates with the condition 0.2 M Na malonate dibasic monohydrate, 20% Peg3350. Crystals appeared overnight at 20 °C. Crystals were automatically harvested with the Crystal Direct harvester and flash frozen at HTX-lab after cryoprotection with glycerol (10% final concentration).

Apo C. apon LDH Structure

Diffraction data were collected on the Proxima-1 beamline at Soleil synchrotron at the selenium edge (0.984 Å). Data were treated anomalously with xds (Kabsch 2010). Structure phasing was performed with Crank-2 (ccp4 suite, Winn et al. 2011 ) using the anomalous signal from 3-Tb sites. Subsequent refinements were done with coot (Emsley et al. 2010) and refmac (ccp4 suite) images prepared with Pymol (https://pymol.org/2/).

The asymmetric unit contains 1 C. apon LDH molecule, 1 TbXo4 molecule, and 3 terbium atoms. The relevant biological tetramer is built through crystal symmetry. The TbXo4 molecule is coordinated to E62 and forms further interactions via its picolinate moiety with W218 of a neighboring tetramer.

C. apon LDH Ternary Complex Structure

Diffraction data were collected on the Massif-1 beamline at ESRF synchrotron. The structure was solved by molecular replacement using the H. sapi LDH-M structure (PDBID 4OJN) as a model. Subsequent refinements were done with coot and buster (https://www.globalphasing.com/). Structure representations were prepared with Pymol. The asymmetric unit contains 1 C. apon LDH tetramer, in which 4 oxamate, 2 FBP, and 4 NADH molecules are bound. No anomalous signal was found indicating the absence of TbXo4 in this structure.

Crystallographic software support was provided by SBGrid (Morin et al. 2013). Data collection and refinement statistics are in Supplementary Table S1.

Results

A New Picture of Eukaryotic LDHs Evolution

Compared to bacterial enzymes, eukaryotic LDHs display different biochemical and structural properties. Our goal was to understand how and when these properties emerged, with a particular emphasis on the allosteric capacity. As a first step, we sought to clarify their evolutionary history and relationships with prokaryotic LDH. An in-depth survey of 2,273 proteomes representative of UniProt reference proteomes led to the identification of 485 canonical LDH sequences present in 438 (19.3%) proteomes (Supplementary Table S2). The ML phylogeny of the 485 LDH sequences is shown as Fig. 1. Mapping experimental data from this work and literature (Table 1) on this tree provides new information that is discussed below. Most of the LDH sequences are present in bacteria [375 sequences in 357 of the 1,737 (20.6%) bacterial proteomes] and in eukaryotes [105 sequences in 77 of the 267 (28.8%) eukaryotic proteomes], while they are rare in archaea [5 sequences in 4 of the 269 (1.5%) archaeal proteomes].

Fig. 1.

Rooted ML tree of the 485 LDH sequences identified in 2,273 proteomes representative of the taxonomic diversity of Archaea, Bacteria, and Eukarya for which complete proteome sequences are available. MDH sequences used to root the tree are in black. Archaeal LDHs are in yellow, bacterial LDH in pink (with cyanobacterial sequences in dark pink), and eukaryotic sequences in blue. Green triangles designate characterized LDH sequences according to this study or literature (see also Table 1). From the innermost to the outermost circle, filled triangles correspond to (i) the presence of Q102 (gray), the critical residue involved in pyruvate recognition, (ii to iv) 3 major residues involved in FBP binding (R173, H188, and Y190, in blue), and (v and vi) 2 histidine that participate to the LDH tetrameric assembly (H183 and H218, olive), while empty triangles indicate the presence of other residues. The presence of the N-terminal extension is shown by gray rectangles of various sizes. The scale bar corresponds to the average number of substitutions per amino acid site in the sequences. Gray circles represent the robustness of branches (ultrafast bootstrap, 1,000 replicates). For clarity, only values > 90% are shown. A larger picture of this tree is shown as Supplementary Fig. S1.

Table 1.

Functional information and properties of 16 enzymes characterized in this work or from the literature

	Organism (taxonomy)	Code	N-term		Allostery		FBP-BS Signature
					Homo	Hetero	R173	H188	Y190	UniProt Id
1	Chloroflexus aurantiacus (Bacteria, Chloroflexi)	C. aura	−	M	−	−	R	N	C	P80040
2	Planctopirus limnophila (Bacteria, Planctomycetes)	P. limn	−	M	−	−	R	D	T	D5SXK9
3	Hungateiclostridium thermocellum (Bacteria, Firmicutes) (Bacteria, Firmicutes)	H. ther	−	L	+	+	R	H	Y	Q8KQC4
4	Moorella thermoacetica (Bacteria, Firmicutes)	M. ther	−	L	+	+	R	H	Y	Q2RHG3
5	Lactobacillus pentosus (Bacteria, Firmicutes)	L. pent	−	L	−	−	R	D	Y	P56512
6	Lacticaseibacillus paracasei (Bacteria, Firmicutes)	L. para	−	L	+	+	R	H	Y	Q034V0
7	Enterococcus mundtii (Bacteria, Firmicutes)	E. mund	−	L	+	+	R	H	Y	V5XPB8
8	Staphylococcus aureus (Bacteria, Firmicutes)	S. aure	−	L	−	−	R	D	Q	Q2G218
9	Phycomyces blakesleeanus (Eucarya, Fungi)	P. blak	−	L	+	+	R	H	Y	A0A167R5F4
10	Deinococcus radiodurans (Bacteria, Deinococcus-Thermus)	D. radi	−	L	+	+	R	H	Y	P50933
11	Thermus thermophilus (Bacteria, Deinococcus-Thermus)	T. ther	−	L	+	+	R	H	Y	Q5SJA1
12	Bifidobacterium longum (Bacteria, Actinobacteria)	B. long	−	L	+	+	R	H	Y	P0CW93
13	Cyanobacterium aponinum (Bacteria, Cyanobacteria)	C. apon	+	L	+	+	R	H	Y	K9Z684
14	Rosa chinensis (Eucarya, Viridiplantae)	R. chin	+	L	+	−	R	Q	Y	A0A2P6P899
15	Drosophila melanogaster (Eucarya, Metazoa)	D. mela	+	L	−	−	R	H	W	Q95028
16	Homo sapiens (Eucarya, Metazoa)	H. sapi	+	L	Low pH	−	R	H	W	Q9UDE9

The presence (+) or absence (−) of the N-terminal extension is indicated in the N-term column. Functionality of the enzymes is reported as M for MDH or L for LDH. The allostery columns report the existence (+) or absence (−) of homotropic activation (Homo) and heterotropic activation (Hetero). With respect to heterotropic activation: Amino acid variabilities in the canonical FBS-BS signature are shown in italic. Residue numbering matches the one from Eventoff et al. (1977), and Uniprot numbers are shown in the last column. References: (1) (Rolstad et al 1988), (2) (Brochier-Armanet and Madern 2021), (3) (Ozkan et al. 2004), (4) (Iwasaki et al. 2017), (5) (Taguchi and Ohta 1992), (6) (Arai et al. 2010), (7) (Matoba et al. 2014), (8) (Yeswanth et al. 2013), (9) (De Arriaga et al. 1982 ), (10, 11) (Coquelle et al. 2007), (12,13) this work, (14) (Karvountzi et al. 1995), (15) (Pasti et al. 2022), and (16, LDH-M) (Dempster et al. 2014). In human, the 3 LDH forms display the same signature sequence for FBP.

