Tbx20 controls the expression of the KCNH2 gene and of hERG … · 2017-01-12 · Tbx20 controls the expression of the KCNH2 gene and of hERG channels Ricardo Caballeroa,b,1, Raquel

Tbx20 controls the expression of the KCNH2 gene andof hERG channelsRicardo Caballeroa,b,1, Raquel G. Utrillaa,b,1, Irene Amorósa,b, Marcos Matamorosa,b, Marta Pérez-Hernándeza,b,David Tinaqueroa,b, Silvia Alfayatea,b, Paloma Nieto-Marína,b, Guadalupe Guerrero-Sernac,d, Qing-hua Liuc,d,Roberto Ramos-Mondragónc,d, Daniela Ponce-Balbuenac,d, Todd Herronc,d, Katherine F. Campbellc,d,David Filgueiras-Ramae,2, Rafael Peinadob,e, José L. López-Sendónb,e, José Jalifec,d,f, Eva Delpóna,b,3,4,and Juan Tamargoa,b,3

aDepartment of Pharmacology, School of Medicine, Instituto de Investigación Sanitaria Gregorio Marañón, Universidad Complutense, 28040 Madrid, Spain;bCentro de Investigación Biomédica en Red (CIBER), Spain; cDepartment of Internal Medicine, Center for Arrhythmia Research, University of Michigan, AnnArbor, MI 48109; dDepartment of Molecular and Integrative Physiology, Center for Arrhythmia Research, University of Michigan, Ann Arbor, MI 48109;eDepartment of Cardiology, Hospital Universitario La Paz, Instituto de Investigación Sanitaria La Paz (IdiPaz), 28046 Madrid, Spain; and fCardiac ArrhythmiaDepartment, Fundación Centro Nacional de Investigaciones Cardiovasculares (CNIC), 28029 Madrid, Spain

Edited by Richard W. Aldrich, The University of Texas at Austin, Austin, TX, and approved November 23, 2016 (received for review July 27, 2016)

LongQT syndrome (LQTS) exhibits great phenotype variability amongfamily members carrying the same mutation, which can be partiallyattributed to genetic factors. We functionally analyzed the KCNH2(encoding for Kv11.1 or hERG channels) and TBX20 (encoding for thetranscription factor Tbx20) variants found by next-generation sequenc-ing in two siblings with LQTS in a Spanish family of African ancestry.Affected relatives harbor a heterozygous mutation in KCNH2 that en-codes for p.T152HfsX180 Kv11.1 (hERG). This peptide, by itself, failed togenerate any current when transfected into Chinese hamster ovary(CHO) cells but, surprisingly, exerted “chaperone-like” effects over na-tive hERG channels in both CHO cells and mouse atrial-derived HL-1cells. Therefore, heterozygous transfection of native (WT) andp.T152HfsX180 hERG channels generated a current that was indistin-guishable from that generated by WT channels alone. Some affectedrelatives also harbor the p.R311Cmutation in Tbx20. In human inducedpluripotent stem cell-derived cardiomyocytes (hiPSC-CMs), Tbx20 en-hanced human KCNH2 gene expression and hERG currents (IhERG)and shortened action-potential duration (APD). However, Tbx20 didnot modify the expression or activity of any other channel involvedin ventricular repolarization. Conversely, p.R311C Tbx20 did not in-crease KCNH2 expression in hiPSC-CMs, which led to decreased IhERGand increased APD. Our results suggest that Tbx20 controls the expres-sion of hERG channels responsible for the rapid component of thedelayed rectifier current. On the contrary, p.R311C Tbx20 specificallydisables the Tbx20 protranscriptional activity over KCNH2. Therefore,TBX20 can be considered a KCNH2-modifying gene.

Tbx20 | hERG channels | long QT syndrome | cardiomyocytes |human induced pluripotent stem cells

Long QT syndrome (LQTS) is characterized by abnormal pro-longation of the QT interval of the electrocardiogram (ECG)

and is due to delayed ventricular repolarization. LQTS increases theoccurrence of ventricular tachyarrhythmias, particularly torsade depointes, leading to recurrent syncope, seizures, ventricular fibrilla-tion, and sudden cardiac death (SCD) (1). At least 15 genes havebeen reported in autosomal-dominant forms of LQTS (1). How-ever, mutations in KCNQ1 (LQT1), KCNH2 (LQT2), and SCN5A(LQT3) represent the most frequent forms of LQTS (∼90%) (1, 2).KCNH2 encodes Kv11.1, or hERG, channels, which generate the

rapid component of the delayed rectifier current (IKr) responsible forventricular repolarization in humans (3). In a Spanish family of Af-rican ancestry suffering LQTS, we identified a frameshift and amissense mutation in KCNH2 that were assumed to be the disease-causing mutations. However, in some family members, we alsoidentified a missense mutation in TBX20 coding for the transcriptionfactor Tbx20, which is necessary in early stages of heart development(4). Importantly, results in flies and mice demonstrated that Tbx20 isalso required for maintaining adult heart function (5, 6).

