Frataxin deficiency in Friedreich’s ataxia is associated with reduced levels of HAX-1, a regulator of cardiomyocyte death and survival

Abstract

Frataxin deficiency, responsible for Friedreich’s ataxia (FRDA), is crucial for cell survival since it critically affects viability of neurons, pancreatic beta cells and cardiomyocytes. In FRDA, the heart is frequently affected with typical manifestation of hypertrophic cardiomyopathy, which can progress to heart failure and cause premature death. A microarray analysis performed on FRDA patient’s lymphoblastoid cells stably reconstituted with frataxin, indicated HS-1-associated protein X-1 (HAX-1) as the most significantly upregulated transcript (FC = +2, P < 0.0006). quantitative Reverse Transcription-Polymerase Chain Reaction (qRT-PCR) and western blot analysis performed on (I) HEK293 stably transfected with empty vector compared to wild-type frataxin and (II) lymphoblasts from FRDA patients show that low frataxin mRNA and protein expression correspond to reduced levels of HAX-1. Frataxin overexpression and silencing were also performed in the AC16 human cardiomyocyte cell line. HAX-1 protein levels are indeed regulated through frataxin modulation. Moreover, correlation between frataxin and HAX-1 was further evaluated in peripheral blood mononuclear cells (PBMCs) from FRDA patients and from non-related healthy controls. A regression model for frataxin which included HAX-1, group membership and group^* HAX-1 interaction revealed that frataxin and HAX-1 are associated both at mRNA and protein levels. Additionally, a linked expression of FXN, HAX-1 and antioxidant defence proteins MnSOD and Nrf2 was observed both in PBMCs and AC16 cardiomyocytes. Our results suggest that HAX-1 could be considered as a potential biomarker of cardiac disease in FRDA and the evaluation of its expression might provide insights into its pathogenesis as well as improving risk stratification strategies.

Introduction

Friedreich’s ataxia (FRDA, OMIM#229300) is a rare hereditary disease due to a deficiency of the mitochondrial protein frataxin (FXN) caused in most cases by homozygous hyperexpansion of Guanine Adenine Adenine (GAA) triplets that severely reduces FXN gene transcription (1). FXN deficiency critically affects viability of neurons, pancreatic beta cells and cardiomyocytes in humans (2). Besides the characteristic features of spinocerebellar ataxia, two-thirds of patients develop hypertrophic cardiomyopathy, which often progress to heart failure, finally causing premature death (3). Searching for genes involved in the disease pathogenesis that could be modulated by FXN levels, a microarray analysis was performed in FXN-deficient lymphoblastoid cells derived from a FRDA patient stably reconstituted with wild-type FXN. HS-1-associated protein X-1 (HAX-1) resulted the highest upregulated transcript in the FXN-reconstituted cells compared to empty-vector transfected cells.

HAX-1 comprises a ubiquitously expressed family of proteins with diverse-binding partners that plays key roles in the regulation of cell survival (4). HAX-1 proteins, ranging in size from 26 to 35 kDa, result from complex alternative splicing events. Deregulation of the expression levels and subcellular distribution of HAX-1 is associated with the development and progression of severe diseases, including cancer, psoriasis, congenital neutropenia and neurological disorders (4–6). HAX-1 has been identified as an antiapoptotic protein required for the survival of different cell types such as cerebellar neurons (7) and cardiomyocytes (8). HAX-1 is indeed involved in the coordination of the network assembly of actin structures with KV3.3 channels, the dysfunction of which is correlated with spinocerebellar ataxia 13 (9). Two different HAX-1 isoforms, one proapoptotic (human v4) and the other antiapoptotic (human v1), have been reported to modulate cardiac cell survival and death through homo- or heterodimerization (10). AQ10: These data suggest the involvement of HAX-1 isoforms in apoptosis modulation and cell survival. HAX-1 is highly expressed in the heart (11) where it plays a role in cardioprotection. HAX-1 was shown to undertake an antiapoptotic function in cardiomyocytes by interacting directly with caspase-9 and therefore inhibiting its activity (12). Moreover, HAX-1 was identified as a regulator of contractility and calcium cycling (13). HAX-1 can associate with sarcoplasmic reticulum Ca-ATPase (SERCA2a) (14) and its regulating protein phospholamban (15), which both play a central role in modulating calcium homeostasis, important for cardiac function (16). Interestingly, HAX-1 heterozygous-deficient hearts exhibit increases in infarct size in response to ischemia/reperfusion injury. Conversely, overexpression of HAX-1 improves contractile recovery coupled with reduced myocardial infarction (8). HAX-1 protective effects can be mediated through the modulation of cyclophilin-D levels that further regulate the mitochondrial permeability transition pore opening (17).

Therefore, we sought to examine the antiapoptotic HAX-1.v1 (named HAX-1 throughout the text) expression in FRDA patients. We found a positive correlation between FXN and HAX-1 expression. In addition, a linked expression of FXN, HAX-1 and antioxidant defence proteins MnSOD and Nrf2 was observed both in peripheral blood mononuclear cells (PBMCs) from FRDA patients and from non-related healthy controls and in AC16 human cardiomyocytes. Moreover, a significant modulation of HAX-1 expression by FXN levels in AC16 human cardiomyocytes therefore suggests HAX-1 involvement in the development of cardiac disease in FRDA.