The very narrow distribution of LDH in archaea suggests that they have been acquired by HGT and not by vertical inheritance from the common ancestor of all archaea. Strengthening this hypothesis, archaeal LDHs are mixed with bacterial sequences and do not form a monophyletic group in the LDH phylogeny (Fig. 1; Supplementary Fig. S1), indicating independent acquisitions from different bacterial donors. For eukaryotes, a similar situation is observed, as eukaryotic sequences are not grouped together in the tree, which suggests again several independent origins. More precisely, eukaryotes LDH emerge at 3 different positions (Figs. 1 and 2). Two fungus sequences are isolated from the others and grouped with 2 archaeal and 2 bacterial sequences [bootstrap value (BV) = 96%]. Other eukaryotes sequences form 2 separated clusters. The smallest (cluster I) gathers sequences from Fungi and 1 sequence from a member of Rhizaria (Figs. 1 and 2) and robustly grouped with a mix of bacterial sequences from various phyla (BV = 97%; Figs. 1 and 2).

Fig. 2.

Focus on the eukaryotic sequences of the ML tree shown on Fig. 1. For clarity, prokaryotic sequences have been collapsed. Colors represent taxonomic groups: light pink corresponds to Alveolata, dark pink to Filasterea, green to Plantae, purple to Fungi, yellow to Metazoa group I (Spiralia and Ciona), brown to Metazoa group II (mainly Ecdysozoa and Cnidaria), and orange to Metazoa group III (Vertebrata). Green triangles designate characterized eukaryotic LDH sequences according to this study (R. chin) or to the literature (the metazoan H. sapi and the fungi P. blak). From the innermost to the outermost circle, filled triangles correspond to (i) the presence of Q102 (gray), the critical residue involved in pyruvate recognition, (ii to iv) 3 major residues involved in FBP binding R173, H188, and Y190 (blue), and (v and vi) 2 histidine that participate to the LDH tetrameric assembly (H183 and H218, olive), while empty triangles indicate the presence of other residues. The scale bar corresponds to the average number of substitutions per amino acid site in the sequences. Gray circles represent the robustness of branches (ultrafast bootstrap, 1,000 replicates). For clarity, only values > 90% are shown. A larger picture of this tree is shown as Supplementary Fig. S1.

The largest group (cluster II) contains sequences of Metazoa (including the 3 human LDHs), Plantae (represented here by Viridiplantae and Rhodophyta), 1 Alveolata, and 1 Filasterea (Fig. 2). Surprisingly, LDH relationships within cluster II show discordance with the phylogeny of eukaryotes. For instance, metazoan sequences are not monophyletic and form 3 distinct groups: group I gathers sequences from Spiralia and Ciona, group II from Ecdysozoa and Cnidaria, and group III from Vertebrata (Fig. 2). Determining whether these discrepancies are the result of tree reconstruction artifacts, lack of phylogenetic signal, gene transfers, or hidden paralogies would require dedicated analyses that are beyond the scope of this study. Cluster II is robustly nested within a clade of Cyanobacteria (BV = 100%; Fig. 1, purple sequences). The split of eukaryotic sequences in 2 separated clusters indicated clearly 2 distinct origins and likely 2 acquisitions through HGT from distinct bacterial donors. Regarding cluster II, a mitochondrial or archaeal origin, as expected according to the evolutionary history of eukaryotes (Dacks et al. 2016), can be excluded since no link to alphaproteobacterial or archaeal sequences is observed (Fig. 1). In fact, an acquisition from Cyanobacteria appears likely, since eukaryotes cluster II sequences are nested within cyanobacterial sequences. In contrast, the situation is less clear for cluster I (fungal and rhizarial sequences), because a few alphaproteobacterial LDHs branch in the vicinity of eukaryotic sequences (Fig. 1). However, these alphaproteobacterial sequences are mixed with sequences from other bacterial phyla, making the exact origin of cluster I sequences difficult to determine. The relationship between cluster II and cyanobacterial LDH sequences is unexpected because, besides endosymbiotic gene transfers linked to the chloroplast acquisition, ancient HGT from Cyanobacteria toward eukaryotes seems to be rare (Rochette et al. 2014). Yet, this relationship was likely not artifactual because it was supported by very high BV (>90%), and the presence of a cysteine at position 35 shared exclusively by cyanobacterial LDH and most of their eukaryotic relatives.

In order to link information gained from the phylogenetic analysis with experimental functionality, we investigated the properties of LDHs from 2 cyanobacteria, C. apo and Cyanobium gracile (C. grac); 1 plant, R. chin; and 1 group I metazoan, Echinococcus granulosus (E. gran). Unfortunately, we did not succeed to refold properly the recombinant C. grac and E. gran LDHs.

Allosteric Activation Capacity of C. apon LDH

We measured enzyme kinetics of recombinant C. apon and R. chin LDHs and compared the data with those published for the canonical allosteric bacterial LDH of Bifidobacterium longum (B. long, Actinobacteria), an enzyme devoid of any N-terminal extension as all bacterial LDHs, except the cyanobacterial ones, and the vertebrate nonallosteric LDH-M and LDH-H (Vesell 1965; Wuntch et al. 1970; Fushinobu et al. 1996). However, accurate comparisons with data from the literature are difficult because kinetic measures are not uniform. To overcome this issue, we present “normalized” substrate saturation profiles recorded at pH 7 in percentage of the maximal activity obtained for each enzyme of this study and from the literature. This allows to visualize and investigate (i) pyruvate affinity changes and (ii) allosteric properties from the shape of the substrate saturation profile. Indeed, a sigmoid activity profile using pyruvate as substrate indicates that the allosteric LDHs are found in an equilibrium between 2 states, a low-affinity T state and high-affinity R state in the absence of allosteric effectors. In the presence of FBP, the equilibrium is displaced toward the R state and becomes hyperbolic.