Here we have tested the KCNH2 and TBX20 mutations toestablish whether they can account for prolongation of re-polarization. Our results demonstrated that more than “one hit” isnecessary to give rise to LQTS in the affected relatives. Moreover,data reveal that the peptide resulting from the KCNH2 frameshiftmutation exerts chaperone-like effects by increasing the membraneexpression of WT hERG channels. Conversely, the p.R311C Tbx20mutation specifically and markedly decreases KCNH2 expression.Therefore, our genetic and functional studies suggest that Tbx20controls the expression of hERG channels in human myocytes and,thus, TBX20 may be considered a KCNH2-modifying gene.

ResultsThe proband (II:4; Fig. 1A) was a 41-y-old male who experiencedsyncope when he got out of bed. The ECG showed sinus rhythm at68 beats per min (bpm) with normal PR (168 ms) and QRS (88 ms)but with a low-amplitude and wide T wave (Fig. 1B). Bazzett-cor-rected QT value (QTc) was 480 ms. Echocardiography, exercise

Significance

Tbx20 is a transcription factor whose critical role in cardiogenesisis well-established. Here we functionally analyzed the electro-physiological effects produced by a mutation (p.R311C) in Tbx20found in some affected individuals belonging to a family withlong QT syndrome (an inherited cardiac arrhythmia due todelayed ventricular repolarization). We demonstrated that Tbx20selectively increases the expression of KCNH2, which encodes forthe channel Kv11.1 (hERG) that generates the main ventricularrepolarizing current. Conversely, the p.R311C mutation disablesthe Tbx20 protranscriptional activity over KCNH2, leading to adecrease in the hERG current and a prolongation of the actionpotentials recorded in human induced pluripotent stem cell-derivedcardiomyocytes. Therefore, we propose that Tbx20, besides itsdescribed role, regulates KCNH2 expression.

Author contributions: R.C., J.J., E.D., and J.T. designed research; R.C., R.G.U., I.A., M.M.,M.P.-H., D.T., S.A., P.N.-M., G.G.-S., Q.-h.L., R.R.-M., D.P.-B., T.H., K.F.C., J.J., E.D., and J.T.performed research; D.F.-R., R.P., and J.L.L.-S. conducted clinical evaluation of familymembers; R.C., R.G.U., I.A., M.M., M.P.-H., D.T., S.A., P.N.-M., G.G.-S., Q.-h.L., R.R.-M.,D.P.-B., J.J., E.D., and J.T. analyzed data; and R.C., J.J., E.D., and J.T. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1R.C. and R.G.U. contributed equally to this work.2Present address: Cardiac Arrhythmia Department, Fundación Centro Nacional de Inves-tigaciones Cardiovasculares (CNIC), 28029 Madrid, Spain.

3E.D. and J.T. contributed equally to this work.4To whom correspondence should be addressed. Email: [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1612383114/-/DCSupplemental.

E416–E425 | PNAS | Published online January 3, 2017 www.pnas.org/cgi/doi/10.1073/pnas.1612383114

Dow

nloa

ded

by g

uest

on

June

4, 2

020

test, and Holter were completely normal. Thereafter, bisoprololtreatment was started. No new episodes have been documented todate. Evaluation of the family identified two sisters who died sud-denly: one at age 19, in the postpartum period (II:2), and another atage 17 (II:5). Both had been diagnosed with epilepsy and treatedwith phenobarbital until death. Interestingly, II:2, who exhibited aQTc of 440 ms, underwent an adrenaline test that was negative.Sister II:1 also presented syncopal episodes since she was 13,

when she was diagnosed with epilepsy and treated with phenobar-bital. After a syncopal episode at rest, the ECG showed a QTc of560 ms, and ECG monitoring documented a polymorphic ventric-ular tachycardia. At age 43, she experienced an aborted SCD de-spite atenolol treatment (50 mg twice daily) and pacemaking at75 bpm with a dual chamber pacemaker (DDD). Thereafter, adual-chamber cardioverter defibrillator (ICD) was implanted. Shehas been asymptomatic since then.Sister II:3 has no cardiac symptoms. Her ECG showed sinus

