Missense variant in TPI1 (Arg189Gln) causes neurologic deficits through structural changes in the triosephosphate isomerase catalytic site and reduced enzyme levels in vivo

Abstract

Mutations in the gene triosephosphate isomerase (TPI) lead to a severe multisystem condition that is characterized by hemolytic anemia, a weakened immune system, and significant neurologic symptoms such as seizures, distal neuropathy, and intellectual disability. No effective therapy is available. Here we report a compound heterozygous patient with a novel TPI pathogenic variant (NM_000365.5:c.569G>A:p.(Arg189Gln)) in combination with the common (NM_000365.5:c.315G>C:p.(Glu104Asp)) allele. We characterized the novel variant by mutating the homologous Arg in Drosophila using a genomic engineering system, demonstrating that missense mutations at this position cause a strong loss of function. Compound heterozygote animals were generated and exhibit motor behavioral deficits and markedly reduced protein levels. Furthermore, examinations of the TPI^Arg189Gln/TPI^Glu104Asp patient fibroblasts confirmed the reduction of TPI levels, suggesting that Arg189Gln may also affect the stability of the protein. The Arg189 residue participates in two salt bridges on the backside of the TPI enzyme dimer, and we reveal that a mutation at this position alters the coordination of the substrate-binding site and important catalytic residues. Collectively, these data reveal a new human pathogenic variant associated with TPI deficiency, identify the Arg189 salt bridge as critical for organizing the catalytic site of the TPI enzyme, and demonstrates that reduced TPI levels are associated with human TPI deficiency. These findings advance our understanding of the molecular pathogenesis of the disease, and suggests new therapeutic avenues for pre-clinical trials.

Introduction:

Triosephosphate isomerase (TPI) is a glycolytic enzyme that converts dihydroxyacetone phosphate (DHAP) into glyceraldehyde 3-phosphate (GAP). The isomerization of these two metabolites doubles the efficiency of glycolysis, though it is not an obligate step for glycolytic flux. Dysfunction within the human TPI gene causes an autosomal recessive disease known as TPI deficiency [1]. TPI deficiency is part of a larger family of diseases known as glycolytic enzymopathies. These disorders are caused by dysfunction of various proteins within the glycolytic pathway and typically lead to hemolytic anemia; however, TPI deficiency has also been shown to lead to neurologic symptoms [2–5]. A common pathogenic mutation has been identified in patients with TPI deficiency (TPI^E104D), while TPI^C41Y, TPI^A62D, TPI^I170V, TPI^V231M, and TPI^F240L have also been associated with the disease [6–11]. TPI mutations produce a spectrum of clinical symptoms in patients, raising questions about penetrance and expressivity. In one noteworthy report, two identical twins with the same TPI genotype had dramatically different health outcomes; one brother exhibited severe distal weakness and hemolytic anemia, while the other showed hemolytic anemia only [11]. These clinical observations demonstrate that environmental or genetic modifiers can modulate the disease.

TPI functions as a homodimer, with each monomer containing an independent catalytic site nested at the top of a (βα)₈ barrel structure known as a triose isomerase (TIM) barrel. Loop3 of the TPI monomer reciprocally extends into its dimer partner, and forms the bulk of the dimer interface (Figure 1). Mutations along this dimer interface have been shown to severely impair the stability of the TPI dimer [12]. Loop6 of the TPI monomer forms the catalytic lid of the enzyme, and its movements dictate accessibility of the substrate to the catalytic site (Figure 1). Substrate isomerization is coordinated by three key catalytic residues, K13, H95, and E165 [13]. E165 has been shown to swing in and out of the catalytic pocket ~2Å in a planar motion, and when swung in, this residue is capable of abstracting and transferring a proton from the C1 to the C2 position of its DHAP substrate [14]. K13 and N95 of the catalytic site stabilize the enediolate intermediate during this transfer, and the closure/hydrogen bonding of Loop6 excludes water from entering the pocket [13]. The exclusion of water is particularly critical for preventing the phosphate elimination reaction, which will occur readily in solution, forming the toxic byproduct methylglyoxal. Mutations of Loop6 or its surrounding hinge residues have been shown to slow the catalytic cycle of TPI [15–17]. It is believed that the closure and reopening of the enzyme lid, followed by substrate diffusion, are the key rate-limiting steps of the enzyme.

Figure 1. Wild type human TPI protein structure.

Human TPI exists as a dimer (center) and its structure is shown here with one monomer in gold and one in blue. The left inset shows the R189 salt bridge with D225 and D227 at the distal end of the TPI dimer, while the right inset shows the key catalytic residues K13, H95, and E165. These catalytic residues are shown juxtaposed next to important components of the TPI dimer interface, Loop3 and E104. PDB ID: 4POC.

Three crystal structures of human disease-associated TPI mutations have revealed that defects in enzyme dimerization and catalytic activity are sufficient to cause TPI deficiency [18–20]. The pathogenic TPI^E104D mutation sits at the dimer interface of these TIM barrels, perturbs dimer formation, and reduces protein stability (Figure 1) [19]. Conversely, the TPI^I170V and TPI^V231M mutations interfere with catalytic activity, rather than reducing dimerization [18, 20]. I170 is part of Loop6, the lid of the catalytic site, and alterations of this aliphatic residue are sufficient to change the hydrogen bonding properties within the catalytic site [20]. V231 sits on the floor of the catalytic site, and the mutation to a methionine at this position will shift the peptide backbone of α8 into the substrate-binding pocket [18]. These three molecular sources of pathogenesis are structurally distinct, but enzymatically related. Catalytic activity of the enzyme is supported by TPI dimerization, and monomeric TPI is predicted to be unstable and likely rapidly degraded [12]. Mutant TPI protein degradation can be attenuated by inhibiting the proteasome, and indeed this ameliorates neurologic dysfunction and early death of a TPI dimer mutation in Drosophila [21]. Thus, understanding the molecular dysfunction associated with each mutation may help target individual therapeutic strategies.