Results

HAX-1 expression is regulated by FXN levels

In this study, we analyzed, by a comparative microarray approach, the transcriptome profile induced by FXN overexpression in FXN-deficient lymphoblastoid cells (GM15850, derived from a FRDA patient) stably reconstituted with wild-type FXN (FXN-cells) compared to GM15850 cells transfected with empty vector (EMPTY-cells). After data normalization, 667 genes were identified as differentially expressed with significant P-value (P < 0.05) (Supplementary Material, Table S1). Among these significantly expressed genes, 91 resulted differentially expressed as either < 0.5 or > 1.5 fold changes (FC, Table 1). Of these differentially expressed genes, only 7 resulted downregulated. Among them, C19orf12 (ID BC004957_1; FC = −2.2; P < 0.05) encodes a small transmembrane protein with unknown function; however, mutations in C19orf12 cause a form of neurodegeneration with brain iron accumulation (NBIA4) (18). Among the upregulated genes, the most significantly upregulated transcript, i.e. clone XM_095613_1 (FC = +2.1; P < 0.0005), was subsequently withdrawn from NCBI database. We thus focused our attention on BC005240_1 transcript (FC = +2; P < 0.0006), which corresponds to HS-1-associated protein X-1 (HAX-1) and analyzed its expression in FRDA.

Table 1

The top 91 differentially expressed genes in GM15850 cells stably reconstituted with frataxin, sorted according to increasing P-values