Without any effector, the B. long LDH saturation profile is typical of homotropic activation with a sigmoid shape (Fig. 3a). When the reaction mixture is supplemented with 0.1 mM of FBP, the enzyme is strongly activated with an increase in maximal activity and a strong shift of pyruvate affinity toward low values (Fig. 3a). The corresponding curves are typical of allosteric behavior, with both homotropic and heterotropic activation, as observed for bacterial LDHs devoid of N-terminal extensions, such as the LDH from Thermus species (Taguchi 2017).

Fig. 3.

Pyruvate saturation curves of 4 LDHs. Measurements were done in the presence of the indicated concentrations of pyruvate with NADH as coenzyme. a) B. long LDH, without or with FBP, left and right panels, respectively. Data are from Fushinobu et al. (1996). b) C. apon LDH, without or with FBP, left and right panels, respectively. c) R. chin LDH without FBP. d) Human muscle (M4-LDH) and heart (H4-LDH). Data are from Vesell (1965).

Then, we investigated the properties of the C. apon LDH, a cyanobacterial enzyme with a long N-terminal extension, closely related to cluster II eukaryotes LDH. Unexpectedly, the pyruvate saturation profile of C. apon LDH is also sigmoid (Fig. 3b), as in B. long (Fig. 3a), with a K_m value for pyruvate of 10 mM, despite the presence of an N-terminal extension. In the presence of FBP, the enzyme of C. apon (Fig. 3b) behaves again like that of B. long (Fig. 3a), as the maximal enzymatic activity increases and the affinity is shifted toward low concentration of substrate, with a K_m value for pyruvate of 0.1 mM. In the presence of FBP, the C. apon LDH is sensitive to inhibition by high concentration of substrate (Fig. 3b), as it is frequently encountered in LDHs (Eszes et al. 1996). When FBP is added, the enzyme turnover (k_cat) is increased from 46 to 182 s⁻¹ (Table S3). The FBP exerts therefore a strong favorable activation effect on the C. apon LDH. We continued our investigation by analyzing the R. china LDH, a plant enzyme with a long N-terminal extension, belonging to cluster II eukaryotes LDH. Here again, the R. china LDH enzymatic activity profile is sigmoid (Fig. 3c). The K_m value for pyruvate is 1.5 mM with a k_cat value of 110 s⁻¹ (Table S3). We tested whether the addition of FBP influences and found no significant effect on activity (data not shown). Therefore, the R. china LDH shows only homotropic activation. The typical profile saturation curves of H. sapi LDHs (M and H forms) display neither sigmoid activation profile nor activation by FBP (Fig. 3d), as expected for nonallosteric enzymes (Vesell 1965). Such a behavior holds also when pig and rabbit LDHs are investigated (Wuntch et al. 1970).

Our data found with R. chin and C. apon LDHs as representative enzymes indicated that plants and cyanobacterial LDHs shared a homotropic capacity; nonetheless, heterotropic activation is absent in this plant. These findings challenge the proposed role of N-terminal extensions as a structural element unfavorable for allosteric regulation.

Distribution of Allostery in LDHs

To go further, we decided to add our data to those from the existing literature. The resulting table allows to present a survey of functional and regulatory properties of 14 LDHs, 2 from this work and 12 from the literature (enzymes 3 to 16 in Table 1). When analyzed with respect to the phylogeny of LDHs, it brings new insight on the evolution of allosteric behavior of these important enzymes. We recall that in the outgroup, tetrameric MalDHs, exemplified here by C. aura and P. limn (enzymes 1 and 2 in Table 1, respectively), are nonallosteric and display hyperbolic substrate saturation profiles (Rolstad et al 1988; Brochier-Armanet and Madern 2021). The 14 LDHs are widespread across the LDH phylogeny and belong to various bacterial and eukaryotic species (Figs. 1 and 2). As expected, they recognize pyruvate as substrate and display Q102, a residue considered a strong signature for LDH functionality (see Brochier-Armanet and Madern 2021 and references therein). The distribution of properties shows that allostery is a dominant phenotype of LDHs (Figs. 1 and 2). In fact, with the exception of the LDHs from L. pent and Staphylococcus aureus (S. aure) (enzymes 5 and 8, respectively), other bacterial and the fungal LDHs (3, 4, 6, 7, and 9 to 12) are both homotropically and heterotropically controlled by pyruvate and FBP (Table 1; Figs. 1 and 2). Furthermore, we noticed that in the part of the LDH tree encompassing these enzymes, the FBP-BS signature sequence is prevailing. The crystal structure of the tetrameric LDH of L. pent (enzyme 5) has shown that the lack of allosteric regulation is due to a strong stabilizing interaction at the AB-like interfaces, which maintains the enzyme in the R-active state and prevents its ability to explore the T state (Uchikoba et al. 2002). Consistently, because there is no regulation by FBP in this enzyme, the 3 amino acids involved in FBP binding are absent. In the case of S. aure LDH (enzyme 8), to our knowledge, there is no structural information explaining its behavior.

Regarding eukaryotic LDHs, the situation is different. Available data show differences between clusters I and II that make sense with 2 distinct origins. In cluster I, the P. Blak LDH displays both homotropic and heterotropic activations (De Arriaga et al. 1982; Soler et al. 1982) and the FBP-BS signature (Table 1) but not the long N-terminal extension observed in cluster II eukaryotic LDHs. These features are shared with enzymes from Deinococcus radiodurans and Thermus thermophilus (enzymes 10 and 11) (Coquelle et al. 2007), 2 bacterial LDHs related to cluster I (Supplementary Fig. S1). In contrast, eukaryotic LDHs of cluster II and their cyanobacterial relatives (enzymes 13 to 16) harbor an N-terminal extension of 20 to 40 amino acids (Fig. 2; Supplementary Fig. S2), the longest extensions being found in plant LDHs, which display different allosteric features. In fact, we showed that C. apon (enzyme 13) is both homotropically and heterotropically regulated as other bacterial LDHs, R. chin (enzyme 14) is only homotropically regulated, while H. sapi (enzymes 16; LDH-M) is homotropically regulated exclusively at low pH and D. mela (enzyme 15) is not allosterically regulated. The dogma suggesting that N-terminal extensions are unfavorable structural features for allostery in LDHs is thus no longer valid.

Apo and Holo C. apon LDH Crystal Structures

To get further insights into the relationship between N-terminal extensions and allosteric behavior, we have solved the structures of both T and R states of C. apon LDH.