rhythm at 68 bpm with normal PR, QRS, and QTc values.Echocardiography, ergometry, Holter, and adrenaline test did

not reveal any structural disease or arrhythmias. However, shesuffers from lupus with a mild decrease of kidney function and isunder treatment with prednisone, hydroxychloroquine, myco-phenolate, and spironolactone (25 mg daily). Her most recentserum K+ concentration was 5.2 mEq per L. Sister II:6 is alsoasymptomatic; her ECG showed normal PR (140 ms) and QRS(90 ms) intervals but sinus bradycardia (48 bpm) and low-voltagewide QT waves (QTc 460 ms) (Fig. S1). An adrenaline test waspositive: on ECG the QTc was prolonged to 618 ms, and therewere T-wave amplitude alternans (Fig. S1) and polymorphicventricular extrasystoles. Therefore, an ICD was implanted andbisoprolol treatment was started. The proband’s mother (66 y old),who had previously been asymptomatic with a normal ECG, wasalso diagnosed with LQTS after an aborted SCD episode duringantibiotic therapy (piperacillin/tazobactam, ciprofloxacin, andtobramycin) in the context of chemotherapy (idarubicin andcytarabine) for the treatment of acute leukemia. An ICD wasimplanted but she died shortly thereafter from the leukemia. Theproband’s father is still alive, asymptomatic, and with a normal

A

II:4

II:1

III:2

II:1

p.T152HfsX180p.T152HfsX180 p.Q1068Rp.Q1068R

2 s

10 p

A/pF

2 s

10 p

A/p

F

2 s

+60 mV-60 mV-80 mV

WT (0.5 g)

p.T152HfsX180(1 g)

WT (0.5 g) + p.T152HfsX180 (0.5 g)

2 s

WT (1 g)

1 135 404 666 1159748 872

PAS Proximal C-linker cNBDS1-S6

25

N-Cap Distal

N-terminus Transmembranesegments

C-terminus

WT

p.T152HfsX1801 135 332

PASChanged aa

25

N-Cap

152

0

10

20

30

HERG WT ( g)p.T152HfsX180 ( g)

0.5 1 00 0 1

0.50.5

I HER

G ta

ilde

nsity

(pA

/pF)

P<0.05 P>0.05P<0.01

B

ED

C

FP<0.05

Fig. 1. (A) Pedigree of the studied family. The arrow indicates the proband. Circles and squares represent females and males, respectively. + and – representsubjects with and without the p.T152HfsX180 and p.Q1068R hERG variants, respectively. (B) Twelve-lead electrocardiogram of the proband (paper speed 25mm/s).(C) DNA sequence chromatograms depicting the heterozygous c.453dupC and the c.3203A>G changes of the KCNH2 gene in different family members.(D) Traces were obtained by applying the protocol (Top) for currents recorded in CHO cells transfected with WT, p.T152HfsX180, and WT/p.T152HfsX180 hERGchannels. (E) Schematic representation of the WT and p.T152HfsX180 hERG protein domains. (F) hERG tail current density recorded in CHO cells transfected withWT, p.T152HfsX180, and WT/p.T152HfsX180 hERG channels after pulses to +60 mV (n ≥ 6). Each bar represents mean ± SEM of the data (n ≥ 8 cells).

Caballero et al. PNAS | Published online January 3, 2017 | E417

PHYS

IOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

June

4, 2

020

ECG. The proband has two nephews (Fig. 1A). III:1 was studied(ECG, echocardiography, ergometry, and Holter) when he was achild; the results revealed an electrically and structurally normalheart. Afterward, an adrenaline test was conducted that wasnegative when he was 23 y old. Conversely, III:2 has experiencedepileptic crises since he was 2 yo. He has been treated withoxcarbazepine since he was 6 y old, and no new episodes havebeen documented (he is 14 now). His ECG, Holter, echocardio-gram, and stress test are normal.

KCNH2 Variants and Functional Analysis. Next-generation sequenc-ing of 82 genes (Table S1) demonstrated that the proband andsister II:1 carried a heterozygous frameshift mutation in theKCNH2 gene (NM_000238.3:c.453dupC) (Fig. 1C) encoding forp.T152HfsX180 hERG. This variant is also present in sister II:6.Sisters II:1 and II:3 and both nephews carry another variant in theKCNH2 gene (NM_000238.3:c.3203A>G) encoding for p.Q1068RhERG (Fig. 1C). Because recombination is a very rare event afterfertilization, expression of the p.T152HfsX180 mutation in oneallele and p.Q1068R in the other allele is more likely to representthe condition of sister II:1 (compound heterozygosity).p.Q1068R is considered a “rare control” variant (7) that appears

