Functional characterization of natural variants found on the major stress inducible 70-kDa heat shock gene, HSPA1A, in humans

Abstract

In this report, we investigated the effects of natural single nucleotide polymorphisms on the function of HSPA1A, the major stress-inducible Hsp70 gene in humans. We first established that all mutant proteins retain their ability to hydrolyze ATP, but three of them had a significantly lower rate of ATP hydrolysis as compared to the wild-type (WT) protein. We also used Isothermal Titration Calorimetry and found that although all mutants bind to protein substrate with dissociation constants similar to the WT protein, four of them had increased reaction entropies. We also tested whether these mutations affect the ability of HSPA1A to refold heat-denatured luciferase. These assays revealed that one mutation resulted in significantly lower levels while a second one resulted in higher levels of the refolded enzyme. We then determined whether the mutations affected the ability of HSPA1A to prevent apoptosis caused by poly-glutamine carrying huntingtin proteins. This assay determined that three of the mutations caused increased cell apoptosis as compared to the WT. Our results reveal that although none of these naturally occurring mutations exists on positions of known function, some alter the molecular chaperone activities of HSPA1A most probably by affecting the allosteric communication between its two major domains.

1. Introduction

Elucidating how genetic variation contributes to human evolution, survival, adaptation, and disease predisposition is an overarching goal in modern human genetics. Such information provides insight into how our species has adapted to almost every environment on the planet, as well as reveals why we are predisposed for particular diseases. For these reasons, several genome-wide projects described the genetic variation in humans and associated specific single nucleotide polymorphisms (SNPs) with diseases [1]. Although these studies have identified several genetic polymorphisms and loci, most of these mutations have not been experimentally tested, and their functional outcomes remain unknown.

Molecular chaperones and in particular the 70-kDa heat shock proteins (Hsp70s) are essential regulators of the cellular stress response and their roles in species adaptation and survival have been linked to several normal and disease phenotypes. Loss of function or overexpression of HSPA1A, the major stress-inducible gene in humans, have been linked with heart, Alzheimer’s, and Parkinson’s diseases, as well as cancer [2]. However, the exact relationship between naturally occurring mutations and their outcome on the function of HSPA1A remains mostly uncharacterized [3].

Given the importance of genetic variation and the critical functions of HSPA1A, we sought to determine whether non-synonymous (amino acid altering) SNPs affect the function of HSPA1A. We identified and collected these mutations from the 1000 Genome project [1], and preliminarily characterized their evolution [3]. This initial characterization determined that these mutations have a very low frequency in humans and none of them alters an amino acid of known function [3]. Furthermore, these analyses revealed that these mutations have subtle effects on protein stability, nucleotide binding, and protein subcellular localization and translocation after heat-shock [3]. However, based on the conservation, radicality, and structural position (Fig. 1), as well as the proximity of these mutations to regions that are affected by substrate binding or ATP hydrolysis [4] we predicted that they might alter the way HSPA1A functions during protein refolding or aggregation. Here, we tested these predictions using recombinant proteins and a human cell system.

Fig. 1.

Three-dimensional model of HspA1A showing the major domains and the positions of the mutations used in the present study. The central panel shows a ribbon representation of the complete HspA1A structure depicting the major domains and subdomains. The inlets show a magnified version of the position of each mutated residue, which are represented as sticks to view the atoms and side chains conformations. All figures were prepared in PyMol (DeLano Scientific).

2. Materials and Methods

2.1. Three-dimensional structural modeling

The HSPA1A structural model (3D model) was generated using I-TASSER (http://zhanglab.ccmb.med.umich.edu/I-TASSER) [5] based on the 2QXL structure [6]. The mutations used in the present study (S16P, S16Y, R36C, I74T, I480N, F592S, and K71A as control) are described in detail in [3]. Their position was mapped onto the 3D model using PyMol (DeLano Scientific).