The present study outlines the clinical profile of a patient with two different mutant TPI alleles, (NM_000365.5:c.569G>A:p.(R189Q)) and (NM_000365.5:c.315G>C:p.(E104D)). The molecular dysfunction associated with TPI^E104D has been shown to be caused by impaired homodimerization, resulting in altered affinity for substrate and reduced thermal stability [18, 19]. However, TPI^R189Q has not previously been described in a TPI deficiency patient. This arginine acts as the basic partner of two acidic aspartates (D225 and D227), and together form an important determinant of protein stability by tethering the C terminus of the 7^th and 8^th folds of the beta barrel to the rest of the protein (Figure 1). Mutations that alter the charge or hydrophilicity at either of the R189 and D227 positions can also impair protein folding [22]. Based on these observations, we introduced similar structural mutations into the homologous TPI position of Drosophila to evaluate whether these alterations were sufficient to induce pathology. Our findings indicate that mutations in this conserved salt bridge are sufficient to reduce protein stability and elicit TPI deficiency pathogenesis, supporting the need for therapeutic interventions aimed at supporting TPI protein stability and reduced enzyme turnover.

Results:

Clinical History.

The propositus was an Italian girl, the first child of non-consanguineous parents, born at 37 weeks gestation after an uneventful pregnancy. Neonatal jaundice was present requiring phototherapy. At 3 weeks of age during a probable viral intercurrent illness, she developed severe anemia (Hb 6 g/dL) and received a blood transfusion. Neuromuscular impairment with respiratory deficiency was apparent at 1 month of age, and at 3 months bilateral diaframatic paralysis was diagnosed. A tracheostomy was performed at 5 months, with subsequent 24 hr mechanical ventilation. The patient’s clinical course was characterised by recurrent infections with haemolytic crises requiring blood transfusions, and her overall neurological condition worsened after a viral infection at 1 year of age. Brain MRI at that time showed cerebral atrophy and periventricular leukomalacia. A progressive ventilator weaning program was implemented, and by age 2 years 10 months she only required night time ventilation. Currently, the patient is 4 years and 8 months of age and remains on night ventilation with intermittent daytime support as needed with significant intellectual and neuromuscular impairment. The severe anemia initially presented following infection, and persisted after the infection was resolved. At that time, the haemoglobin level in the proband was 8.8 g⁄dL. Infant anemia can be caused by a spectrum of factors, including infections, immune system disorders, or hereditary diseases [23]. The sustained anemia combined with progressive neuromuscular indications suggested the possibility of a hereditary condition.

Molecular and enzyme studies

Molecular testing for genes of glycolytic disorders identified two missense mutations in the TPI coding region: c.315 G>C (p.E104D) and the new variant c.569 G>A (p.R189Q) (Figure 2). This numbering uses the established nomenclature for TPI mutations, assuming the start methionine is removed following translation [24]; and for consistency all residue numbering in this study uses the same convention. Sequencing of parents TPI1 gene revealed they were each heterozygous for one of the mutations (Figure 2). The TPI^E104D mutation is the most prevalent TPI missense mutation identified and leads to TPI deficiency in homozygous patients or in compound heterozygotes when paired with a second TPI mutation [8, 25]. Database searches of the Human Gene Mutation Database, the 1000 Genomes, and ExAC databases did not identify the TPI^R189Q mutation, yet the protein prediction tools Predict SNP [26] and M-CAP [27] suggested that it is likely to be pathogenic. TPI activity from red blood cells was markedly reduced (289 UI/g Hb; reference range 1407–2133), whereas the parents displayed intermediate activity compatible with the heterozygous state (father 1446 UI/g Hb, mother 584 UI/g Hb). Collectively, these findings strongly suggest that the TPI^R189Q mutation is a loss-of-function allele capable of eliciting TPI deficiency.

Figure 2. Pedigree analysis of proband and parents reveals the inheritance of two different missense mutations in the TPI1 gene.

DNA samples were collected from proband and parents, and the TPI1 gene was sequenced. A is traced in green, G is traced in black, C is traced in blue, and T is traced in red. Mixed peaks were identified at c.315 G>C and c.569 G>A, in the father and mother, respectively. These affect codons resulting in E104D and R189Q amino acid substitutions, respectively. The patient inherited both of these mutations and is thus a trans-or compound-heterozygote.

TPI^R189 mutations recapitulate neurologic dysfunction in animal models

To confirm the pathogenicity of the TPI^R189 mutant allele, we utilized a genomic engineering (GE) strategy in D. melanogaster. We have previously created a GE system to evaluate modifications of TPI [28], and have since used this system to evaluate molecular determinants of TPI deficiency [20, 28, 29]. R189 is a conserved residue TPI, and we mutated the homologous position in the TPI gene of Drosophila (TPI^R187) (Figure 3A). The Drosophila TPI^R187A, TPI^R187K, TPI^R187L, and TPI^R187S alleles were designed to examine the molecular determinants of TPI^R187 dysfunction in vivo, and were generated using the TPI genomic engineering system outlined previously [28]. The TPI^R187A, TPI^R187L, and TPI^R187S alleles proved to be homozygous lethal and were maintained with balancer chromosomes, while TPI^R187K homozygotes were viable and capable of propagating a stable stock. These results demonstrate that alterations of the basic charge at position R187 are sufficient to elicit a strong loss-of-function.