with a 0.03% frequency in the African population (Table S2).Functional analysis has demonstrated that the p.Q1068R mutationaccelerates both inactivation and recovery from inactivation, whosetime constants suffer ∼10- to 20-mV shifts in their voltage de-pendence (8). However, the p.Q1068R variant has not been con-sidered pathogenic by itself (7). In contrast, p.T152HfsX180 hERGis a 332-aa peptide (Fig. 1E) that we considered to be highlypathogenic and responsible for the LQTS in this family. Thus, weconducted a functional analysis by transfecting CHO cells withcDNA encoding either WT (n = 7) or p.T152HfsX180 hERG (n =6) channels (1 μg). hERG channels generated a slowly activatingcurrent whose amplitude progressively increased with pulses up to0 mV and then progressively decreased at potentials >0 mV owingto the fast C-type inactivation (9), resulting in the bell-shaped cur-rent density–voltage curve typical of hERG channels (Fig. 2A). Fig.1D shows that, as expected, p.T152HfsX180 hERG channels didnot generate any current. To simulate the heterozygous condition ofall of the mutation carriers, cells (n = 17) were transfected with WTplus p.T152HfsX180 hERG channels (0.5 + 0.5 μg). Surprisingly,maximum current amplitudes generated by depolarizing pulses (Fig.2A) and tail currents recorded upon repolarization to −60 mV(Figs. 1F and 2B) were not statistically different from those gen-erated by WT hERG channels (1 μg) (P > 0.05). We surmised thatthe p.T152HfsX180 hERG peptide could exert a “chaperone-like”effect by increasing membrane expression of WT hERG channels.In fact, Fig. 1F demonstrates that addition of the peptide (0.5 μg) tohERG WT (0.5 μg) generated significantly greater currents thanthose generated by hERG channels alone (P < 0.05). Furthermore,p.T152HfsX180 hERG did not modify the voltage dependence ofhERG activation (Fig. 2B) but slowed deactivation (Table S3). Toconfirm this chaperone-like effect, we transfected mouse atrial-derived HL-1 cells with the p.T152HfsX180 hERG. Some HL-1cells (36%) exhibited IKr as the main repolarizing current (IKr-pre-dominant cells), whereas in other cells (36%) IKr overlapped withthe slow component of the delayed rectifier current (IKs) (in-termediate cells). Thus, IKr was measured in IKr-predominant and-intermediate cells as the dofetilide-sensitive current, because it wascompletely inhibited by this selective Kr blocker (1 μmol/L) (10).Fig. 2C demonstrates that p.T152HfsX180 hERG significantly in-creases both the maximum and tail amplitudes of IKr (P < 0.05).Furthermore, the tail current increase and the slowing of tail cur-rent deactivation depended on the amount of cDNA transfected(Fig. 2 D and F and Table S3).We used a previously validated in silico model of the human

ventricular action potential (AP) (11) to test for the effects of theheterozygous p.T152HfsX180 hERG mutation. The model was run

for endocardial and epicardial cells at different frequencies rangingbetween 0.1 and 3 Hz. The voltage- and time-dependent character-istics of currents generated by WT+p.T152HfsX180 hERG channelswere incorporated into the model to simulate mutation effects. Fig.2E shows superimposed human endocardial APs driven at 0.1 Hzgenerated by WT andWT+p.T152HfsX180 hERG channels. As canbe observed, the duration of the heterologous mutant case AP(APD; action-potential duration) was slightly briefer. Further-more, APD measured at 90% of repolarization (APD90) of sim-ulated WT+p.T152HfsX180 endocardial and epicardial cells wasonly slightly abbreviated (∼3%) at either driving frequency (Fig. 2G).Overall, these results suggested that the heterozygous

p.T152HfsX180 hERGmutation produced subtle effects over the IKr,even when confirmation on a more physiological setting is needed.