Apo C. apon LDH crystal structure was solved by phasing using the anomalous signal of the terbium cation from the Crystallophore, a metallo-organic complex used as a nucleating and phasing agent (Engilberge et al. 2017). Apo C. apon LDH tetrameric arrangement agrees with the canonical LDHs “dimer of dimers” structures (named here dimers A//B and C//D) (Coquelle et al. 2007; Friberg et al. 2020), with the formation of 4 active sites, each of them being able to bind 1 pyruvate molecule, along with a NADH-binding site in the Rossmann-like motif, and 2 FBP-BSs located at the interface between the 2 dimers (Fig. 4a). The latest interface is described hereafter as to the AD-like interface. Indeed, because of the crystallographic symmetry, AD and BC interfaces are equivalent. No substrates are bound to the Apo C. apon LDH leaving the MbL (residues 101 to 106) open. As observed in crystal structures of LDHs or MDHs (Talon et al. 2014), there is a large solvent-accessible cavity (SAC) located in between the dimer of dimers that makes the tetrameric assembly.

Fig. 4.

Ribbon drawing of C. apon LDH crystal structure in Apo and Holo states. a) The Apo state of C. apon LDH assembly is a tetramer with 4 active sites (red bubbles) organized as 2 functional dimers (A//B and C//D, respectively, in red-blue and green-yellow). Two FBP-BSs (purple wheels) are found at the AD-like interfaces. The N-terminal extensions complete the interactions within the tetramer with, for instance, the N-terminal extensions of monomer B interacting with monomers A and D. AB- and AD-like interface position is indicated as gray arrows. b) The Holo state of C. apon LDH contains 4 oxamate and NADH molecules (shown as red spheres) and the 2 FBP-BSs are filled up with FBP molecules (shown as purple spheres). The tetramer is represented along the R-axis according to the conventional definition for LDHs (P, Q, and R-axes). Position of the MbL is indicated on both models. SAC corresponds to the solvent-accessible cavity.

The C. apon LDH ternary complex crystal structure (Holo) was solved by molecular replacement using a H. sapi LDH-M as search model (Dempster et al. 2014) as significant conformation changes between the Apo and Holo C. apon LDH prevented to find the molecular replacement solution from the Apo C. apon LDH structure. The C. apon LDH ternary complex is similarly tetrameric with the cofactor and active sites of each monomer occupied with a NADH and an oxamate molecule inducing the closure of the MbL (Fig. 4b). As expected, 2 FBP molecules are found in the FBP-BSs at the AD-like interface.

The fine description of conformation changes between the Apo and Holo states in allosteric LDHs (without any N-term extensions) is well documented (see Taguchi 2017 for review). C. apon LDH undergoes similar conformational changes with local changes induced by substrate and FBP binding impacting the tertiary and quaternary structures of the tetramers. Therefore, we will compare C. apon LDH crystal structures with bacterial and eukaryotic LDHs of B. long LDH (Apo and T-inactive states: pdb ID 1LLD; R-active state: pdb ID 1LTH, chain R) and H. sapi LDH-M (Apo: pdb ID 4L4R; ternary complex: pdb ID 4OKN), respectively, into an evolutionary perspective.

Role of N-terminal Extensions of C. apon LDH Folding

The driving point for getting structural information on C. apon LDH was to determine the position of its N-terminal extensions and in particular whether they match the position of the N-terminal extensions as described in vertebrate LDH structures. α-Carbon representations are shown in Fig. 5.

Fig. 5.

Comparison of the FBP-BS and N-terminal extensions in C. apon LDH with representative allosteric and nonallosteric Apo LDHs. a) B. long LDH (allosteric), b) C. apon LDH (allosteric, this work), and c) H. sapi LDH (nonallosteric) as Apo proteins. Structures are represented along the P-axis for observation of the FBP-BSs (purple wheel in B. long LDH and C. apon LDH) and the N-terminal extensions (in C. apon LDH and H. sapi LDH). Monomers are colored as on Fig. 4. With the exception of amino acid of the N-terminal extension (blue and green), the positions are normalized with respect to LDH nomenclature (Eventoff et al. 1977).

For the description of the structural differences in the N-terminal extensions, we use the linear structural numbering according to PDB coordinates for each enzyme. To help the reader, a sequence alignment is presented in Supplementary Fig. S3.

In the allosteric C. apon LDH, the N-terminal extension of 1 monomer (i.e. B) interacts with 2 other monomers (A and D in that example) (Figs. 4 and 5b), by creating 2 areas of contacts that are absent in allosteric LDHs lacking the N-terminal extension. In C. apon, amino acids N9 to R21 of monomer B interact with monomer D, while amino acids F2 to S8 favor the creation of a “staple-like” element (SLE) between monomers A and D that reinforces AD-like interactions by a grafting process. In fact, in the SLE, residues F2 to I5 form a short helix (αA) allowing the 2 hydrophobic residues, F2 and I5, to dive into a hydrophobic pocket located in between α1G/2G helix of monomer A and β sheet (βM) of monomer D (Fig. 5b). The hydrophobic pocket is part of the allosteric core (AlCo), a structural feature involved in the signal communication between LDH monomers (Supplementary Fig. S4) (Taguchi 2017; Iorio et al. 2021). The other extremity of SLE (L6, L7, and S8) makes a short β-sheet (β+), which completes the β-sheet motif (βK, βL, and βM) of monomer D, resulting in a supra molecular layer of these secondary structure elements. Two interfaces (AD-like and BD-like) that contribute to the tetrameric scaffold are thus strengthened by the presence of extensions whereas the AB-like interface is not.

Interestingly, the positioning of the N-terminal extensions in H. sapi LDH-M is very similar with a minor difference that hydrophobic residues (L3 and L7) of the short first helix enter deeper in the hydrophobic pocket (Fig. 5c). It has been reported that in Heart H. sapi LDH, the 8 first residues of the N-terminal extensions, with in particular residues L3 and L7, are responsible for tetramerization (Thabault et al. 2020).

In the case of B. long LDH, no extensions are present to bring closer monomers A and D. Consequences are the absence of βM (on monomer D) and the lack of constraints on the loop before α1G/2G that adopts a relaxed conformation (Fig. 5a). While the N-terminal extensions of C. apon and H. sapi LDH-M differ in length and sequence, they share similar conformation and equivalent structural stabilizing features. We next analyzed the amplitude of conformational change frequently encountered in allosteric LDHs.