2.2. Generation of recombinant DNA clones and site-directed mutagenesis

The HSPA1A clone used in the present study was subcloned into pet-22b and peGFP-C2 vectors, as described in [3, 7]. In this study, two different constructs of the huntingtin gene exon 1: a 23-glutamine control (Q23) and 74-glutamine experimental (Q74) were used. The original plasmids (pHM6-Q23pHM6-Q74) were a gift from David Rubinsztein (Addgene plasmid # 40264) [8]. To monitor transfection efficiency, Q23 and Q74 were C-terminally fused by PCR to the coding sequence of the red fluorescent protein (RFP) and then subcloned to the pcDNA3.1(-)/myc-His C vector (Thermo Scientific). DNA mutations were created through site-directed mutagenesis as described in [3]. The luciferase pGL4 vector was purchased from Promega.

2.3. Generation of recombinant proteins

Recombinant HSPA1A corresponding to the wild-type and mutated proteins was generated from sequence-verified recombinant clones as described in [3, 7].

2.4. Cell culture and transfection

HeLa (ATCC^® CCL-2^™) were maintained in a humidified 5% CO₂ atmosphere at 37°C in complete medium consisting of MEM supplemented with 10% fetal bovine serum, 2 mM L-glutamine, penicillin-streptomycin, and 0.1 mM non-essential amino acids and sodium pyruvate. A day before transfection, cells were split into 24-well plates at 2.0 × 10⁴ cells/well. After 18 hours, cells were transiently transfected with the appropriate construct using the PolyJet In Vitro DNA Transfection Reagent (SignaGen) as per the manufacturer’s instructions. Transfection was allowed to continue for 18 hours, and then transfection media was removed and replaced with fresh complete media. The amounts of DNA used per ml were: 0.7 μg of HSPA1A and 0.3 μg of luciferase (for the refolding assay) and 0.5 μg of HSPA1A and 0.5 μg of Q23 or Q74 huntingtin variants (for the aggregation assay).

2.5. Activity tests using recombinant proteins

To test the effects of the mutations on the function of HSPA1A, we determined whether they affect the ability of HSPA1A to hydrolyze ATP. The ATPase assay used was performed as described in [7]. The phosphate released in three independent experiments was plotted against time and the rate of hydrolysis was quantified using the slope generated by linear regression [9].

We next employed Isothermal Titration Calorimetry (ITC) to determine whether mutated variants of HSPA1A bind to protein substrates similarly to the WT protein. ITC measurements were performed as described in [3] using 4 μM of each protein and 16 mM of protein substrate (peptide NIVAKKK) [10, 11]. The data were processed with the NanoITC software (NanoAnalyze Software v3.7.0). Each measurement was repeated three times with different protein batches and values for each experiment were presented as independent values.

2.6. Activity tests using a mammalian cell system

To test the chaperone functions of HSPA1A within a cell, we determined whether the mutations affect (i) refolding of heat-denatured luciferase and (ii) protection against protein-aggregate stress.

(i) The intracellular refolding assay was performed exactly as described in [12] with one change: the denaturation step was performed at 45 °C for 16 min. The luminescence was detected using the Dual-Glo Luciferase reagent (Promega) and measured using the GloMax^® 96 Microplate Luminometer (Promega). The raw luminescence measurements per experimental replication were averaged and normalized as percentages of the ratio of the experimental average to the GFP 37 °C control average, and standard deviation was calculated from three biological replicates. Total fluorescence was measured using the SpectraMax® M3 Multi-Mode Microplate Reader (Molecular Devices).

(ii) The aggregation assay was performed as described in [13]. Transfection was allowed to continue for 48 hours and apoptotic cells were determined using the Caspase-3/7 Glo reagent (Promega). Raw luminescence was measured using the GloMax^® 96 Microplate Luminometer (Promega). The results for each experimental construct were averaged and relative caspase activity were determined by calculating the ratio of the average luminesce for the experimental construct to the average luminescence of the GFP-Q74 control (which was set at 100%).

2.7. Statistical tests

Statistical significance was determined by an unpaired t-test. A P value < 0.05 was considered statistically significant. The boxplots were generated using R software (http://shiny.chemgrid.org/boxplotr/).

3. Results

3.1. Structural predictions

The functional effects of these non-synonymous SNPs were predicted based on several criteria [3]. These criteria established that: none of these mutations change an amino acid of known function; they occurred on highly conserved amino acid positions; were radical; and were predicted to alter protein functions. Additionally, the mapping of these mutations on an HSPA1A 3D structural model predicted that these mutations might alter local conformations or the molecule surface (Fig. 1). Furthermore, some of these mutations (Fig. 1) are in close proximity to the lysine at position 71, known to alter ATP hydrolysis [14], and the aspartic acid at position 481, known to affect inter-domain communication [4].