TBX20 Mutation and Functional Analysis.Next-generation sequencingof the proband also identified the heterozygous mutationNM_001077653.2:c.931C>T at the TBX20 gene (Table S2), whichwas confirmed by Sanger analysis (Fig. 3B). The mutation leads to

-100 -75 -50 -25 0 25 50 750

10

20

30

WT (1 g)

WT (0.5 g) +p.T152HfsX180 (0.5 g)p.T152HfsX180 (1 g)WT (0.5 g)

Membrane potential (mV)

I hER

G ta

il de

nsity

(pA

/pF)

* * * * * ** * * * * * * *

CA

-100 -75 -50 -25 0 25 50 750

1

2

3

4

5

p.T152HfsX180 (-), n=13p.T152HfsX180 (0.5 g), n=10p.T152HfsX180 (1 g), n=12

*****

* * *

***** ***

Membrane potential (mV)

I Kr t

ail d

ensi

ty (p

A/p

F)

1 s

2 pA

/pF

+60 mV

-60 mV-40 mV

100 ms

0.1

pA/p

F

WT+p.T152HfsX180 WT

100 ms

50 m

V

0 mV

WT +p.T152HfsX180

WT

B

CHO cells HL-1 cells

0 1 2 30

10

20

30

EpiEndo

Frequency (Hz)

AP

D90

sho

rteni

ng (%

)

-100 -75 -50 -25 0 25 50 75

0

5

10

15

20WT (1 g)

WT (0.5 g) + p.T152HfsX180 (0.5 g)p.T152HfsX180 (1 g)WT (0.5 g)

* * * *** * * * *

Membrane potential (mV)

I hER

G d

ensi

ty (p

A/p

F)

* *

0

1

2

3

4

5

p.T152HfsX180 ( g)0 0.5 1

**

I Kr t

ail

dens

ity (p

A/p

F)

#

n=13

n=10

n=12

D

E

FGrandi-BersMathematical Model

G

p.T152HfsX180 (-)

p.T152HfsX180 (1 g)

1 s

2 pA

/pF

Fig. 2. (A and B) Maximum current density (current density–voltage rela-tionships) (A) and tail currents (activation curves) (B) generated by WT andp.T152HfsX180 hERG channels alone or when they are cotransfected in CHOcells, as a function of the membrane potential. In B, solid lines represent the fitof a Boltzmann equation. *P < 0.05 vs. hERG WT (1 μg) (n ≥ 6). (C) IKr tracesrecorded in IKr-predominant HL-1 cells transfected or not with p.T152HfsX180hERG. (D) IKr tail current densities recorded in HL-1 cells transfected or not withp.T152HfsX180 hERG. (E) Simulated IKr traces (Top) and APs (Bottom) obtained at0.1 Hz by using the Grandi–Bers mathematical model of human ventricular en-docardial cells by introducing the modifications produced by p.T152HfsX180 hERGon the IKr. (F) IKr tail current densities recorded in HL-1 cells transfected or not withp.T152HfsX180 hERG after pulses to +60 mV. (G) Percentage of APD90 shorteningin APs simulated at different frequencies in epicardial and endocardial cells. Points/bars representmean± SEMof the data. InD and F, n, number of cells; *P< 0.05 vs.nontransfected cells; #P < 0.05 vs. p.T152HfsX180 0.5 μg transfected cells.

E418 | www.pnas.org/cgi/doi/10.1073/pnas.1612383114 Caballero et al.

Dow

nloa

ded

by g

uest

on

June

4, 2

020

the substitution of Arg311 by Cys (p.R311C Tbx20). Arg311,which is highly conserved among different species (Fig. 3C), islocated in the transactivation region of Tbx20 (Fig. 3D). Two af-fected (II:6 and III:2) and another unaffected (II:3) of the pro-band’s relatives also carry the p.R311C Tbx20 mutation (Fig. 3A).The p.R311C Tbx20 variation was annotated with a 0.01% fre-quency in Africans. Other nonsynonymous variants identified inthe proband are listed in Table S2.Tbx20 binds to the consensus sequence “AGGTGTG” within the

DNA of target genes (6). We hypothesized that Tbx20 might reg-ulate the expression of cardiac ion channels involved in the controlof human cardiac repolarization as it does in fly and mouse adulthearts (5, 6). Sequence analysis of mouse and human promoters ofgenes encoding ion-channel subunits revealed that a Tbx20 bindingsite appears in both KCNH2 genes (Table S4). Thus, we aimed atidentifying the effects of WT and p.R311C Tbx20 on the expressionof hERG in HL-1 cells by recording IKr. Transfection with WT(60.8 ± 7.2 pF, n = 72) or p.R311C Tbx20 (65.6 ± 9.4 pF, n = 65)plasmids did not modify HL-1 cell capacitance (55.2 ± 9.0 pF,n = 68) (P > 0.05).Fig. 3E shows dofetilide-sensitive currents recorded in IKr-pre-

dominant HL-1 cells. IKr was recorded again in IKr-predominantand -intermediate HL-1 cells whose distribution percentages