Figure 3. Mutations of TPI^R187 in Drosophila are sufficient to elicit neurologic dysfunction associated with TPI deficiency.

A) Boxshade alignment of the region surrounding the human R189 residue (indicated with #). Sequence accession numbers are listed in the Materials and Methods. B) Animals heterozygous for M80T, paired with mutations affecting the Drosophila R187 residue that alter its basic charge elicit mechanical stress sensitivity. Animals were reared at 25°C and tested on day 3, n≥20 animals, maximum responses capped at 360 sec. Variance in data was tested with a one-way ANOVA, comparisons between groups made with Tukey’s post hoc test. * indicates p<0.05, *** p<0.001, **** p<0.0001.

The patient’s heterozygous TPI genotype suggested that the TPI^R189Q allele was interacting with TPI^E104D to lead to TPI deficiency. We tested whether TPI^R187 mutations were sufficient to elicit TPI deficiency using a complementation test with the pathogenic Drosophila TPI allele, TPI^M80T. A Met-to-Thr mutation at position 80 of Drosophila TPI (also called TPI^sugarkill [30] and TPI^wstd [31]) is located at the enzyme dimer interface, similar to the human mutation TPI^E104D. TPI^M80T is a recessive mutation that leads to progressive neurologic defects in homozygotes, and demonstrates a similarly severe phenotype in combination with the null allele TPI^JS10, an allele with a 1.6kb deletion of two of its exons (also known as TPI^null) [30, 32]. We crossed the TPI^R187 and TPI^M80T mutant flies and measured complementation of the TPI^M80T behavioural dysfunction (Figure 3B). The TPI^R187K allele was capable of complementing the mechanical stress sensitivity of TPI^M80T and TPI^null (Figure 3B). Conversely, TPI^R187A, TPI^R187L, and TPI^R187S failed to complement TPI^M80T, and in fact exacerbated the mechanical stress sensitivity relative to TPI^M80T/M80T(Figure 3B). These data suggest that TPI alleles encoding a non-basic amino acid at residue 187 interact with TPI^M80T similarly to TPI^null [30].

Non-basic R187 mutations reduce TPI protein levels in Drosophila and patient fibroblasts

TPI protein levels are an important determinant of pathogenesis in TPI deficiency. Mutations at the dimer interface have been shown to lead to protein degradation, and pharmacological or genetic inhibition of the protein turnover via the proteasome is capable of ameliorating behavioural dysfunction in Drosophila models of TPI deficiency [21, 32]. Thus, we sought to determine whether the R187 substitutions altered protein levels in vivo. We crossed the TPI^R187 alleles with an allele of TPI bearing a cyan fluorescent protein (CFP) tag, TPI^WT-CFP, to measure the protein produced by mutations in TPI^R187. The addition of the CFP tag to a WT’ allele facilitated the discrimination of the two protein products based on molecular weight, while still supporting animal viability. Lysates from animals revealed that the TPI^R187A, TPI^R187L, and TPI^R187S mutations reduced TPI protein levels to a greater extent than the dimer mutation TPI^M80T (Figure 4A, B). In contrast, the TPI^R187K protein was present at a level equivalent to TPI^WT (Figure 4A, B). These data demonstrate that non-basic substitutions at TPI^R187 lead to reduced protein levels in vivo. Finally, none of the R187 substitutions significantly reduced TPI^WT-CFP levels, though the TPI^R187L did seem to elicit a slight reduction (~25%) in TPI^WT-CFP (Figure 4C).

Figure 4. Mutations that alter the basic chemistry of TPI^R187 reduce protein levels in vivo.

Animals heterozygous for a CFP-tagged wild type TPI were generated, lysates collected, and probed via western blot. A) A representative image of a western blot probed for anti-TPI or anti-ATPα. ATPα was used as a loading control. Untagged TPI variants migrated at ~27 kDa, and the CFP-tagged version at ~54 kDa. B) and C) are quantifications of the densities of untagged (~27 kDa) and CFP-tagged (~54 kDa) bands, respectively. Quantification was made using density measurements of bands produced from three independent replicates (n=3, ± S.D.). Variance was assessed with a one-way ANOVA, and comparisons to WT were made using Tukey’s post hoc test. *** indicates p<0.001, while ns indicates no significance.

The R189 residue in human TPI is conserved in eukaryotes and prokaryotes (Figure 3A), suggesting that it may have a conserved role in protein folding or stability. The mutational analysis in Drosophila demonstrated that non-basic substitutions contribute to reduced protein in vivo, leading us to hypothesize that the TPI^E104D/TPI^R189Q patient may also exhibit reduced levels of TPI. We collected fibroblast samples from the proband and a healthy control, and probed them for TPI protein levels (Figure 5). TPI enzyme levels were reduced ~75% of that from the unaffected control (Figure 5). The TPI deficient patient is a compound-heterozygote, and although we were unable to differentiate between TPI^E104D and TPI^R189Q variants, we believe greater than 50% reduction in TPI suggests that TPI^R189Q is influencing the folding or stability of TPI in the patient.

Figure 5. Total TPI protein levels are reduced in fibroblasts from TPI^E104D/TPI^R189Q patient.