3.2. Effects of mutations on the ATPase function of HSPA1A

To determine the effect of the SNPs on the ability of HSPA1A to hydrolyze ATP, we used recombinant proteins (Suppl. Fig. 1) and a malachite green based assay (Fig. 2). The results of the ATPase assay revealed that all mutant HSPA1A proteins retained the ability to hydrolyze ATP (Fig. 2 and Suppl. Table 1). However, the rate of ATP hydrolysis was altered when compared to either the WT-HSPA1A or the K71A mutation, which is known to have negligible ATPase activity [14]. Specifically, three mutations (R36C, I74T, and F592S) had a significantly lower rate of ATP hydrolysis as compared to the WT protein (Fig. 2 and Suppl. Table 1). However, the rate of ATP hydrolysis of the I74T and F592S mutations was significantly higher than the K71A control. Two of these mutations, R36C and I74T, are located at the lobe IB of the nucleotide-binding domain (NBD) of HSPA1A (Fig. 1), and these positions have not been implicated with the ATPase function of HSPA1A. The last mutation, F592S, is found at the α-helical lid subdomain of the substrate-binding domain (SBD) of the protein, which has not been implicated with ATP hydrolysis.

Fig. 2.

HSPA1A mutated variants retain their ability to hydrolyze ATP, but the rate of hydrolysis is different than the wild-type (WT) protein. Quantification of the phosphate released after incubation of HSPA1A with ATP for different periods of time. Centerlines show the medians; box limits indicate the 25th and 75th percentiles as determined by R software; whiskers extend 1.5 times the interquartile range from the 25th and 75th percentiles; crosses represent sample means. The experiment was repeated three times and the slope of the reaction is shown below the protein name. Star (*) denotes statistical significance versus the wild-type protein (see Suppl. Table 1).

3.3. Effects of mutations on the binding of HSPA1A to a protein substrate

To determine whether any of these mutations alter the binding to protein substrates we used recombinant proteins (Suppl. Fig. 1) and Isothermal Titration Calorimetry (ITC). Our results (Table 1 and Suppl. Fig. 2) are summarized below. First, all variants bind to protein substrate with similar affinity to the WT and the K71A (which is known not to affect substrate-binding) proteins. Second, the enthalpy (delta H) of the interactions is in all cases negative, suggesting that the noncovalent interactions (hydrogen bonds and van der Waals interactions) at the binding interface remain largely unaffected. However, the S16Y mutation results in a larger negative enthalpy change as compared to the WT protein and the S16P result is a smaller negative enthalpy change, similar to the K71A mutation. Third, the entropy change (delta S) of the interaction is very different for all mutations studied (Table 1). Specifically, there is a large entropy gain (reversed as compared to the WT protein) for the binding to protein substrate of the S16P, R36C, I480N, and F592S.

Table 1.

Binding and thermal results of the Isothermal Titration Calorimetry (ITC) assays using purified recombinant HspA1A proteins (WT and mutated variants) with protein substrate

Cell	Syringe	N	Kd (mM)	dH (kJ/mol)	dS (J/mol*k)
WT1	Peptide	0.993	3.35E-04	−25.84	−36.93
WT2	Peptide	1.002	5.29E-04	−19.75	−23.52
WT3	Peptide	1.023	3.64E-04	−20.03	−31.35
K71A_1	Peptide	0.982	5.28E-04	−11.58	21.91
K71A_2	Peptide	1.002	6.15E-04	−13.87	20.25
K71A_3	Peptide	0.989	6.35E-04	−11.97	19.89
S16P_1	Peptide	0.991	7.38E-04	−12.47	16.92
S16P_2	Peptide	1.005	6.79E-04	−13.43	15.61
S16P_3	Peptide	1.027	5.99E-04	−13.48	16.47
S16Y_1	Peptide	1.038	3.35E-04	−42.71	−72.72
S16Y_2	Peptide	1.021	3.31E-04	−41.66	−73.09
S16Y_3	Peptide	1.007	3.15E-04	−40.76	−71.18
R346C_1	Peptide	1.028	5.76E-04	−15.71	11.53
R346C_2	Peptide	1.001	4.72E-04	−15.95	12.52
R346C_3	Peptide	0.984	5.09E-04	−16.24	10.55
I74T_1	Peptide	0.998	9.99E-04	−18.91	−5.973
I74T_2	Peptide	0.985	9.95E-04	−19.67	−4.888
I74T_3	Peptide	1.014	9.35E-04	−20.28	−6.559
I480N_1	Peptide	1.017	7.10E-04	−18.53	23.28
I480N_2	Peptide	1.007	6.61E-04	−21.18	22.95
I480N_3	Peptide	1.025	6.97E-04	−19.74	24.83
F592S_1	Peptide	1.071	5.59E-04	−16.07	8.356
F592S_2	Peptide	1.037	5.00E-04	−19.52	11.56
F592S_3	Peptide	1.062	3.74E-04	−15.32	10.92