	ID	Name	FC	P-value
1	XM_095613_1	Similar to polycystic kidney disease 1 homolog; polycystin-1; loc169286	2.1	0.0005
2	BC005240_1	Homo sapiens HCLS1-associated protein X-1	2	0.0006
3	AL138828_3	ba472e5.2.2 (novel protein; isoform 2); ba472e5.2	2	0.0007
4	NM_004852_1	One cut domain; family member 2; onecut2	1.9	0.0007
5	XM_171898_1	Similar to transmembrane activator and CAML interactor; loc255262	1.9	0.0008
6	AB051077_1	Peroxin pex6p; pex6	1.6	0.0009
7	AJ010482_1	Myopodin protein; myopodin	2.1	0.001
8	XM_092643_1	Similar to b29 protein; loc165524	1.9	0.001
9	AK075317_1	cDNA clone moderately similar to Mus musculus gng3lg mRNA; unnamed protein product	2.1	0.001
10	AB062787_1	Triggering receptor trem-2v	2.3	0.0011
11	AY014403_1	Kinesin-like protein rbkin1; rbkin	1.5	0.0013
12	AF241254_1	Angiotensin converting enzyme-like protein	1.8	0.0016
13	XM_070666_1	Similar to junctophilin 1; mitsugumin72; junctophilin type1; jph1	2.1	0.0017
14	XM_171974_1	Similar to zinc finger protein 91 (hpf7; htf10); loc256696	1.7	0.0018
15	BC028101_1	Bruno-like 5; RNA-binding protein (drosophila)	1.5	0.0019
16	XM_063253_1	Similar to NADH dehydrogenase (ubiquinone) 1 alpha subcomplex; 4; loc122667	2.1	0.0021
17	XM_058961_1	Similar to riken cDNA 4021401a16; loc126003	1.6	0.0026
18	XM_167626_1	Hypothetical protein xp_167626; loc220932	1.6	0.0027
19	NM_139120_1	yy1-associated protein; isoform 4; yap	1.5	0.0031
20	AK022552_1	cDNA clone highly similar to Rattus norvegicus rexo70 mRNA; unnamed protein product.	1.5	0.0031
21	XM_106383_1	Hypothetical protein xp_106383; loc166944	1.8	0.0031
22	AK097855_1	cDNA clone unnamed protein product	1.6	0.0032
23	AK074215_1	cDNA clone unnamed protein product	1.5	0.00324
24	AK074686_1	cDNA clone unnamed protein product	1.6	0.0036
25	XM_067012_1	Similar to ptpn13-like; y-linked; testis-specific PTP-BL-related protein on y; loc140016	1.6	0.0041
26	XM_170930_1	Similar to dj830a10.1 (ecotropic viral integration site 5); loc255978	1.6	0.0045
27	XM_171784_1	Similar to cg15057 gene product; loc256313	2	0.0047
28	XM_167035_1	Similar to heparan sulfate d-glucosaminyl 3-o-sulfotransferase 1 precursor; heparin-glucosamine 3-o-sulfotransferase; loc222537	1.6	0.0047
29	XM_115902_1	Similar to poly(a) binding protein; cytoplasmic 1; loc207061	1.5	0.0053
30	AK091802_1	cDNA clone unnamed protein product.	1.6	0.0053
31	XM_097203_1	Hypothetical protein xp_097203; loc147157	1.5	0.0053
32	BC028048_1	Unknown (protein for mgc:40108)	1.6	0.0058
33	XM_166973_1	Similar to b24o18.3 (pom121 membrane glycoprotein (rat homolog)-like 2); loc219549	1.5	0.0058
34	AK056874_1	cDNA clone unnamed protein product.	1.7	0.0062
35	XM_064853_1	Similar to zinc finger protein 347; zinc finger 1111; loc125891	1.6	0.0066
36	XM_119769_1	Hypothetical protein xp_119769; loc205405	1.6	0.0066
37	AL138875_3	ba103j18.1.2 (novel protein; isoform 2); ba103j18.1	1.8	0.0067
38	BC012183_1	Unknown (protein for mgc:20470)	1.7	0.0074
39	XM_065532_1	Similar to hypothetical protein flj13725; kiaa1930 protein; loc130017	1.5	0.0075
40	XM_032161_1	Hypothetical protein xp_032161; loc90494	1.7	0.0079
41	XM_033004_1	Hypothetical protein bc005107; loc90625	1.8	0.0083
42	XM_090444_1	Similar to synaptotagmin 10; loc160699	1.6	0.0085
43	AK091473_1	cDNA clone unnamed protein product	2	0.0086
44	XM_166111_1	Similar to snap-25-interacting protein; loc219672	0.5	0.01
45	AL035464_1	dj1043e3.1 (novel protein); dj1043e3.1	1.6	0.0111
46	AK026702_1	Homo sapiens cDNA: flj23049 fis; clone lng02559; unnamed protein product	2	0.0112
47	AB014569_1	kiaa0669 protein; kiaa0669	0.4	0.0126
48	BC030602_1	Similar to myosin IC	1.8	0.0126
49	AB062487_1	ok/sw-cl.68; ok/sw-cl.68	1.6	0.0134
50	AF123074_1	Cytoplasmic dynein intermediate chain 1	1.8	0.0139
51	BC034327_1	Similar to kiaa1683 protein	1.8	0.0160
52	NM_002258_1	Killer cell lectin-like receptor subfamily b; member 1; klrb1	1.5	0.0167
53	AF254067_1	Interleukin 21 receptor; IL21R	1.6	0.0198
54	AF230904_1	c-Cbl-interacting protein; cin85	0.5	0.0201
55	XM_096879_1	Hypothetical protein xp_096879; loc145828	1.6	0.0212
56	NM_001813_1	Centromere protein e; CENPE	2.5	0.0212
57	XM_173595_1	Hypothetical protein xp_173595; loc254234	0.4	0.0213
58	BC009870_1	Gamma tubulin ring complex protein (76p gene)	1.5	0.0226
59	XM_045632_1	Similar to hypothetical protein flj11292; loc92539	1.5	0.0227
60	XM_116109_1	Similar to cg1364 gene product; loc201615	2.2	0.0229
61	AK091107_1	cDNA clone unnamed protein product.	0.4	0.0252
62	XM_171089_1	Similar to nephronectin; loc255743	1.6	0.0255
63	XM_170392_1	Similar to beta-1;3-galactosyl-o-glycosyl-glycoprotein beta-1;6-n-acetylglucosaminyltransferase (core 2 branching enzyme) (core2-glcnac-transferase) (c2gnt) (core 2 gnt); loc223101	1.7	0.0277
64	XM_167601_1	Similar to knockout; loc219735	1.5	0.0306
65	AF269161_1	c21orf7 form a	1.6	0.0311
66	XM_058513_1	Similar to riken cDNA 4921513o20; loc120892	1.5	0.0320
67	XM_071224_1	Similar to putative gene with similarity to zinc finger proteins; loc140160	1.6	0.0321
68	BC020955_1	Similar to hypothetical protein flj11046	1.7	0.0323
69	BC031660_1	Similar to loc166075	1.8	0.0336
70	NM_015847_1	Methyl-cpg binding domain protein 1; isoform pcm1; mbd1	1.6	0.0339
71	XM_066779_1	Similar to sarcoma antigen; putative tumor antigen; loc139587	1.6	0.0341
72	BC004957_1	Similar to riken cDNA 1600014c10 gene	0.5	0.0345
73	BC012184_1	Unknown (protein for mgc:20471)	1.7	0.0347
74	AK095926_1	cDNA clone unnamed protein product.	1.5	0.0373
75	XM_067647_1	Similar to 60s ribosomal protein l23a; loc132066	1.6	0.0383
76	BC015338_1	Unknown (protein for mgc:21282)	1.7	0.0383
77	XM_166207_1	Similar to riken cDNA 2 410 018 m14; loc219541	1.6	0.0388
78	XM_094365_1	Similar to early b-cell factor (olfactory neuronal transcription factor 1); loc167252	1.5	0.0392
79	D01038_1	vla-3 alpha subunit	1.5	0.0393
80	BC007653_1	Unknown (protein for image:3610013)	1.6	0.0399
81	NM_022442_1	Ubiquitin-conjugating enzyme e2 variant 1; isoform c; ube2v1	1.6	0.0409
82	AF268387_1	Laryngeal carcinoma-related protein 1	1.6	0.0416
83	XM_168471_1	Hypothetical protein xp_168471; loc222182	1.5	0.0426
84	XM_088886_1	Similar to c4b-binding protein alpha chain precursor (c4bp) (proline-rich protein) (prp); loc163426	1.7	0.0430
85	XM_116262_1	Similar to platelet basic protein precursor (PBP) (small inducible cytokine b7) (cxcl7); loc201960	0.4	0.0431
86	XM_097883_1	Hypothetical protein xp_097883; loc150273	1.7	0.0440
87	XM_118021_1	Hypothetical protein xp_118021; loc203766	1.5	0.0441
88	AK001268_1	cDNA clone unnamed protein product	1.8	0.0447
89	NM_139238_1	adam-ts-related protein 1; isoform 1; adamtsl1	1.5	0.0471
90	AF116720_1	pro3015	1.6	0.0487
91	AF319633_1	Putative potassium channel kiaa0027	1.7	0.0490

GenBank identity number (ID) and FC in gene expression are also displayed.

We first confirmed by qRT-PCR the microarray result and further assessed FXN and HAX-1 protein levels in FXN cells and EMPTY cells (Fig. 1A).