(39% and 33%, respectively) were not modified by Tbx20.Tbx20 WT significantly increased (P < 0.05) the maximumoutward current recorded upon depolarization (Figs. 3E and4A) without modifying the activation kinetics (Table S3). Tbx20also significantly increased the IKr tail density (Figs. 3E and 4B)(P < 0.05), whereas it did not modify tail current deactivation(Table S3). Surprisingly, p.R311C Tbx20 was unable to increasemaximum outward IKr (Figs. 3E and 4A) and the tail currentdensity elicited upon repolarization (Figs. 3E and 4B). Consis-tently, the mutated transcription factor did not modify either theactivation or the deactivation kinetics of the current (Table S3)(P > 0.05). Importantly, p.R311C Tbx20 did not change the per-centage of IKr-predominant and -intermediate HL-1 cells (39%and 28%, respectively). Fig. 4 B and C demonstrate that trans-fection with Tbx20, either WT or mutated, did not significantlymodify the voltage dependence of Kr channel activation (TableS3). Western blot analysis in HL-1 cells showed that WT Tbx20significantly increased (n = 5, P < 0.05), whereas p.R311C Tbx20significantly decreased, the expression of hERG channels (Fig. 4Dand F) (n = 5, P < 0.05). It has been proposed that MiRP1encoded by KCNE2 is also present in the channels generating theIKr in the human heart (12). Fig. 4 E and G demonstrate thatTbx20, either WT or mutated, did not modify MiRP1 expressionin HL-1 cells.To further demonstrate the transcriptional effect of Tbx20 over

the mouse KCNH2 gene, the IKr density was also assessed in HL-1cells, in which endogenous Tbx20 was silenced using lentiviralconstructs containing short hairpin RNA (shRNA) for Tbx20 to-gether with GFP. Control cells were infected with a lentiviruscontaining a scrambled shRNA and GFP. At 48 h postinfection,Tbx20 expression decreased by 58% (Fig. S2). The results demon-strated that the IKr density significantly decreased in Tbx20-silencedcells (Fig. S2) (n ≥ 5, P < 0.05).To test whether Tbx20 regulates the expression of human

KCNH2, we measured the luciferase activity in HL-1 cellsexpressing the minimum human KCNH2 promoter. The luciferaseassay demonstrated that Tbx20 WT, but not p.R311C, significantlyincreased (P < 0.05) the transcription of the human KCNH2 gene(Fig. 4H) (n = 8 dishes per group). Similarly, SP1, a transcriptionfactor whose binding site is also present in the minimal promoterof the human KCNH2 gene, significantly increased the transcrip-tion of this gene (Fig. 4H) (n = 8, P < 0.05). We also tested theeffects of the combined action of both forms of Tbx20 (WT andmutated). Fig. 4H shows that in the joint presence of WT andp.R311C Tbx20 the protranscriptional effect was reduced (n = 3,P < 0.05 vs. Tbx20 alone), probably because of a competitionbetween the WT and mutated forms for the Tbx20 consensusbinding site in the KCNH2 minimal promoter. Interestingly,p.R311C Tbx20 alone did not decrease basal luciferase activity,probably because the sensitivity of our system is limited. Themouse KCNE2 gene promoter does not exhibit the Tbx20 bindingsite, whereas the human gene does (Table S3). Unfortunately, itwas not possible to construct the minimal human promoter of theKCNE2 gene for a luciferase assay.To test for the effects of the Tbx20 mutation on human ven-

tricular AP characteristics, again the mathematical model wasused. Fig. 4I shows superimposed human endocardial APs drivenat 0.1 Hz in control conditions and when cells were transfectedwith TBX20 either WT or mutated. Tbx20 WT shortened theAPD measured at 50% of repolarization (APD50) and theAPD90 as a consequence of an IKr increase (Fig. 4I, Top).Conversely, p.R311C Tbx20 prolonged the APD50 (by 25%compared with Tbx20 WT) and APD90 (by 23%), due to thedecrease of the IKr conductance (Fig. 4I, Top).

Effects of Tbx20 in Human Induced Pluripotent Stem Cell-DerivedCardiomyocytes. Next, we determined the effects of the Tbx20p.R311C mutation on IKr and AP characteristics in a more

A

B

Tbx20(NP_001071121)

II:4

E

Tbx20 WTTbx20 (-)

+60 mV

-60 mV-80 mV

Tbx20 p.R311C

D

1 s

1pA

/pF

C

Fig. 3. (A) Pedigree of the studied family. The arrow indicates the proband.Circles and squares represent females and males, respectively. + and – rep-resent subjects with and without the p.R311C Tbx20 mutation, respectively.(B) DNA sequence chromatograms of the proband depicting the heterozy-gous change (c.931C>T) of the TBX20 gene. (C) Sequence alignment of theregion surrounding R311 in Tbx20 in several species. The box highlights theconservation of this residue. (D) Schematic representation of the Tbx20 se-quence, indicating the T box, and the transactivation and transrepressorregions. (E) IKr traces recorded in IKr-predominant HL-1 cells transfected ornot with either WT or p.R311C Tbx20 by applying the pulse protocol (Top).