A) A representative western blot probing fibroblast samples from the TPI deficient proband (P) and a healthy patient as a control (C). Beta-tubulin (Tub) was used as a loading control in the experiment. B) Quantification of relative TPI signal reveal a significant reduction in total TPI protein levels in the proband compared to control. Differences were evaluated with Student’s t-test, *** p value <0.0001, n=3.

TPI^R189 influences protein stability without altering dimerization

Previous studies have demonstrated that TPI folding defects are capable of inducing TPI deficiency [18, 19, 29]. Mutagenesis of residues crucial for homodimer stabilization have profound neuropathologic effects in Drosophila melanogaster [28, 29]. In addition to the dimer interface, the conserved R189 – D225/D227 salt bridge is a known modulator of TPI stability and function in Trypanosoma brucei (5). R189 and D225/D227 are located in helix 6 and in the C-terminal end of helix 7, respectively; linking the C-terminal β7α7β8α8 unit to the rest of the protein (Figure 1). Also, it has been previously shown that this salt bridge is an important mediator of a hydrogen bonding network imperative for the catalytic capacity and stability of TPI (5). Mutations at this salt bridge could impair human TPI dimer stability, or lead to TPI aggregation.

To examine how R189 mutations influence the physical and biochemical properties of TPI, we purified human TPI^WT, TPI^R189A, and TPI^R189Q. TPI purification was performed as outlined previously [20], and although successful, the R189 mutants yielded far less enzyme compared to TPI^WT. The migration of these enzymes through native gel electrophoresis was used as a primary strategy to assess quaternary structure. In this assay, any aggregated protein would not be able to migrate into the gel and would be retained at the top, or changes in protein homogeneity could be observed as differences in the coherence of the observed band. We observe that both R189A and R189Q mutants have a slightly higher mobility compared to WT, consistent with the loss of a charged residue (Figure S1). In all cases, no protein was detected in the well or in the stacking layer of the gel, leading us to conclude that the R189A and R189Q mutants do not cause significant protein aggregation in vitro. To evaluate whether these mutations may influence dimerization, we measured the hydrodynamic radius of all enzyme variants using dynamic light scattering (DLS). DLS measurements revealed that the two mutations of R189 do not reduce the hydrodynamic radius of the enzyme, suggesting that the dimer is intact (Table 1). Conversely, the radius is slightly larger for both mutants (Table 1), suggesting that the enzyme in solution may be more flexible than the WT variant.

Table 1.

R189 substitutions alter protein stability but not dimerization.

Sample	Radius (nm)	Melting Temperature (°C)
Wild Type	3.5 ± 0.1	49.8 ± 0.4
R189A	3.8 ± 0.1	62.1 ± 0.8
R189Q	4.0 ± 0.2	59.2 ± 0.7

We next tested the stability of these proteins using differential scanning fluorimetry. In this assay, the fluorescence of a hydrophobic dye increases as it interacts with non-polar residues of a protein that are exposed during thermal denaturation. Thermal denaturation revealed that both R189A and R189Q mutants exhibit single-phase, non-reversible denaturation (Figure S2). WT TPI exhibits a T_m of 49.8 ± 0.4°C, while both mutations of R189 shifted the melting temperature to 62.1 ± 0.8°C and 59.2 ± 0.7°C for the R189A and R189Q, respectively (Table 1). These data suggest that all three TPI enzymes are stably folded, and that the folded R189 mutants require more energy to intrinsically denature.

Structural studies indicate abnormal coordination of catalytic residues proximal to R189

In an effort to understand the effect that mutations at the R189 position might have on the structure of TPI, we have purified, crystallized, and determined the structure of TPI^R189A at 2.2Å resolution. Crystals were obtained in conditions similar to wild-type [20] in an effort to minimize the impact the crystallization environment might play on changes induced by the mutant. Despite this, the morphology of the crystals were different, and produced rod-like crystals which belong to space group P1, and contained 8 copies of TPI^R189A in the asymmetric unit, arranged into 4 TPI dimers. Wild-Type TPI buries 1705 Ų at its dimer interface, while TPI^R189A buries between 1692–1728 Ų with its 4 dimers; therefore, the structure indicates that the R189A mutant protein is also a dimer. Overall, the fold of each TPI^R189A was nearly identical to the previously reported TPI^WT (r.m.s.d = 0.13–0.27 for the eight TPI^R189A chains in the structure). In the vicinity of amino acid 189, we find that the side chains for residues D225 and D227 are both repositioned very slightly, partially filling in the void left by the absent arginine side chain (Figure 6A).

Figure 6. TPI^R189A breaks a conserved salt bridge in helix 6 and consequently influences loop6 open and close conformations.

A) The R189 salt bridge in wild-type (white) and R189A mutant (green). Wild-type enzyme mediates two hydrogen bonding interactions with D225 and D227 via R189. There is not a significant repositioning of the D225/227 loop in response to the R189A alteration, just a slight rearrangement in the positioning of these side chains. B) The TPI catalytic lid. Wild-type TPI dimer contains one lid in the catalytically competent position (yellow, closed) and one in the incompetent, open position (blue). All 8 monomers of R189A are either in the open position or are disordered (only 3 of the R189A monomers are shown her for clarity, in green), with the catalytic E165 residue in the “swung out” orientation. C) The N-hinge. Residues coloured as in (B) without backbone for positional clarity. Arrows indicate residue movements from ‘open’ to ‘closed’ positions within the wild-type enzyme. D) The loop7. Cartoon is coloured as in (B), with loop7 residues Y208 and S211 shown in sticks. A hydrogen bond to the phosphate ion is dashed, with the phosphate ion shown in sticks.