The Active Site of C. apon LDH Samples the T and R states

Based on structural studies of vertebrate LDHs, it is thought that the T–R conformational equilibrium is completely shifted toward the R-active state, independently of the presence of ligands, because of additional interactions between monomers due to N-terminal extension (Abad-Zapatero et al. 1987; Read et al. 2001; Nowicki et al. 2015 ). The sigmoidal pyruvate saturation profile of C. apon LDH suggested that the enzyme can sample the T-inactive state despite the presence of N-terminal extension. So we performed a structural analysis (Fig. 6) of the C. apon LDH catalytic site using the Apo and Holo crystal structures, with a special emphasis on the substrate-binding residue R171 for which the side chain position is a relevant structural proxy for T or R states in LDHs (Coquelle et al. 2007; Colletier et al. 2010; Taguchi 2017; Iorio et al. 2021).

Fig. 6.

Detailed view of R171 and the formation of the active site in LDHs. a) B. long LDH, b) C. apon LDH, and c) H. sapi LDH-M with a comparison of the unbound and bound forms. The side chain position of R171 (OUT or IN position) is indicated. H68 from the adjacent monomer is also indicated. The purple arrows represent the movement of domains when substrates bind. It is worth noting monomer D (yellow) getting closer to monomer B (blue) and the closure of the MbL being similar for the 3 enzymes when substrate binds. The partial view of the tetramers is represented along the R-axis. αC helices from monomer B of all models were superposed to generate this detailed view. Monomers are colored as in Fig. 4. The location of the substrate analog oxamate (OXM) in the different Holo states is indicated.

In Apo C. apon LDH (no oxamate bound), the active site R171 residue points to the outside of the protein (OUT position) (Fig. 6b). In order to bind oxamate, rearrangements of helices αC, α2F, and α1G/α2G of C. apon LDH need to happen, so R171 can move in the active site (IN position) (Fig. 6b). R171 exchanges position with H68 from the adjacent monomer. Indeed, In LDHs, H68 and R171 are neighboring residues with anticoordinated side chain conformations. When H68 is in the conformation observed in the T-state structure, it prevents the side chain of R171 from accessing the active site and adopting the favorable configuration that binds the substrate analog, oxamate (Coquelle et al. 2007; Colletier et al. 2010; Taguchi 2017, Iorio et al. 2021). Despite the presence of N-terminal extension, the catalytic site of C. apon LDH samples T and R states as observed in B. long LDH structures taken as representative of allosteric structural reorganization. (Fig. 6a). Catalytic site of C. apon LDH differs therefore, from the one in the H. sapi Apo LDH-M structure, in which the R171 side chain is described as IN, pointing towards the active site, showing the enzyme is in the R-active state (Fig. 6c). In fact, with H. sapi LDH-M, the active site is in a precompetent state for catalysis with H195 and R171 ready to receive a pyruvate molecule that will trigger the closure of the MbL via interaction with Q102 (Fig. 6c).

We thus present here an unexpected sampling capacity of C. apon LDH between the T-inactive and R-state conformation as for bacterial LDHs despite the presence of N-terminal extensions as in eukaryotic LDHs. We next analyzed the amplitude of conformational change frequently encountered in allosteric LDHs.

Large Conformational Changes upon Ligand Binding

First, we analyzed the consequences of ligand binding on C. apon LDH N-terminal extensions. Upon binding of substrates, the enzyme undergoes several rearrangements as observed in the allosteric B. long LDH (Iwata and Ohta 1993; Iwata, Kamata, et al. 1994; Iwata, Yoshida, and Ohta 1994), with the closure of the MbL, the movement of α2F helix, and the bending of α1G/2G helix to reach the R-active state (Fig. 6a and b). The binding of FBP helps the movement of α2F helix by triggering the switch of the R171 position from OUT to IN (Fig. 6a and b). In the case of C. apon LDH, the movement of α2F helix and α1G/2G helix increases the size of the hydrophobic pocket below α1G/2G helix and allows the hydrophobic residues of the N-terminal extensions to enter deeper in it. In this sense, the C. apon LDH R-active state resembles to the one of H. sapi LDH-M (Fig. 6b and c). Second, we evaluated the overall reorganization of the tetramer upon binding of substrates. The additional movements in B. long and C. apon LDHs (see purple arrows in Fig. 6) drag along the β-sheet motif (βKLM) and α3G helix of, i.e., monomer D toward the opposite A//B dimer inducing further contacts within the tetramer. These interactions correspond to those existing in H. sapi LDH-M (Fig. 6). Moreover, in B. long and C. apon LDHs, the movements have a longer-range impact on the compaction of the tetramer. While Apo and complex H. sapi LDH are mostly found in a compact form, B. long and C. apon LDHs are breathing between the open T-inactive state and the compact R-active form (Fig. 7). When NAD, FBP, and oxamate bind to B. long LDH, the tetramer compacts and the access to the SAC decreases.

Fig. 7.

Surface representation of the tetramers. a) B. long LDH, b) C. apon LDH, and c) H. sapi LDH-M with comparison of their Apo and Holo states. To symbolize the compaction along the Q-axis, distances between Cα of A36, a symmetry-related residue at the A//B interface is represented in gray. The tetramer is represented along the R-axis. LDH monomers are colored as in Fig. 4.

An open assembly is necessary for the diffusion of the FBP molecules as their binding sites are only attainable from the internal SAC of the tetramer. Slice views showing the cavity are presented in Supplementary Fig. S5. Despite the N-terminal extension, C. apon LDH assembly adopts an open assembly when unliganded showing that the extensions do not hinder the access to the SAC by the FBP molecules (Fig. 7b). In the case of H. sapi LDH-M, no rearrangements other than the closure of the active loop happens upon substrate analog binding (Fig. 7c).

In conclusion, the C. apon LDH N-terminal extensions are not sufficient to shift strongly the allosteric equilibrium toward the R-active state and consequently hold it in a compact tetrameric form, as it is the case with eukaryotic counterparts.

Its SAC is still accessible and resembles the opening in B. long LDH. So, the FBP molecules can diffuse to reach their binding sites and rearrangements will have to happen to force R171 to the IN position.

The Deletion of N-terminal Extensions Affects C. apon LDH Properties

In order to evaluate the impact of the N-terminal extensions on the enzyme properties, we designed a mutant in which the N-terminal sequence (MFEKILLSNPSAENPSSLRP) has been deleted (ΔN-ter C. apon LDH). We monitored the oligomeric state of ΔN-ter C. apon LDH in parallel with the wild-type (Wt) enzyme (Wt C. apon LDH) using SEC-MALLS analysis (Fig. 8).

Fig. 8.