Caballero et al. PNAS | Published online January 3, 2017 | E419

PHYS

IOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

June

4, 2

020

physiologically relevant setting. Unfortunately, genetically mod-ified mice were not an option, because adult mouse heart doesnot generate IKr (12). Therefore, we drew upon human inducedpluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) in-fected or not with lentiviral constructs encoding Tbx20 WT ormutated. IKr was measured as the dofetilide-sensitive current(1 μmol/L). Fig. 5A shows IKr tail currents recorded at −60 mVafter depolarizing pulses from −60 mV to −40 and +60 mV in anoninfected hiPSC-CM. Tbx20 WT produced a 1.6-fold increasein IKr tail current density (P < 0.05), whereas Tbx20 p.R311C

reduced the tail current density by ∼60% (P < 0.05 vs. Tbx20 WTand noninfected cells; n ≥ 7) (Fig. 5B). Tbx20, either WT orp.R311C, did not modify the voltage dependence of Kr channelactivation (Table S3). APs were recorded in hiPSC-CMs thatexhibited automatic activity (13). In cells infected with Tbx20WT (n = 10), maximum diastolic potential and AP amplitudeaveraged −68.4 ± 1.9 and 98.7 ± 14.2 mV, respectively. At 1 Hz,APD50 and APD90 were 108.1 ± 17.4 and 163.2 ± 23.2 ms, re-spectively, and Tbx20 WT shortened whereas Tbx20 p.R311Csignificantly prolonged the APD90 (Fig. 5C) (n = 10, P < 0.01).

-50 -25 0 25 50 75

0

1

2

3

Tbx20 WT (n=13)Tbx20 p.R311C (n=14)

Tbx20 (-) (n=11)

**

*

**

# #

Membrane potential (mV)

Dof

etili

de-s

ensi

tive

curre

ntde

nsity

(pA

/pF)

-60 -20 20 600

1

2

3

4

Tbx20 WT (n=12)Tbx20 p.R311C (n=12)

Tbx20 (-) (n=10)

*

#

*******

######

Membrane potential (mV)

Dof

etili

de-s

ensi

tive

tail

curre

nt d

ensi

ty(p

A/p

F)

A B

0.0

0.5

1.0

1.5

2.0 *

Tbx20(-)

Tbx20WT

Tbx20p.R311C

Rel

ativ

ehE

RG

expr

essi

on

* #

0.0

0.4

0.8

1.2

Tbx20(-)

Tbx20WT

Tbx20p.R311C

Rel

ativ

eM

iRP

1ex

pres

sion

0.0

0.5

1.0

1.5 ** **

Human KCNH2 promoter

Tbx20 WTTbx20 p.R311C

SP1

(-)(-)(-)

#

Nor

mal

ized

luci

fera

seac

tivity

*

(+)(-)(-)

(-)(+)(-)

(+)(+)(-)

(-)(-)(+)

C

GF

-60 -20 20 600.0

0.2

0.4

0.6

0.8

1.0

Tbx20 (-)Tbx20 WTTbx20 p.R311C

Membrane potential (mV)

Nor

mal

ized

I Kr

tail

curre

nt

H100 ms

0.1

pA/p

F

Tbx20 WT

Tbx20 p.R311C

Tbx20 (-)

100 ms

50m

V

0 mV

Tbx20WT Tbx20 (-)