In wild-type TPI structures both from our group and others [14, 15, 20, 33–36], the formation of a catalytically competent active site is controlled by movements of a lid formed by residues 165–175, which has been observed in both open and closed conformations. Under these crystallization conditions, the closed conformation is associated with the presence of an ordered phosphate and bromide ion, which occupy positions taken by phosphate and triose moieties in the TPI substrate co-complex [20]. Only one monomer in the TPI dimer adopts this closed conformation in the 4POC wild-type structure [20]. Motions of the lid are controlled by movements at the N-hinge and C-hinge regions (residues 165–167, and 173–175 respectively) [15, 37]. Movement of the lid into the closed position, associated with a substrate bound state, results in E165 being shifted into a catalytically competent position. Interestingly, we find numerous differences between TPI^WT and TPI^R189A in the catalytic lid region. First, we find the positioning of the E165 side chain is in the catalytically inactive position in all 8 monomers within the asymmetric unit (Figure 6B). This is true regardless of the presence of phosphate and bromide ions which are found in 4 monomers. Thus, despite containing interactions which mimic the substrate bound state, TPI^R189A does not reposition E165. Second, in active sites that contain phosphate and bromide ions, the lid was not ordered, indicating that its position was flexible and not being held into the closed state (Figure 6B). The positioning of N-hinge residues 166–168 are significantly different than the occupied wild-type, consistent with those residues being tucked into the core of the protein (Figure 6C). The internal orientation of residues 166–168 is observed within the “inactive” monomer of the TPI^WT dimer with the lid open (Figure 6C); however, in TPI^R189A the catalytic lid is disordered. The movement of the conserved W168 of the N-hinge is known to be critical to the integrity of loop6 [38]. The nitrogen of the indole ring hydrogen bonds to the carboxyl of E129 in the closed state, and swaps to the hydroxyl of Y164 in the open conformation (Figure S3). These three residues are constitutively bonded, but the organization of the bonding changes as the lid closes and W168 moves away from Y164 (Figure 6C, S3). The alternating coupling of Y164::W168 and E129::W168 are thought to be a part of how the conserved P166 fulcrums E165 toward the “swung in” and “swung out” positions [15, 37, 39]. In the TPI^R189A structure, W168 does not alter its organization within this network regardless of catalytic site occupancy (Figure 6C, S3). Finally, the occupied state of loop6 has also been shown to be stabilized by the loop-clamping side chains Y208 and S211 of loop7 (Figure 6D) [40]. The hydroxyl of Y208 hydrogen bonds with the amide backbone of A176 in the occupied state, while the hydroxyl of S211 establishes two hydrogen bonds to the peptide backbone of A169 and G173 [40]. Further, the closure of loop6 helps to position the amide backbone of S211 into a position where it hydrogen bonds with the active site phosphate [41]. Interestingly, we find that the loop7 residues in TPI^R189A are positioned halfway between the “open” and “closed” conformations, with the amide backbone of S211 more closely aligning with the closed conformation, while the side-chain hydroxyl protrudes toward where the loop6 in a way similar to occupied TPI^WT (Figure 6D). Collectively, these observations support the conclusion that the TPI^R189A active site is still competent to bind substrate (as mimicked through phosphate and bromide ions); however, it does not adopt a conformation consistent with a closed catalytically competent enzyme.

DISCUSSION:

We report here evidence of a new pathogenic allele associated with TPI deficiency. The proband of this study is a compound heterozygote carrying TPI^E104D and TPI^R189Q alleles. TPI^E104D is one of six known TPI deficiency mutations, and the most common disease-associated allele. However, mutations in TPI^R189 had not previously been associated with the disease. The patient exhibited increased susceptibility to infections, neurologic impairment, and anemia; hallmarks of the disease. To determine whether TPI^R189 mutations could themselves be pathogenic, we mutated this conserved R189 in Drosophila. The homologous arginine in Drosophila is at position 187 (Figure 3A), and mutations that altered the basic chemistry of Drosophila TPI^R187 elicited homozygous lethality. Conversely, a functionally conserved substitution, TPI^R187K, was homozygous viable and behaviourally normal. Finally, we demonstrated that TPI^R187 mutations could exacerbate the neurologic deficits caused by the TPI^M80T mutation in compound heterozygote animals. These findings suggest that TPI^R189Q elicits a strong loss-of-function that can elicit neurologic disease in compound heterozygote patients.

Mutating the conserved R189 reduces protein levels in patient fibroblasts and Drosophila animal models. Mutations that alter the folding or stability of a protein can lead to reduced protein levels in vivo, leading us to purify and examine the protein. Previous TPI structural studies have highlighted the presence of the solvent-exposed salt bridge formed by R189 and D225/D227 [42, 43]. This bridge links helix 6 and helix 7 on the side of the TIM barrel opposite of the dimer interface. Mutational analysis of this conserved arginine in Trypanosoma revealed that it was an important determinant of protein stability [22]. Further, altering the acidic chemistry of D225 in yeast impaired TPI folding [43]. Our findings here suggest that the patient mutation is likely impacting both folding and stability, culminating in reduced protein levels. We purified human TPI^R189A and TPI^R189Q proteins, and gel electrophoresis (Figure S1) and DLS confirmed that these R189 mutants were forming dimers. Yet the hydrodynamic radius of the R189 mutants suggested that these enzymes were more flexible than WT (Table 1). The crystal structure of TPI^R189A demonstrates that the enzyme improperly folds the catalytic lid, loop6 (Figure 6), and the disordered character of this loop could explain its increased hydrodynamic radius. Interestingly, thermal denaturing experiments using a hydrophobic fluorescent dye suggest that the R189 mutants were more resistant than WT to thermal denaturation (Table 1, Figure S2). These results are the opposite of what was reported in the Trypanosomal protein using circular dichroism, and may be attributable to differences between the enzymes or methodologies.