N: reaction stoichiometry; dH: enthalpy; dS: entropy; Kd: dissociation constant; Cell: instrument cell containing protein; Syringe: instrument syringe for ligand titration

3.4. Effects of mutations on the protein refolding function of HSPA1A

To determine whether the mutations affect the refolding ability of HSPA1A, we assessed the refolding of heat-denatured luciferase in HeLa cells (Fig. 3A) overexpressing the wild-type or mutated HSPA1A variants (Suppl. Fig. 3A and 4). These results revealed that the majority of mutations refold denatured luciferase at the same rate as the WT. However, the R36C and I480N mutations significantly altered the percent of refolded luciferase. Specifically, for the R36C variant, the percent of refolded luciferase increased by approximately 16% compared to the WT. Differently, the I480N variant showed a decrease in refolded luciferase by approximately 9%, which is similar to the rate of refolding we observed using the K71A mutation. These results suggest that key properties of the amino acids found at positions 36 and 480 of HSPA1A appear to be essential to the protein’s ability to promote denatured protein refolding.

Fig. 3.

HSPA1A natural variants alter the refolding and the anti-apoptotic properties of the protein. (A) All mutations refold heat-denatured luciferase at a level similar to the wild-type (WT) protein, except the R36C and I480N variants, which show significant changes as compared to the wild-type protein. (B) The R36C, I480N, and F592S variants result in altered caspase-3/-7 activity after poly-glutamine aggregation stress. In all graphs, centerlines show the medians; box limits indicate the 25th and 75th percentiles as determined by R software; whiskers extend 1.5 times the interquartile range from the 25th and 75th percentiles; crosses represent sample means. Each experiment was repeated three times. The P values of the student t-test were: (B) GFP/WT=0.0010; GFP/S16P=0.0010; GFP/S16Y=0.0010; GFP/I74T=0.0010.

3.5. Effects of mutations on the protecting function of HSPA1A against protein-aggregate cell toxicity

We next sought to determine the effect of SNPs on the ability of HSPA1A to protect against protein-aggregate stress. To this end, we overexpressed elongated huntingtin protein mutants and characterized their effect on inducing cell apoptosis (Fig. 3B and Suppl. Fig. 3B). As a negative control, we measured the caspase activity in cells overexpressing the Q23 construct (Suppl. Fig. 5), which should not induce apoptosis [13]. Our results using the Q74 constructs revealed that the WT protein results in a 15% decrease in Caspase-3/-7 activity compared to the GFP-Q74 control (Fig. 3B). A similar decrease is also observed in the variants S16P and S16Y. The I74T variant also inhibits caspase activity, but to a greater extent than the WT protein. Lastly, the constructs R36C, I480N, and F592S show no decrease of caspase activity (Fig. 3B), suggesting that these proteins may fail to protect cells from protein aggregation stress.

4. Discussion

This study aimed to characterize the effects of six naturally occurring mutations on the function of HSPA1A protein in humans. Our results suggest that although none of these mutations are found on positions of known function, some alter the molecular chaperone activities of HSPA1A.