Effect of N-terminal deletion on enzyme properties. Oligomeric state determination using SEC-MALLS analysis. The chromatogram shows the elution profile monitored by excess refractive index (left ordinate axis) and the Mw as dashed line (right ordinate axis) derived from MALLS and refractometry measurements. The estimated average Mw is indicated on the graph. a) Wt C. apon LDH. b) ΔN-ter C. apon LDH. c) Enzymatic activity profiles using pyruvate. Measurements were done in the presence of the indicated concentrations of substrates with NADH as coenzyme. Closed and open circles are for Wt and N-terminal mutant of C. apon LDH, respectively.

On the size exclusion column, the ΔN-ter C. apon LDH is eluted at a higher volume than C. apon Wt LDH, showing it does not behave as a tetramer. The experimental weight-averaged molecular mass of 143 kDa for Wt C. apon LDH is close to the theoretical value of 144 kDa for a tetramer (Fig. 8a). In contrast, the value of 71 kDa for the mutant is consistent with a dimeric species instead of a tetramer (Fig. 8b).

We then recorded and compared the pyruvate saturation profile of ΔN-ter C. apon LDH with that of Wt C. apon LDH (Fig. 8c). The activity profile of the mutant remains sigmoid. However, the K_m value (30 mM) for pyruvate is increased for ΔN-ter C. apon LDH compared to the Wt enzyme (10 mM), demonstrating that the deletion lowers the substrate-binding affinity of the dimeric species. The k_cat value of the ΔN-ter C. apon LDH shows that the catalytic turnover is lowered compared to the native enzyme (Table S3).

The data demonstrate that the dimeric species is a less efficient enzyme compared to the tetrameric Wt C. apon LDH. A similar conclusion was drawn using the ΔN-ter mutant of the H. sapi LDH-H (Thabault et al. 2020).

We then tested if the ΔN-ter C. apon LDH can be activated by FBP and found no noticeable enhancement of activity when 0.1 or 3 mM FBP was added to the assay (data not shown). These data demonstrate that (i) the N-terminal extensions in C. apon LDH contribute to the stability of the tetrameric assembly, which is, therefore, the prerequisite for the FBP-BS formation, and (ii) that the dimeric species of C. apon LDH sustains the homotropic activation.

We monitored that deletion of the N-terminal extension lowers the apparent stability by 10 °C as assessed by residual activity measurement (Supplementary Table S3). We also designed a R. chin ΔN-terminal LDH mutant, but the enzyme did not refold properly preventing its characterization.

Altogether, the complementary structural and biochemical characterizations of new LDHs with N-terminal extensions indicate that their presence is not responsible for the absence of allosteric activation by FBP as it was thought for a long time based on data from H. sapi LDH-M only.

Detailing the FBP-BS

After demonstrating that there is no link between the absence of allosteric activation by FBP and the presence of the N-terminal extensions, we took a closer look at the AD-like interface that contributes to the formation of the FBP-BS. Structural studies have shown that residues R173, H188, and Y190 of each monomer participate to the binding of FBP (Fig. 9) and can be considered as the signature of the FBP-BS (Iwata and Ohta 1993; Iwata, Kamata, et al. 1994; Iwata, Yoshida, and Ohta 1994; Coquelle et al. 2007; Taguchi 2017). These 3 main residues are present at equivalent position in the C. apon LDH sequence. Compared to C. apon LDH, R. chin and H. sapi LDHs show each a single mutation, with the sequences R173, H188Q, and Y190 and R173, H188, and Y190W, respectively. Such variability in the signature sequence may impact the FBP-BS properties of these enzymes. To have a more precise picture, we mapped the presence of the R173, H188, and Y190 onto the LDH phylogeny (Figs. 1 and 2). The corresponding trees show that the majority of bacteria harbor indeed the FBP-BS amino acid signature (Fig. 1), while in eukaryotes (clusters I and II), the favorable combination is, in most cases, not achieved (Figs. 2 and 9). Even if there are no experimental reports in the literature of LDHs displaying an (incomplete) amino acids signature unfavorable for FBP binding, we estimate that these kinds of LDHs would not be heterotropically activated. Consequently, the P. blak LDH (Table 1) displaying both the homotropic and heterotropic allosteric activation corresponds therefore rather to an exception in cluster I (De Arriaga et al. 1982; Soler et al. 1982).

Fig. 9.

Conservation of the 3 critical residues involved in FBP-BS (R173, H188, and Y190) of allosteric bacterial LDH in eukaryotes. For clarity, taxonomic groups have been collapsed. Prokaryotic sequences are in black, while eukaryotic sequences have been colored according to their taxonomy. The conservation of the 3 residues involved in FBP-BS is reported in % (black). The most frequent alternative amino acids are indicated in red. On the top of the figure, a simplified LDH cartoon indicating the location of FBP-BS and a close-up view of the FBP-BS in C. apon LDH are shown.

Regarding cluster II, while the FBP-BS signature is present in 88% of the cyanobacterial LDH sequences, it drops down from 55% to 6% in cluster II eukaryotic lineages (Fig. 9). In fact, the R173 residue is strictly conserved in all cluster II members, while, in plants, a H188Q replacement and, in metazoan, a Y190W replacement are observed (Fig. 9). This suggests that these 2 amino acid positions could play a role in the allosteric capacity changes in eukaryotic LDH irrespective of the N-terminal extension.

Tuning C. apon LDH Allosteric Properties by Single Eukaryotic-Like Mutation

Knowing that R. chin LDH that harbors the R173, Q188, and Y190 FBP-BS signature has lost the ability to bind FBP but still displays a homotropic behavior, while the activity of the H. sapi LDH-M (R173, H188, W190 FBP-BS signature) is not controlled allosterically at neutral pH, we designed 2 single-point mutants, H188Q and Y190W in C. apon LDH.

The pyruvate saturation profile of H188Q C. apon LDH recorded at pH 7, as for the Wt enzyme, is sigmoid typical of homotropic activation (Fig. 10a). Yet, this mutant exhibits an unexpected increased affinity for pyruvate with a K_m value of 1.1 mM compared to the 10 mM value for the Wt enzyme (Table S3). Such a K_m value is close to the one determined for the R. chin LDH (1.5 mM). We did not record any significant change of activity when the assay was done in the presence of FBP, demonstrating that the H188Q mutation has abolished the FBP recognition.

Fig. 10.

Catalytic properties of C. apon LDH mutants. a) Pyruvate saturation curve for the H188Q mutant. b) Pyruvate saturation curve for the Y190W mutant.