The catalytic lid of human TPI is influenced by R189. We purified, crystallized, and determined the structure of TPI^R189A to examine how this mutation could physically alter the biochemistry of this enzyme. Unexpectedly, the primary difference in the wild type and R189A structures was the organization of the catalytic site; specifically i) the catalytic lid (loop6), ii) the catalytic E165, iii) the N-hinge, and iv) the loop6-clamping residues of loop7. These four structures are mechanistically linked: the N-hinge immediately proceeds the catalytic E165, and the orientation of this acidic residue is in part influenced by its neighbouring P166. The peptide backbone in this region is thought to be torqued by the movement of the conserved W168 (Figure 6C). The indole ring of W168 hydrogen bonds with two nearby residues, Y164 and E129 (Figure S3). The organization of this hydrogen bonding triad is necessary for catalysis, and the activity of the enzyme is reduced as the movement of the catalytic lid is impaired [15, 17]. Indeed, a W168F substitution at the N-hinge has been shown to disorder the lid [38].

It is currently unclear how the R189A mutation physically influences the organization of the N-hinge. One hypothesis is that the movement of the N-hinge may be influenced by helix 6 through V184, which buttresses Y164 through van der Waals interactions (Figure 6C). V184 is one helix turn from R189, therefore changes to this basic residue could be predicted to break the salt bridge necessary for maintaining the integrity of helix 6, and thereby supporting the movements of the catalytic lid. However, we do not see any changes in the density of V184 in the TPI^R189A structure, therefore our current data cannot rigorously test this hypothesis. It is possible that the confines of the crystal lattice may be preventing additional motions in the V184 region. A second possibility is that the R189::D225/D227 salt bridge is important for positioning the backbone of the loop7 residues Y208 and S211. Loop7 helps to stabilize the “closed” or occupied state of the catalytic site through hydrogen bonds formed with the backbone of the catalytic lid as well as the substrate. Residues Y208 and S211 are seen in an intermediate conformation in the TPI^R189A structure, relative to the positions taken when in the “open” and “closed” WT enzyme. The inability for these residues to fully shift and clamp the catalytic lid in the occupied state may contribute to the instability of loop6 in the TPI^R189A mutant.

The rigid flexibility of TPI’s catalytic lid (loop6) is a key physical mechanism for directing the isomerization of DHAP [36]. The closure of this enzyme lid limits the access of water molecules to the substrate intermediate [44]. As the proton of DHAP is being shuttled by E165, the former carbonyl backbone of the enediol phosphate intermediate is exceedingly sensitive to attack by water [36, 45, 46]. In solution, this reaction readily forms the toxic by-product, methylglyoxal [47]. However, TPI facilitates the selective isomerization of its substrate by protecting these reaction intermediates and stabilizing the charged transition state in the conversion of DHAP to G3P [48]. A previous study has even gone so far as to purposefully create an artificial methylglyoxal synthase by simply deleting the loop6 lid of the TPI enzyme, thus stabilizing the substrate intermediates without protecting them from solvent [36]. The disordered nature of this catalytic lid could thus potentially trigger recognition by quality control mechanisms within the cell. The TPI^M80T (a.k.a. TPI^sugarkill) mutation in Drosophila is known to be recognized by Hsp70 and Hsp90, and then targeted to the proteasome for degradation and genetic or pharmacologic impairment of HsP70/90 or the proteasome improves longevity, demonstrating that mutant TPI retains function and increased turnover is a critical part of disease pathogenesis [21]. Therefore, protein misfolding, recognition, and proteasome degradation could explain why TPI^R189 mutations lead to reduced protein levels in humans and flies, without substantially reducing thermostability in vitro. Furthermore, since TPI is an essential enzyme in organisms from E coli to humans, it is reasonable to propose that one common class of disease-causing TPI mutations may retain TPI catalytic function but have reduced steady-state levels due to increased turnover of the mutant protein.

We propose that mitigating mutant TPI degradation may be an effective therapy for several, if not most, pathogenic TPI deficiency mutations. There are no current therapies for treating the neurological symptoms of TPI deficient patients. Bone marrow transplants have been tested in mice [49] and humans (personal communication), and an animal study found that it corrected red blood cell defects in a murine model of TPI deficiency [49]. These results are promising for the hematologic components of the disease, but a recent study has suggested that the neurologic dysfunction associated with TPI deficiency is intrinsic to vesicle recycling defects in neurons [29]. These observations suggests that an effective neurologic therapy may need to target the peripheral or central nervous system. In an animal model of TPI deficiency, proteasome inhibitors were systemically administered and shown to attenuate behavioural dysfunction in a Drosophila model of TPI deficiency [21]. The proteasome inhibition was shown to support levels of the dimer mutant, TPI^M80T, which is prone to protein degradation. We have reported here that TPI^E104D/TPI^R189Q patient fibroblasts have approximately 25% the normal levels of TPI enzyme (Figure 5). Therefore, proteasome inhibitors may provide a new therapeutic avenue for treating neurologic dysfunction in TPI deficiency. Indeed, several proteasome inhibitors have already obtained FDA approval for oncology indications. First and second-generation proteasome inhibitors have been used to treat multiple myeloma and mantle-cell lymphoma, with side-effects including peripheral neuropathy [50–52]. Although these secondary responses may seem counterproductive for treating neurologic disease, they indicate that these drugs are having an impact on the nervous system and may require alternative administration or dosing. Pre-clinical and clinical studies will be needed to more thoroughly evaluate the therapeutic potential of these compounds in TPI deficiency.