In a previous report, we found that these mutations have subtle effects on binding to nucleotides, protein subcellular localization, or structural stability (ref). Our current study reveals that the R36C, I74T, and F592S have a significantly lower rate of ATP hydrolysis. Among these mutations, the R36C and I74T, are topologically in close proximity to the lysine at position 71 and the ATP-binding pocket. Based on these observations, we suggest that they might affect ATP hydrolysis by indirectly altering the bonds formed and the secondary movements of the molecule. This prediction is indirectly supported by the finding that C17, which does not bind directly to ATP, alters the ATPase activity of Hsp70 by disrupting the hydrogen-bond network within the active site [15]. Differently, the F592S mutation, which is located at the lid subdomain of the SBD and showed an increase in ADP-binding entropy [3], may alter the conformation or movement of the lid subdomain and thus indirectly affect ATP hydrolysis.

Furthermore, our results suggest that the S16P, R36C, I480N, and F592S mutations cause noticeable alterations in the molecular forces that govern the ability of HSPA1A to interact with protein substrates by increasing the reaction entropy. This positive change in the entropy indicates an overall increase in the degree of the freedom of the system and could be explained by the release of water molecules from the complex surface or represent changes in the conformational degree of freedom of both HSPA1A and the protein substrate [16].

Furthermore, the R36C, I480N, and F592S mutations show different levels of refolding and anti-apoptotic activities. The R36C and F592S mutations, which result in slower ATP hydrolysis and increased substrate-binding reaction entropies, had contrasting refolding properties. Furthermore, both mutations failed to inhibit apoptosis caused by protein-aggregation stress because either their binding to specific anti-apoptotic clients is altered or inhibited [17]. Differently, the I480N mutation, which had ATPase activity comparable to the WT HSPA1A, had almost no refolding activity. These observations imply that there is a fine balance between ATP-hydrolysis and protein binding and any major disturbance of this balance may result in major changes on the chaperoning functions of HSPA1A.

The observations described above suggest that the mutations affect ATP hydrolysis and protein-ligand binding stability because they alter the local conformation of HSPA1A and thus may affect the allosteric communication between the two major domains of the chaperone. The rational of this prediction is based on the fact that the function of Hsp70s is allosterically controlled by ATP hydrolysis and the conformational changes that occur [18, 19]. These findings suggest that even mutations found in the NBD could lead to alterations in substrate binding or protein refolding. This idea is further supported by the finding that the K71A mutation, which minimizes ATP hydrolysis, results in stabilization of HspA8-substrate interactions [20]. These concepts could explain the behavior of the R36C mutation, which although shows lower rates of ATP hydrolysis, it results in increased refolding of denatured luciferase. On the other end of the molecule, the I480N mutation, which results in loss of refolding activity, may also affect the local conformation or alter the hydrogen bonds formed between the SBD-β-subdomain and the NBD, as it has been shown for the aspartic acid in position 481 in the bacterial Hsp70, DnaK [4]. These two variants (R36C and I480N), appear to show how protein refolding and the rate of ATP hydrolysis are in some instances inversely related and reveal for one more time how amino acids from completely different parts of the protein affect the entire allosteric process and alter the cellular stress response.

Although these mutations result in significant functional alterations, both of these variants have very low frequency or are private mutations and thus, are predicted to have very low penetrance in the human population, exemplifying further the notion that HSPA1A is under strong purifying selection that aims to constraint major changes in the function of the molecule.

Highlights.

• Mutations, R36C, I74T, and F592S, lower the rate of ATP hydrolysis of HSPA1A.

• S16P, R36C, I480N, and F592S alter the interaction of HSPA1A and protein substrates.

• R36C increases while the I480N decreases the refolding of heat-denatured luciferase.

• R36C, I480N, and F592S failed to protect cells from aggregated protein stress.

Abbreviations:

3D model

three-dimensional structural protein model

EGFP

enhanced green fluorescent protein

HeLa

Henrietta Lacks' 'Immortal' cells

Hsp70

Seventy-kilodalton heat shock protein

ITC

Isothermal Titration Calorimetry

MEM

Minimum Essential Media

NBD

nucleotide-binding domain

PBS

Phosphate-buffered saline

RFP

red fluorescent protein

SBD

substrate-binding domain

SDS-PAGE

sodium dodecyl sulfate polyacrylamide gel electrophoresis

SNPs

single nucleotide polymorphisms