Figure 1

HAX-1 expression is regulated by FXN levels. Overexpression of FXN upregulates HAX-1 in GM15850 FRDA patient-derived B cells (A) and in HEK293 cells (B). (A) Left panels: mRNA relative quantity was analyzed in immortalized GM15850 FRDA patient-derived B cells stably reconstituted with empty vector (EV) and wild-type frataxin (FXN) by qRT-PCR using the housekeeping invariant reference gene TUBULIN. Values are means ± SEM (n = 2). Center panel: Expression of FXN, HAX-1 and Tubulin as loading control was analyzed by western blot. Data are representative of two independent experiments. Right panels: Densitometric quantification. FXN and HAX-1 protein expression were normalized with Tubulin. Data represent the mean ± SEM of two independent experiments. (B) Left panels: mRNA level was analyzed in HEK293 cells stably reconstituted with empty vector (EV) and wild-type frataxin (FXN) by qRT-PCR using the housekeeping invariant reference gene TUBULIN. Values are means ± SEM (n = 2). Center panel: Expression of FXN, HAX-1 and Tubulin as loading control was analyzed by western blot. Data are representative of two independent experiments. Right panels: Densitometric quantification. HAX-1 and FXN protein expression were normalized with Tubulin. Data represent the mean ± SEM (n = 2). (C). HEK293 cells were transiently transfected with plasmid for FXN gene interference (shFXN, +) or empty vector (−) every 2 days. ATP synthase (ATP S.), FXN and HAX-1 were analyzed after the fourth (IV) round of transfection. Data are representative of 3 independent experiments. (D) FXN and HAX-1 expression analyzed in C were normalized with ATP synthase. Values are means ± SEM (n = 3). P-values were calculated with Student’s t-test (^*P < 0.05; ^**P < 0.01, ^****P < 0.0001).

Furthermore, FXN and HAX-1 expression were analyzed both at mRNA and protein levels in HEK293 stably transfected with empty vector or wild-type FXN. Overexpression of FXN increases HAX-1 both at mRNA and protein levels (Fig. 1B). Conversely, overexpression of HAX-1 did not alter FXN levels (Supplementary Material, Fig. S1).

To further address this relationship, we also interfered with FXN expression using a shRNA-based approach in HEK293 cell line achieving a 55–60% FXN knockdown after four rounds of transient transfection. FXN silencing could reduce HAX-1 expression up to 34% (Fig. 1C and D).

Modulating FXN levels in cardiomyocytes regulate HAX-1 protein levels and their sensitivity to H₂O₂ treatment

Considering that the heart represents one of the most affected tissues in Friedreich’s ataxia and that HAX-1 is involved in cardioprotection, we subsequently analyzed the association between FXN and HAX-1 protein levels in human cardiac cells. AC16 cardiomyocytes were transiently transfected with wild-type FXN or empty vector. Overexpression of FXN results in a significant 2-fold upregulation of HAX-1 expression (Fig. 2A and B). Moreover, FXN silencing shows that a reduction of FXN protein levels after four rounds of ShFxn plasmid transfection results in decreased HAX-1 expression of approximately 55% (Fig. 2C and D). These data suggest that the coregulation of FXN and HAX-1 is present also in human cardiomyocytes, thus evoking a common pathway active in the heart.

Figure 2

Modulating frataxin levels in cardiomyocytes regulate HAX-1, Nrf2 and MnSOD protein levels and their sensitivity to H₂O₂ treatment. (A) AC16 cardiac cells were transiently transfected with pIRES2-FXN (FXN) or empty vector (EV). Cells extracts were collected 24- and 72-h post-transfection. Representative blot and corresponding stain-free gel used for FXN and HAX-1 protein (at 72 h) normalization of three independent experiments are shown. (B) Nrf2 (24 h) and MnSOD (72 h) were normalized with ATP synthase (ATP S.). FXN and HAX-1 expression analyzed in A were normalized using stain-free technology. Values are means ± SEM (n = 3). (C) AC16 cardiac cells were transiently transfected with plasmid for FXN gene interference (shFXN, +) or empty vector (−) every 2 days. FXN, HAX-1, Nrf2 and MnSOD were analyzed after the fourth (IV) round of transfection. Representative blot and corresponding stain-free gel used for FXN and HAX-1 protein normalization of three independent experiments are shown. (D) Nrf2 and MnSOD were normalized with ATP synthase (ATP S.). FXN and HAX-1 expression analyzed in (C) were normalized with using stain-free technology. Values are means ± SEM (n = 3). (E) AC16 cardiac cells were transiently transfected with plasmid for frataxin gene interference (shFXN) or empty vector (pSUPER) each 2 days. After three rounds of transfections, AC16 were additionally transfected with empty vector pcDNA3.1 or HAX-1. At 24 h post transfection, cells were treated with 0.5 mm of H₂O₂ for 2 h. Cell viability was measured by countess staining selectively the dead cells with Trypan Blue. Relative cell viability was expressed as mean ratio of the number of living cells in H₂O₂-treated sample versus the number of cells seeded. Values are means ± SEM (n = 2). P-values were calculated with Student’s t-test (^*P < 0.05; ^**P < 0.01).

FRDA-derived cells have been reported to have disabled antioxidant responses such as the ones mediated by MnSOD (manganese-dependent superoxide dismutase) (19,20) and Nrf2 (nuclear factor erythroid derived 2-like 2) (21). We therefore analyzed whether these antioxidant proteins could be regulated by FXN modulation in cardiomyocytes. Indeed, upon FXN overexpression, HAX-1, Nrf2 and MnSOD levels are increased. Intriguingly, Nrf2 upregulation anticipates the increase of HAX-1 levels (Fig. 2A and B). Silencing of FXN upon four rounds of ShFxn plasmid transfection also resulted in a decreased expression of 57% of Nrf2 and 30% of MnSOD levels (Fig. 2C and D).

We then analyzed the antioxidant response of cardiomyocytes upon treatment with H₂O₂ evaluating whether HAX-1 transfection following FXN silencing could affect cell survival. Interestingly, we observed that HAX-1 overexpression could compensate the toxicity of H₂O₂ treatment in cardiomyocytes in which FXN was silenced, thus suggesting that HAX-1 can hamper the oxidative stress in FXN-deficient cells (Fig. 2E).