The second single mutation Y190W has a strong effect on the C. apon LDH enzymatic profile, which becomes hyperbolic (Fig. 10a). However, the pyruvate affinity (K_m = 10 mM) stays close to the value obtained with the Wt enzyme. As it was the case with the first mutant, when the assay wass done in the presence of FBP, the enzymatic activity of the Y190W C. apon LDH mutant was not impacted.

Therefore, introducing each individual mutation of the FBP-BS carried by R. chin and H. sapi in the sequence of C. apon LDH is sufficient to abolish the heterotropic activation by FBP. However, their impact on the T-inactive/R-active state equilibrium is different, with the Y190W mutation inducing a strong shift toward the R-active state.

Structure and Electrostatic Property Changes of the AlCo

The nonallosteric properties of vertebrate LDHs have been recently challenged (Iacovino et al. 2022; Pasti et al. 2022). By recording pyruvate saturation profiles, it was found that their hyperbolic shapes turned sigmoidal when measurements were done at low pH values (pH 5), a phenomenon due to the pH-dependent dissociation of the tetramer into dimeric species. Because Y190W C. apon LDH can be considered as “mimicking” a vertebrate enzyme, we wondered whether it also displayed pH-dependent allosteric properties. Thus, we compared its pyruvate saturation profiles at pH 5 and 6 with that obtained at pH 7 (Fig. 10). The 3 profiles were hyperbolic, demonstrating that Y190W C. apon LDH activity is not sensitive to pH (Supplementary Fig. S6). In addition, we verified that Y190W C. apon LDH remains tetrameric in all tested conditions using analytical ultracentrifugation (Supplementary Fig. S6).

Strongly intrigued by the difference of allosteric properties between C. apon LDH-M and H. sapi LDH-M induced by the pH conditions, we analyzed the occurrence of histidine residues in the close vicinity of the AD-like interfaces that participate to the tetrameric assembly. Indeed, because histidine has a pKa of approximately 6.0, its ionization state may have an influence on the conformational stability. We found that in H. sapi LDH-M sequence, there are 2 histidine residues at positions 183 and 218 (Figs. 1 and 2; Supplementary Figs. S2, and S4), in the hydrophobic rich region considered as the AlCo of LDHs (Taguchi 2017; Iorio et al. 2021). They are closely located to the SLE that establishes additional bridging interactions between monomers A and D. In C. apon LDH. there are no ionizable amino acids in the same region. The structural comparison allows to see that the presence of a SLE in C. apon and H. sapi LDH-M strongly modifies the AlCo structure compared to the conformation observed in B. long LDH.

To get insights into their relative effect on allosteric behavior, we mapped the presence of these histidines onto the LDH phylogeny (Figs. 1 and 2). Interestingly, they are found almost exclusively in metazoan group III LDHs within cluster II (Fig. 2). More precisely, H183 is found in all metazoan group III sequences, whereas H218 is found in LDH-A. Structurally speaking, due to the symmetry imposed by the tetrameric scaffold, H183 and H218 are located at the P-related interface, so that they make a cluster of 8 charges that should be responsible for the pH-dependent dissociation and resulting allosteric transitions as observed in LDH-M (Iacovino et al. 2022; Pasti et al. 2022). This strongly suggests that the combination of H183 with H218 has a specific role to play in LDH from muscle.

The Fate of Allosteric Regulation in LDHs

LDH is an important enzyme catalyzing the reversible conversion of pyruvate into lactate. Deciphering the structure–function relationship of LDH is therefore of great importance to explain their subtle functionality tuning with respect to their various roles in cells, in particular in human. Indeed, lactate is a crucial compound for physiological cellular function, metabolism, and signal transduction (Adeva-Andany et al. 2014). In particular, the alteration of lactate homeostasis participates in human health and disease (reviewed by Li et al. 2022). Consequently, according to Medline, there are more than 43,000 publications devoted to LDHs. In sharp contrast, the number of studies aimed to understand the structure–function evolution of LDH is very low, and in particular, most of those focused on eukaryotic LDH are outdated. The origin of LDH functionality was the result of a MDH gene sequence drift due to a small set of mutations. This phenomenon occurred independently several times (Madern 2002; Boucher et al. 2014; Steindel et al. 2016; Brochier-Armanet and Madern 2021 ). More precisely, phylogenetic analyses have revealed that the clade of stricto sensu LDH evolved from an intermediate group of enzymes, which harbor a mix of functional properties in between the canonical tetrameric MalDHs type 3 and the clade of stricto sensu LDH (Brochier-Armanet and Madern 2021). The present study shows that canonical LDHs are not uniformly widespread over the 3 domains of life. They are almost exclusively found in Bacteria and Eucarya. The most parsimonious scenario suggests that LDH originated from MDH in bacteria and was acquired secondarily via HGT by eukaryotes and a few archaea. The quasi-absence of LDHs in Archaea is very likely due to strong differences in their carbohydrate metabolism with respect to Bacteria and Eucarya (Madern 2002; Bräsen et al. 2014). Regarding eukaryotes, 2 distinct major events led to the LDH in Fungi on the one hand and in other eukaryotes on the other hand.

Previous studies have revealed atypical features of vertebrates LDH (M, H, and C forms). Starting from these reports, our work allows depicting the fate of allostery in LDHs. Our survey of the literature indicated that most of the bacterial LDHs display both homotropic and heterotropic activations. Recently, it was shown that the addition of 2 evolutionary-related mutations (strictly found in all LDHs) in a nonallosteric enzyme from the intermediate group with MDH functionality was sufficient to give rise a homotropically activated LDH (Iorio et al 2021). These dynamically enhancing mutations have changed the conformational equilibrium of the MDH catalytic site, always found in the R-active state, allowing the resulting new enzyme (i.e. a MDH with LDH capacity) to explore the T-inactive state. In the superfamily of MDH/LDH, their selection in a nonallosteric ancestral enzyme is considered as the first evolutionary step at the origin of LDHs (Iorio et al 2021). How and when the second set of evolutionary mutations allowing heterotropic activation to emerge from a homotropicaly activated LDHs remains an open question.

The intimate relationship of eukaryotes cluster II LDHs with cyanobacterial LDHs is puzzling as it cannot be linked to the chloroplastic endosymbiosis, i.e. the capture of a cyanobacterial endosymbiont at the origin of plastids in plants and algae. Even if ancient endosymbiotic-independent HGTs between eukaryotes and cyanobacteria are considered as rare (Rochette et al. 2014), a recent study has shown that HGT transfer of genes coding for citrullinating enzymes peptidyl arginine deiminases from cyanobacteria to animals introduced new enzymatic regulatory capability through posttranslational modification (Cummings et al. 2022).