In conclusion, the data presented in this study highlights a new pathogenic mutation that contributes to TPI deficiency in patients. Mutations affecting the R189 residue in humans or in flies (TPI^R187) causes reduced TPI enzyme levels, and this may be structurally induced by changes in the folding of the enzyme lid. These results are critical for directing future pre-clinical and clinical studies aimed at developing therapeutic strategies for attenuating the neurologic dysfunction seen in TPI deficient patients.

MATERIALS AND METHODS:

Haematological and enzyme assays

Blood samples from patient and parents were collected after obtaining informed consent. Routine haematological investigations were carried out according to [53]. TPI activity was determined according to [54]. All the diagnostic procedures and investigations were performed in accordance with the Helsinki Declaration of 1975.

Molecular analysis of TPI gene

Genomic DNA was extracted from leucocytes using standard manual methods. The entire codifying region and intronic flanking regions of TPI1 gene were analysed by direct sequencing (ABI PRISM 310 Genetic Analyzer; Applied Biosystems, Warrington, UK) using Big Dye Terminator Cycle Sequencing kit (Applied Biosystems). Nucleotide numbering refers to NCBI Reference Sequence NM_000365.5. TPI sequences used for alignments were Hs_TPI (CAA49379.1), Dm_TPI (CAA40804.1), Mm_TPI (CAA37420.1), Gg_TPI (NP_990782.1), Sc_TPI (ONH79887.1), Tb_TPI (CAA27559.1), and Ec_TPI (ADX52952.1).

Patient fibroblasts

Patient cells were obtained from a 4-year-old female TPI deficiency patient via skin punch using University of Pittsburgh approved IRB 0404017. Cells were cultured using standard methods (37C, 5% CO₂) in complete media (DMEM with 10% serum, 100u penicillin/100 ug streptomycin/ml (Lonza), 2 mM L-glutamine (Gibco) and supplemental non-essential amino acids (Gibco). The cells were de-identified and are known only as FB887.

Drosophila husbandry and strains

Drosophila strains were maintained on standard cornmeal molasses media at room temperature in a 12:12 light:dark cycle, unless indicated.

Mutagenesis and Genomic Engineering

Site directed mutagenesis was achieved using the QuikChange Lightening Site-Directed Mutagenesis Kit (Agilent Technologies). Oligomeric mutagenic primers were engineered (Integrated DNA technologies) to introduce an arginine-to-alanine/lysine/leucine/serine change at position 187 (TPI^R187A, TPI^R187K, TPI^R187L, TPI^R187S). Mutations were generated on the pGE-attBTPI⁺ plasmid and confirmed via sequencing [28]. Once generated, the constructs were injected into the embryos generated by mating of ;;PGX-TPI founder line males crossed with vasa-phiC31^ZH−2A;; virgin females for in genomic engineering [55]. F1 progeny were screened for integration indicated by w⁺ expression, and integration was confirmed by sequencing.

Drosophila behavioural testing

Approximately equivalent numbers of males and females were collected and aged 3 days at 25°C on standard molasses-based media. At testing, mechanical stress sensitivity was determined by vortexing flies in a standard vial for 20 seconds and measuring their subsequent time to recovery [56]. Recovery was scored following two purposeful movements, such as righting, grooming, or walking, and behavioural responses were capped at 360 seconds. A oneway ANOVA assessed variance in the behavioural responses, and Tukey’s post hoc test was used to compare genotypes.

Western Blotting

Drosophila immunoblotting was performed as outlined previously [29]. Blots were incubated with anti-TPI (1:5000; rabbit polyclonal FL-249; Santa Cruz Biotechnology) or anti-ATPalpha (1:10,000; mouse monoclonal alpha5; Developmental Studies Hybridoma Bank). Densitometric analyses of the scanned films were performed on unsaturated exposures using ImageJ software available from the National Institutes of Health. A one-way ANOVA was performed to assess variance of TPI levels and data sets were compared using Tukey’s post-hoc analysis.

Human fibroblasts were trypsinized (0.05% for 5 minutes), pelleted, resuspended in RIPA buffer with protease inhibitors (PMSF (100uM), Leupeptin (1ug/uL), Pepstatin A (0.5ug/uL)) and were pulse sonicated. Protein concentrations were determined using a BCA assay (Pierce). Immunoblotting was performed on whole protein cell lysates following the addition of an equal volume of 2X SDS PAGE sample buffer (4% SDS, 4% β-mercaptoethanol, 130 mM Tris HCl pH 6.8, 20% glycerol). Proteins were resolved by SDS-PAGE (12%), transferred onto 0.45 μm PVDF membrane. The blots were blocked in Odyssey Blocking Buffer (Licor) and incubated with anti-TPI (1:5000; rabbit polyclonal FL-249; Santa Cruz Biotechnology) or anti-Beta-tubulin (1:1000; mouse polyclonal E7-C; Developmental Studies Hybridoma Bank) diluted in Odyssey Blocking Buffer (Licor). Following washes in PBST, the blots were incubated with anti-mouse-IR800 (Fisher Scientific) and anti-rabbit-IR680 (Molecular Probes) both diluted to 1:20,000 in 0.1% Tween 20 blocking buffer. Blots were washed in PBST and developed using Odyssey Infrared Imaging System. Quantification of the scanned films was performed digitally using ImageJ software available from the National Institutes of Health. Differences of TPI levels were evaluated by a two-tailed Student’s t-test.