Low levels of FXN in FRDA-derived cell lines and PBMCs correspond to reduced levels of HAX-1

To corroborate the link between FXN and HAX-1 in FRDA cells, their expression level was analyzed in 11 FXN-deficient lymphoblastoid cell lines derived from FRDA patients and their unaffected carrier relatives. A significant decrease of HAX-1 both at mRNA and protein levels was confirmed in FRDA cell lines (Fig. 3).

Figure 3

Low levels of FXN in FRDA cell lines correspond to reduced levels of HAX-1. FXN and HAX-1 mRNA and protein levels were analyzed in 11 lymphoblastoid cell lines derived from FRDA patients compared to 11 clinically unaffected heterozygous relatives. The mean ratio (Patients/Controls) is illustrated. (A) FXN and HAX-1 mRNA expression (2^-ΔCt) analysis was performed with one sample t-test (^*P < 0.05, ^****P < 0.0001), testing the null hypothesis of a mean = 1. Values are means of independent experiments (n = 2–5). Both overall mean and data range are illustrated. (B) FXN and HAX-1 protein expression were normalized with ATP synthase. Values are means of independent experiments (n = 3). Both overall mean and data range are illustrated. Analysis was performed with one sample t-test (^*P < 0.05, ^****P < 0.0001), testing the null hypothesis of a mean = 1.

Subsequently, we obtained and assessed baseline data for 53 patients with Friedreich’s ataxia [27 (51%) females and 26 (49%) males], with ages ranging from 20 to 72 years (mean 39 years ±12). Fifty of 53 patients (94%) were homozygous for expanded GAA repeats in FXN gene, with the larger allele containing at least 200 GAA triplets (Table 2); the remaining 3 patients (6%) were compound heterozygous, containing an expanded GAA repeat on one allele and an FXN point mutation on the other. We excluded these latter from further subgroup expression analyses. Twenty-two patients (44%) had early-onset, 16 (32%) intermediate-onset and 12 (24%) late-onset disease (22). Twenty patients (38%) developed a hypertrophic cardiomyopathy (Table 2 and Supplementary Material, Table S2).

Table 2

Characteristics of FRDA patients

FRDA patients (N)	53
Sex (F/M)	27 (51%)/26 (49%)
Age (mean/range)	41 years (20–72)
Expanded GAA repeats in the larger allele (mean/range)	780 (200–1200)
Cardiac disease (N/%)	20 (38%)

In order to evaluate HAX-1 expression as a potential biomarker in FRDA, FXN and HAX-1 expression were analyzed in PBMCs from these FRDA patients and non-related healthy controls. For mRNA expression analysis, values were available for 39 patients (74%, mean age 42 years ±13) and 23 non-related healthy controls (mean age 41 years ±13). Data for protein expression were available for 41 patients (77%, mean age 39 years ±13) and 27 non-related healthy controls (mean age 41 years ±12). We observed that FXN and HAX-1 expression are coregulated, both at mRNA (Fig. 4A and B) and protein levels (Fig. 4D and E).

Figure 4

FRDA PBMCs express lower levels of HAX-1. FXN and HAX-1 expression were analyzed in PBMCs derived from FRDA patients (Patients) and healthy individuals who did not exhibit any type of heart problems (Controls). mRNA expression was normalized with ATP SYNTHASE (2^-△Ct) in (A, B) Patients (n = 39) and Controls (n = 23). Values are means of independent experiments (n = 2–5). Both overall mean and data range are illustrated. FXN and HAX-1 protein expression were normalized with ATP Synthase in (D, E) Patients (n = 41) and Controls (n = 27). Values are means of independent experiments (n = 2–5). Both overall mean and data range are illustrated. Scatter plot showing Frataxin (FXN) versus HAX-1 mRNA (C) and protein (F) values across all subjects. Dashed line: Fitted linear regression model. R²: R squared value of the regression model. r = Pearson correlation. P = P-value related to the slope of the regression model. (G) FXN and HAX-1 mRNA levels were analyzed in 13 FRDA patients with hypertrophic cardiomyopathy (HCM), 26 FRDA patients without hypertrophic cardiomyopathy (w/o HCM) and 23 healthy individuals who did not exhibit any type of heart problems (Controls). Values are means of independent experiments (n = 2–5). Both overall mean and data range are illustrated. (H) FXN and HAX-1 protein levels were analyzed in 18 FRDA patients with hypertrophic cardiomyopathy (HCM), 23 FRDA patients without hypertrophic cardiomyopathy (w/o HCM) and 27 healthy individuals who did not exhibit any type of heart problems (Controls). Values are means of independent experiments (n = 2–5). Both overall mean and data range are illustrated. Analysis was performed with Student’s t-test (^**P < 0.01, ^***P < 0.001, ^****P < 0.0001).

Regression analysis showed a statistically significant association between FXN and HAX-1 both at mRNA and protein levels (Fig. 4C and F). In particular, when analyzing a possible mRNA-level association, we found a statistically significant, positive relationship between FXN and HAX-1 (R² = 0.55, P < 0.05, Pearson r = 0.54) (Fig. 4C). This association was confirmed, with an even higher effect size (i.e. correlation), at protein level, where we again found a statistically significant, positive relationship between FXN and HAX-1 (model R² = 0.6, P < 0.001, Pearson r = 0.81) (Fig. 4F). The difference of FXN and HAX-1 expression in FRDA patients with and without hypertrophic cardiomyopathy is not statistically significant, although the HAX mRNA expression is lower in FRDA patients with hypertrophic cardiomyopathy compared to healthy controls (Fig. 4G and H).