The distribution of eukaryotic LDHs into 2 distinct groups of sequences of different lengths (with or without N-terminal extensions) raises the question of the benefit of extensions. Contrary to eukaryote cluster II sequences, fungal sequences belonging to cluster I do not harbor N-terminal extensions. Regarding cluster I, only a single enzyme from fungi (P. blak) was characterized and described as activated by FBP (De Arriaga et al. 1982; Soler et al. 1982), and consistently, it harbors the 3 critical residues in the FBP-BS (R173, H188, and Y190). However, these residues are poorly conserved across cluster I, as 66% of enzymes from cluster I display an incomplete FBP-BS signature. This suggests they have secondarily lost the heterotropic activation capacity. Further investigations will be necessary to describe the evolution of the fungal sequences in more detail.

Our work on C. apon LDH revealed that long N-terminal extensions are neither specific of eukaryotes cluster II LDHs nor a structural feature preventing allosteric capacity in LDHs as thought for a long time. In contrast, structural data show that long N-terminal extension creates SLE that is anchored to the AlCo of adjacent monomers. We show that the presence of these extensions impacts fold, compactness, and local dynamics of AlCo. Accessing new regions of sequence space in enzyme evolution via insertions and deletions is considered as an important mechanism to promote functional and regulatory innovation (Banavali and Roux 2005; Vahidi et al. 2018; Emond et al. 2020). Owing to that, we suggest that the addition of N-terminal extensions allowed the ancestral eukaryotic LDH to evolve further specific lineage extinction mechanism of allosteric regulation that would not be achieved without N-terminal arms. In particular, we observed that the allosteric capacities might be partially or fully switched off by single mutations that target 2 positions of the FBP-BS sequence signature at the P-related interface. Our data on C. apon illustrate this effect. With respect to allostery, a restricted capacity is observed in plants, as all reports, including our data on R. chin, indicate a homotropic activation capacity with an absence of heterotropic activation by FBP (Betsche 1981; Tihanyi et al. 1989; O'Carra and Mulcahy 1996; Sugiyama and Taniguchi 1997). Moreover, we could indeed recapitulate the plant phenotype with the H188Q C. apon LDH mutant, which loses the heterotropic activation by FBP but keeps its homotropic activation capacity, demonstrating clearly that a single mutation at position 188 can mimic the partial extinction of the allosteric activation capacity in plants. Note that the absence of regulatory effect by FBP does not exclude that activity cannot be modulated by different mechanisms such as inhibition by ATP or other metabolites in order to fit metabolic requirements of each plants as observed with lettuce LDH (Betsche 1981).

Our present work shows that the introduction of the single-point mutation, Y190W, in C. apon LDH results in completely shifting the enzyme toward the R-active state and thus highlights the role of this residue in vertebrate LDH abolished allosteric capacity. On this basis, LDH from Metazoa groups I and III that also display W190 would behave as nonallosteric LDH. Several studies have shown that single amino acid mutations may trigger allosteric activation, i.e. the capacity to explore the T and R states, using nonallosteric (always in the R state) enzymes as a starting point (Kuo et al. 1989; First and Fersht 1993; Zhou et al. 2003; Farsi et al. 2012; Iorio et al. 2021). In contrast, the complete shift of an allosteric enzyme toward an enzyme with Michaelian kinetics, i.e. with a strongly reduced capacity to sample the T state, has been rarely documented (Stebbins and Kantrowitz 1992; Arai et al. 2011). In a molecular dynamic study using thermophilic LDHs, it has been shown that the FBP-BS pocket and the AlCo act as a tandem of microswitches, controlling the propagation of dynamics and consequently influencing the allosteric capacity (Iorio et al. 2021). The concept of microswitch implies that the allosteric signal propagation may be impacted by changes in small part of an enzyme, frequently hydrophobic rich (Vogel et al. 2006; Steen et al. 2013; White et al. 2018; Fleetwood et al. 2020; Gerrard Wheeler et al. 2021; Mariño Pérez et al. 2022). In these areas, the importance of tryptophan as a key amino acid has been documented in a family of kinases (Chopra et al. 2016). In a work using a transcriptional repressor EthR, the introduction of a W by a single mutation within a small ligand-binding pocket was sufficient to shift the enzyme into the R-active state (Carette et al. 2012). We consider therefore that the FBP-BS is a reversible microswitch, depending on the presence or absence of the effector. In eukaryotic LDH group II, the presence of W190 because of its bulky side chain, knocks down the switching capacity allowing these enzymes to stay constrained into the R-active state.

Recent studies have challenged the concept stating that LDH-A from muscle in vertebrates is nonallosteric enzymes by showing that some reminiscent allosteric capacity can be detected, in particular upon low pH dissociation of the tetrameric state (Katava et al. 2020; Iacovino et al. 2022; Pasti et al. 2022). We noticed that this phenomenon correlates with the apparition of 2 ionizable histidines (per monomer) in the AlCo of vertebrates LDHs at positions 183 and 218. We suggest that, at low pH, the 8 histidine residues mainly located in the close vicinity of the AD-like interfaces are protonated, inducing a local repulsive effect that favors dissociation of the tetramer into A//B dimeric species. A//B-like dimers are less structurally constrained and the opportunity for the enzymes to explore both T-inactive and R-active states reappears. It agrees with the fact that the minimal catalytic unit, which can sustain enzymatic activity, is the A//B dimeric species (Madern et al. 2000). We suggest that the selection of the 2 histidine residues by muscle LDHs of vertebrate, allowing to be either in a regulated or in a nonregulated state is a specific adaptive response of muscles subjected to lactic acidosis due to intense exercise.

Our integrative approach demonstrates that linking phylogenetic studies to biochemical and structural information of LDHs have upset their evolutionary history that was previously based on a restricted amount of data.

Supplementary Material

Supplementary data are available at Molecular Biology and Evolution online.

Author Contributions

D.M., C.B.-A., and E.G. conceived the project. D.M. and C.B. performed the sample preparation and biochemical characterization. D.M. and A.Y.R. performed biophysical characterizations and searched for crystal growth conditions. C.B.-A. and D.M. conducted evolutionary analyses. A.Y.R. collected X-ray diffraction data and constructed the resulting structural models. A.Y.R., E.G., and D.M. analyzed the structural data. All the authors analyzed the data and wrote the manuscript.

Data Availability

The data are available in the article and its online Supplementary material. The crystallographic data are deposited at the Protein Data Bank. Accession numbers 8AB2 and 8AB3.