TPI Protein Purification

Wild-type and mutant TPI proteins were expressed in E.coli, BL21 (DE3) Codon+ cells using ZY autoinduction media [57]. Cell were collected by centrifugation and lysed via homogenization in a buffer containing 25 mM Tris pH 8.0, 500 mM NaCl, 10% glycerol, 5 mM imidazole, and 1 mM β-mercaptoethanol. The lysate was cleared by centrifugation, and TPI was purified by nickel affinity chromatography. Overnight digestion using TEV protease followed by a second round of Nickel affinity chromatography was used to remove the His₆-MBP tag and other contaminants. Anion exchange chromatography and gel filtration chromatography were performed as final polishing steps. Peak fractions of TPI^R189A from gel filtration that were used for crystallization were dialyzed into 20 mM Tris pH 8.8, 25 mM NaCl, 2.0% glycerol, and 1 mM β-mercaptoethanol and concentrated to 4.55 mg/mL.

Crystallization and structure determination

TPI^R189A crystals were obtained using the sitting drop vapor diffusion method at 4°C. Crystals grew within 2 days in a drop containing 1 μL of protein and 2 μL of the 28% PEG 2000, 50 mM KBr reservoir solution. Crystals were cryoprotected in 40% PEG 2000, 10% glycerol, and 50 mM KBr and flash frozen in liquid nitrogen prior to x-ray diffraction. Diffraction data was collected at our home source using a Rigaku FR-E generator and RAXIS4-HTC detector. Diffraction data were integrated, scaled, and merged using HKL2000 [58] using an I/σI cutoff of 2.0. hTPI^R189A crystals belong to the space group P1 (a = 65.142 Å, b = 73.661 Å, c = 92.842 Å; β = 90.026°) and contain 4 dimers in the asymmetric unit. Initial phases were estimated via molecular replacement using a search model derived from an independent structure of human TPI (2JK2) [19]. The model was then refined and improved by manual rebuilding within Coot [59] combined with simulated annealing, positional, and isotropic B factor refinement within Phenix. Model quality was assessed using MolProbity [60]. Structural figures were generated using PyMol (PyMOL Molecular Graphics System, Version 1.5.0.4, Schrödinger, LLC.). The coordinates and structure factors associated with hTPI^R189A structures have been deposited within the Protein Databank under accession code 6NLH.

Thermal Shift Assays

The thermal stability of hTPI^R189A and hTPI^R189Q mutants relative to wild type were determined using a protein thermal shift assay [61]. Proteins assayed were diluted to 0.1 mg/mL in a buffer containing 15 mM Tris pH 8.8, 200 mM NaCl, 5% glycerol, 1mM β-mercaptoethanol, and 1.7x SYPRO Orange. Fluorescence of SYPRO Orange was measured and plotted as a function of temperature to quantify protein unfolding. The T_m was defined as the temperature with the maximum rate of change in fluorescence.

Dynamic Light Scattering

TPI^R189A, TPI^R189Q, and wild type proteins were diluted to a final concentration of 0.5 mg/mL in a buffer containing 15 mM Tris pH 8.8, 200 mM NaCl, 5% glycerol, and 1mM β-mercaptoethanol. Dynamic light scattering measurements were taken on a DynaPro NanoStar and analyzed using the DYNAMICS software package.

Table 2.

Data Collection and Refinement Statistics

	Human TPI R189A
PDB ID: 6NLH
Data collection
Space group	P1
Cell dimensions
a, b, c (Å)	65.142 73.661 92.842
α, β, γ (°)	90.031 90.026 89.999
Unique Reflections	84,091
Resolution (Å)	50.0 – 2.20(2.24–2.20)a
Rmerge (%)b	8.1 (41.4)
I / σI	16.8 (3.29)
Completeness (%)	96.0 (93.8)
Redundancy	2.5
Refinement
Resolution (Å)	20.0 – 2.20
Rworkc/ Rfreed (%)	17.8/21.4
Number of non-hydrogen atoms
Protein	15,388
Solvent and Ligands	818
B-factors (Å2)
Protein	35.0
Solvent	29.5
R.m.s. deviations
Bond lengths (Å)	0.006
Bond angles (°)	0.67
Ramachandrian
Outliers (%)	0.0
Allowed (%)	2.6
Favored (%)	97.4
Clashscore	3.15

^aValues in parentheses are for highest-resolution shell.

^bR_merge = (|(ΣI - <I>)|)/(ΣI), where <I> is the average intensity of multiple measurements

^cR_work = Σ_hkl||F_obs(hkl)|| - F_calc (hkl)||/Σ_hkl|F_obs(hkl)|.

^dR_free represents the cross-validation R factor for 1.5% (~1,250) of the reflections against which the model was not refined.

Highlights:

• We report a novel TPI pathogenic variant Arg189Gln found in a compound heterozygote patient with the “common” mutation.

• Modeling in Drosophila demonstrates TPI^R189Q is a strong loss-of-function mutation that disrupts critical TPI dimer salt bridges.

• Structural studies demonstrate that mutations affecting the Arg189 position alter the coordination of the substrate-binding site and important catalytic residues.