We also analyzed whether these FRDA-derived PBMCs could have altered expression of antioxidant enzymes MnSOD and Nrf2. A significant difference was observed only for MnSOD protein expression, which is expressed at lower levels in FRDA patients compared to healthy individuals (Supplementary Material, Fig. S2A). In addition, we found a statistically significant, positive relationship between FXN and Nrf2 (P < 0.05, Spearman r = 0.38) as well as between HAX-1 and Nrf2 (P < 0.01, Spearman r = 0.41, Supplementary Material, Fig. S2B).

Discussion

This study shows for the first time a positive correlation between FXN and HAX-1 expression both at protein and mRNA levels in FRDA. Interestingly, low levels of FXN in both FRDA cell lines and PBMCs of FRDA patients correspond to reduced levels of HAX-1.

Cardiac dysfunction may develop early in the course of FRDA and is recognized as the leading cause of death in patients (23). Both FXN and HAX-1 are crucial for cell survival. Indeed, FXN KO in mice is lethal (24) and overexpression of FXN in human cells protects them from apoptotic stimuli (25). Conversely, HAX-1 is highly expressed in the heart and is involved in protection of cardiomyocytes from injury (8). Cardiomyocytes are extremely sensitive to hypoxia and subsequent oxidative stress-induced damage. It is noteworthy that HAX-1 increases in hypoxic environment (26) as well as FXN that indeed participates in the stress response of tumor adaptation to hypoxia (27). Furthermore, under oxidative stress conditions, the correct assembling of actin fibers is impaired in FRDA patients thus preventing the translocation of nuclear factor (erythroid derived 2)-like 2 (Nrf2) transcription factor into the nucleus and consequently hampering the antioxidative responses (28). Moreover, HAX-1 downregulation in different types of cells results in increased apoptosis due to oxidative stress (17,29,30). We have observed that modulation of FXN in AC16 cardiomyocytes regulates consequently HAX-1 levels but also Nrf2 and MnSOD antioxidant protein levels. All these proteins are upregulated upon overexpression of FXN or downregulated upon FXN silencing. Interestingly, we observed that HAX-1 overexpression could compensate for the toxicity of H₂O₂ treatment in cardiomyocytes in which FXN was silenced. Therefore, HAX-1 upregulation could rescue the impaired antioxidant cellular responses caused by FXN deficiency. The linked expression of FXN, HAX-1 and antioxidant proteins MnSOD and Nrf2 was also verified ex vivo in PBMCs. Although we could only detect statistically significant differences in MnSOD expression between healthy controls and FRDA patients, an interesting correlation between Nrf2-FXN-HAX-1 was observed (Supplementary Material, Fig. S2), underlining the already known importance of antioxidant defects in FRDA, and especially because Nrf2 is considered a therapeutic target for this disease (31,32).

In addition, since in vivo data on rat models and in vitro overexpression experiments suggest an antagonistic role for HAX1 variant II (HAX-1 v4) (10), we also performed a quantitative analysis of this proapoptotic HAX1 variant in all FRDA cells previously analyzed for HAX-1 v1 levels. However, both in healthy controls and FRDA cells, HAX-1 v4 mRNA expression is about 10-fold less than variant v1 (Supplementary Material, Fig. S3). Finally, HAX1 v1/HAX-1 v4 mRNA expression ratio is quite similar in control and FRDA cells (Supplementary Material, Fig. S3). Our data therefore do not support the hypothesis of an imbalance between these two isoforms as the major cause of apoptosis modulation and cell survival due to FXN deficiency.

As HAX-1 is required for the survival of cerebellar neurons (7) and involved in the spinocerebellar ataxia 13 through a physical interaction with KV3.3 channels (9), we therefore also investigated whether a direct interaction between HAX-1 and FXN would explain the pathogenic mechanism in FRDA. However, we could not observe HAX-1/FXN interactions neither in HEK293 cells nor in AC16 cardiomyocytes (Supplementary Material, Fig. S4). Hence, all together these data suggest that FXN could contribute to regulate the cellular oxidative resistance by modulating HAX-1 levels, without a direct interaction between these proteins.

Considering that both FXN and HAX-1 protect from apoptotic stimuli and that dysregulated apoptosis underlies the progression of cardiomyopathies and neurodegeneration, low levels of FXN and consequently of HAX-1 could contribute to the development of hypertrophic cardiomyopathy as well as neuronal deficit, important features in FRDA. Interestingly, in this regard, several microRNAs have been described as modulators of FXN and putative biomarkers for cardiomyopathy (33–38). Further investigation would be necessary to clarify whether and which specific epigenetic mechanisms could coordinate FXN and HAX-1 expression in neurons.

The identification of new disease biomarkers associated to cardiac alteration could be important to define a risk prediction and reduce morbidity and mortality. Additional analyses of other diagnostic features of cardiac disease in a larger group of FRDA patients would allow to define HAX-1 as a potential prognostic biomarker, when found associated with clinical benefit. Finally, it also warrants further investigation to evaluate its use as a possible therapeutic target of cardiac disease.

Materials and Methods

DNA constructs

The construct pIRES2-FXN^(1–210) containing human FXN cDNA cloned into the bicistronic expression vector pIRES2-EGFP (Clontech Laboratories) and the shRNA construct (shFxn) were generated in this laboratory as previously reported (25,27). The construct pcDNA3.1 + N-HA-HAX-1 v1 (clone ID: OHu19870) was provided by GenScript (Piscataway, NJ, USA).

Cell cultures and transfections

Human embryonic kidney cells (HEK-293 Flp-In) stably expressing FXN^(1–210) were previously described (39) and cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS). AC16 human cardiomyocyte cell line (Merck Millipore, CA, USA) were cultured in DMEM-F12 medium supplemented with 12.5% FBS.

To overexpress the wild-type FXN and HAX-1 v1, AC16 cardiac cells were respectively transfected with pIRES2-FXN and pcDNA3.1 + N-HA-HAX-1 v1 using Lipofectamine 2000 (Invitrogen), following the manufacturer’s instructions.

To interfere with FXN expression, HEK-293 cells and AC16 cardiac cells were transfected with shRNA plasmid for FXN gene (shFXN) using Lipofectamine 2000 (Invitrogen), following the manufacturer’s instructions. Four rounds of transfection of HEK-293 cells every 2 days were performed to achieve a 55–60% FXN knockdown, and AC16 cardiac cells were transfected until four rounds of transfection to achieve a 55% of FXN reduction.

Immortalized GM16798, GM16223, GM16216, GM16228, GM14518, GM16210, GM16214, GM16203, GM16243, GM16205 and GM15850 FRDA patient-derived B cells and immortalized GM16241, GM16236, GM16213, GM16229, GM14519, GM16212, GM16215, GM16202, GM16237, GM16239 and GM15849 lymphoblasts from clinically unaffected heterozygous parents were obtained from NIGMS Human Genetic Cell Repository, Coriell Institute for Medical Research (Camden, NJ, USA) and cultured in RPMI 1640 supplemented with 15% FBS.

Measurement of cell proliferation and viability using cell counting

For cell viability assay, AC16 cardiac cells were transiently transfected with plasmid for frataxin gene interference (shFXN) or empty vector (pSUPER) each 2 days. After three rounds of transfections, AC16 were additionally transfected with empty vector pcDNA3.1 or pcDNA3.1 + N-HA-HAX-1 v1. 24 h post transfection, 1.4 × 10⁵ cells were seeded in 12-well plates and then treated with 0.5 mm of H₂O₂ for 2 h. Cell viability was measured by Countess Automated Cell Counter (Thermo Fisher Scientific) staining selectively the dead cells with Trypan Blue.

Microarray analysis and processing

Total RNA of GM15850 FRDA patient-derived B cells stably reconstituted either with empty vector or FXN was extracted and purified using TriZol reagent (Thermo Fisher Scientific). RNA quality and quantity was assessed using a Nanodrop spectrophotometer (Thermo Scientific) and agarose gel electrophoresis. Methods for sample processing and microarray experiments have been previously described (40). Briefly, direct labeling of second strand (ss) cDNA was conducted according to manufacturer’s instructions (Invitrogen) with starting material of 20 μg of total RNA. The amino-allyl labeled dNTP mix was added to the reaction to generate amino-allyl-labeled ss cDNA. PCR purification kit (Qiagen) was used to remove unincorporated dNTPs, fluorescent dyes and primer. Alexa555- and Alexa647-labeled cDNA was then combined in one tube and evaporated almost to dryness. The combined labeled cDNAs were stored on ice protected from light until hybridization. Hybridization was performed in duplicate, with one dye swap and slides were incubated in a water bath at 42°C for an overnight in the dark.

We used the Human 40K A OciChip™ array, consisting of 20 000 genes with full functional characterization and content referencing and the Human 40K B OciChip™ array consisting of genes referenced against several NCBI databases (www.ocimumbio.com). Each slide was scanned on the GenePix 4000B Microarray Scanner at the optimal wavelength for the Alexa555 (F532) and Alexa647 (F635) dyes. Each spot was automatically segmented; total intensities as well as the fluorescence ratios of the two dyes for each spot were then calculated. The spots were flagged when they exhibited poor hybridization signals and when they were saturated (F635 or F532 median = 65 535). Flagged spots were filtered and systematic bias in the data was removed by applying dye-swap and printiploess intra-array normalization followed by inter-array quantile normalization. Dye-swap normalization makes use of the reverse labeling in the two microarray replicates directly (41). To establish the significance of observed regulation for each gene, we used one sample t-test. Finally, only genes with a satisfactory effect (FC ≥ ±1.5) were considered. The expression data of microarray experiments are available as a specific GEO Sample record and may be linked as follows: http://www.ncbi.nlm.nih.gov/geo/query/GSE108200.

FRDA patient enrollment and characteristics

FRDA patients were recruited from three different centers: University of Rome ‘Tor vergata’, Sapienza University of Rome and Istituto Neurologico Carlo Besta. Inclusion criteria were (a) genetic diagnosis of FRDA, (b) age ≥ 18 years and (c) available echocardiography and electrocardiogram evaluations. Exclusion criteria were the presence of active substance abuse, hematological disorders or major comorbidities requiring chronic pharmacological treatment, not related to FRDA. All the enrolled patients gave informed consent prior to the inclusion in the study. The investigation is conformed to the principles outlined in the Declaration of Helsinki and approval was granted by Institutional Ethics Committee (n.47/16 and n.56).

Isolation of mononuclear cells from peripheral blood

PBMCs from FRDA patients and healthy individuals were isolated using density-gradient medium. For western blot analysis, fresh blood was diluted in equal volume of PBS, gently layered over top of the Lympholyte (Lympholyte-H; Cedarlane Laboratories) and then centrifuged at 750g for 30′ at room temperature (RT) without brake. The enriched mononuclear cell layer was pipetted off in a new tube and washed two times with PBS. To remove the contaminating red blood cells, the cell pellet was resuspended in 2 ml of ACK lysing buffer (Lonza). Cells were incubated at RT for 5′, gently mixed every minute, then centrifuged at 300g for another 5′ and subsequently washed three times with PBS.

For RNA expression analysis, fresh blood was gently layered on equal volume of Ficoll-Paque PLUS (GE Healthcare) and then centrifuged at 400g for 40′ at RT without brake. The mononuclear cell layer was transferred in a new tube and washed two times with PBS. The cell pellet was resuspended in 1 ml of TRIzol (Thermo Fisher Scientific).

Western blot analysis

Total cell extracts were prepared in ice-cold modified RIPA buffer (10 mmol/l sodium phosphate, pH 7.2, 150 mmol/l NaCl, 1% Na deoxycholate, 0.1% SDS, 1% Igepal CA-630 and 2 mmol/L EDTA) or IP buffer (50 mmol/l Tris–HCl, pH 7.5, 150 mmol/l NaCl, 1% Igepal CA-630, 5 mmol/l EDTA, 5 mmol/l EGTA) supplemented with complete protease inhibitor cocktail (Roche Diagnostics, Milan, Italy). About 50 or 100 μg of the total cell lysates were resolved by SDS-PAGE and analyzed by immunoblot with specific mAb anti-frataxin (Abcam, ab110328), mAb anti-HAX-1 (BD Biosciences, 610 825), mAb anti-Nrf2 (Abcam, ab62352), pAb anti-MnSOD (Enzo Life Science), mAb anti-ATP synthase β (BD Biosciences, 612 519), mAb anti-Tubulin (Sigma, T9026) and secondary antibody horseradish peroxidase-conjugated goat anti-mouse (Pierce) using ECL system detection (GE Healthcare Europe GmbH, Milan, Italy). Densitometric analyses were performed using ImageLab software (Biorad). FXN and HAX-1 protein quantification on AC16 human cardiomyocyte cell line was carried out using stain-free technology (42), where the volume density of each target protein band was normalized to the total protein loaded into each lane using ImageLab software (Biorad).

Quantitative RT-PCR analysis

Total RNA was extracted from HEK-293 Flp-In cells, immortalized B cells and PBMCs using TRIzol reagent (Thermo Fisher Scientific) according to the manufacturer’s instructions. Then, 1.5 μg of total RNA was retrotranscribed in cDNA using the High Capacity cDNA Reverse Transcription Kit (Thermo Fisher) according to the manufacturer’s instructions. Primers to quantify mRNA expression were designed to span exon/intron junctions to avoid amplification of genomic DNA by using Primer blast (https://www.ncbi.nlm.nih.gov/tools/primer-blast). The primers sequences were as follows: FXN-Fw: 5′-CACCCAGGCTCTCTAGATG-3′ and FXN-Rev: 5’-CACCACTCCCAAAGGAGAC-3′; HAX1.v1-Fw: 5’-CTGGACCTCGGAGCCACAGA-3′ and HAX1.v1-Rev: 5’-GTTCCCACGGCCCCATGAG-3′.; α4-TUBULIN-Fw: 5-‘GAGCACTCAGACTGTGCCTT-3′ and Rev: 5’-CGATTGAGGTTGGTGTAGGT-3′; ATP-SYNTHASE-Fw: 5′-CTCAGCCATTCCAGGTTGC-3′ and Rev: 5′-CATATTCACCTGCCAAAATCTG -3′. HAX1 primers were variant-specific for full-length or splice variant 1 (HAX1.v1, NM_006118.3). α4-TUBULIN and ATP-SYNTHASE have been used for data normalization in HEK-293 and GM15850 or immortalized lymphoblasts and PBMC, respectively. Each qRT-PCR analysis has been performed in triplicate for at least two times. To evaluate the expression level of FXN and HAX1.v1 splice variants, the threshold cycle (Ct) values of each sample has been extracted with the Real Time PCR 7500 Software (Applied Biosystems, Waltham, MA, USA). The Ct values express the relative measure of the concentration of an mRNA target in the PCR reaction. The normalization of the Ct values with the Ct of a selected housekeeping gene, allows the calculation of the 2^-ΔCt value for each sample. The 2^-ΔCt value represents the absolute expression level of the target gene and it is shown in the graphs.

Statistical analysis

Kolmogorov–Smirnov test has been used to analyze the distribution of the data, which are expressed as median and range. For protein analysis, the significance of differences between populations of data was assessed according to the Student’s t-test with a level of significance of at least P ≤ 0.05. For mRNA expression, data from HEK-293, GM15850 and PBMCs were normally distributed so Student’s t-test has been used; for data not normally distributed (immortalized lymphoblasts), Mann–Whitney test for comparisons between two groups has been used. Significance was set at P ≤ 0.05. Statistical analyses were performed using GraphPad Prism ver. 6.0 (GraphPad Software, San Diego, CA, USA).

In order to examine putative associations between frataxin (FXN) versus HAX-1 values, we employed a linear regression model where FXN was modeled as a function of HAX-1 as well as an offset term. The association was considered statistically significant if the P-value associated with the slope of the regression model was P < 0.05. The effect size was further examined through the Pearson correlation coefficient.

Acknowledgements

We thank all colleagues in our laboratory for helpful discussion.

Funding

Non Communicable Disease (NCDS-2013-00000333 to G.N.); European Research Council (Advanced Grant number 293699, FAST to R.T.); National Ataxia Foundation and Mission Sustainability (CARDIMIRAX, CUP: E81I18000370005) to F.M.; Telethon (grant number GGP15004 to I.C.).

Conflict of Interest statement

None declared.

Author notes

The authors wish it to be known that, in their opinion, the first Francesca Tiano and Francesca Amati authors should be regarded as joint First Authors.