I

Tbx20 p.R311C

N=5

N=5

N=5

N=3 N=3 N=3

N=8

N=8

N=8

N=8

D E

N=3

Fig. 4. (A and B) Maximum IKr density–voltage relationships (A) and activation curves (B) for currents recorded in IKr-predominant and -intermediate HL-1 cellstransfected or not with either WT or p.R311C Tbx20. (C) Normalized activation curves for currents recorded in the three experimental groups. In B and C, solidlines represent the fit of a Boltzmann equation. (D and E) Western blot (WB) images and their corresponding stain-free gels showing hERG (arrows in D) andmiRP1 (E) expression in HL-1 cells transfected or not with either WT or p.R311C Tbx20. In D, the sample of the last right lane was run in the same gel but wasseparated (continuous line) when incubating with the primary antibody together with the antigenic peptide. (F and G) Mean densitometric analysis of hERG (F)and MiRP1 (G) levels normalized to total protein. (H) Normalized luciferase activity in HL-1 cells expressing the pLightSwitch_Prom vector carrying the humanKCNH2 promoter cotransfected or not with SP1 and either WT or p.R311C Tbx20. (I) Simulated IKr traces (Top) and APs (Bottom) obtained at 0.1 Hz by using theGrandi–Bers mathematical model of human ventricular endocardial cells by introducing the modifications produced by Tbx20 WT and p.R311C on the IKr. Points/bars represent mean ± SEM of the data. n, number of cells; N, number of dishes. *P < 0.05 vs. Tbx20 (-); #P < 0.05 vs. Tbx20 WT; **P < 0.01 vs. Tbx20 (-).

E420 | www.pnas.org/cgi/doi/10.1073/pnas.1612383114 Caballero et al.

Dow

nloa

ded

by g

uest

on

June

4, 2

020

Furthermore, as shown in Fig. 5D, the prolongation of theAPD90 was greater at slow- (158% at 0.1 Hz) than at fast-drivingfrequencies (102% at 2 Hz).To reproduce the genetic condition of the proband, the math-

ematical model was run considering both the heterozygousp.T152HfsX180 hERG mutation together with the p.R311CTbx20 mutation. Fig. 5E demonstrated that p.R311C Tbx20 wouldlengthen the APD in the presence of the p.T152HfsX180 hERGmutation. Furthermore, the prolongation was greater in endo-cardial than in epicardial cells and also at slow- than at fast-drivingfrequencies (Fig. 5F).

Effects of p.R311C on the Expression of Other Cardiac K Channels. In thehuman ventricular myocardium Kv7.1 (encoded by KCNQ1)homotetramers associate with minK proteins (encoded by KCNE1)to form the channels that generate IKs (14). Some HL-1 cells (27%)exhibit IKs as the main repolarizing current (IKs-predominant cells).Thus, IKs was measured in IKs-predominant and -intermediate cellsas dofetilide-resistant current, which was completely inhibited by aselective Ks channel blocker (HMR-1556; 1 μmol/L) (10). Fig. 6Ashows IKs traces recorded in IKs-predominant cells transfected withWT or mutated Tbx20. Neither WT nor mutant Tbx20 modified theIKs density (n ≥ 6, P > 0.05) (Fig. 6B). Moreover, neither modifiedthe percentage of IKs-predominant or -intermediate cells (28% and

33% in the presence of WT and p.R311C Tbx20, respectively).Consistently, Tbx20 did not modify either the voltage dependenceof Ks channel activation or the activation and deactivation kinetics(Table S3). Western blot analysis (Fig. S3) confirmed that WT andp.R311C Tbx20 were not able to modify the expression of Kv7.1channels (Fig. 6C) (n ≥ 3, P > 0.05). Sequence analysis of the mouseKCNQ1 gene promoter demonstrated that there are two Tbx20consensus binding sites far away from the transcription start site(−2474 and −1992) (Table S4). Importantly, human KCNQ1 lacksthe Tbx20 binding site (Table S4). Consistently, neither WT norp.R311C Tbx20 modified the expression of human KCNQ1 mea-sured with a luciferase assay (Fig. 6E). Conversely, SP1, whose con-sensus binding site is present in the minimal promoter of the humangene (16), actually increased KCNQ1 transcription significantly (Fig.6E). Human and mouse KCNE1 gene promoters exhibit consensusTbx20 binding sites (Table S4) and, indeed, both WT and p.R311CTbx20 significantly increased minK expression (Fig. 6D) in HL-1cells and transcription as measured by luciferase assay (Fig. 6F).Neither human, rat, nor mouse KCNJ2 gene promoters (which

encode for the inward rectifier K+ channel Kir2.1) exhibit theTbx20 consensus binding site (Table S4). Fig. 6G and Fig. S3demonstrate that Kir2.1 protein expression was not modified bythe presence of any of the forms of Tbx20 (n = 3 dishes per group,P > 0.05). However, even when Kir2.1 channels were expressed,

E

F

GRANDI-BERS MATHEMATICAL MODEL

250 ms

50pA

+60 mV

-60 mV-40 mV

Dofe-sensitivetail currents

A C

B D

Tbx20(-)

-60 -20 20 600.0

0.5

1.0

1.5

Tbx20 WT, n=9Tbx20 p.R311C, n=9

Tbx20 (-), n=7

******

## # # #

Membrane potential (mV)

Dof

e-se

nsiti

ve ta

ilcu

rrent

dens

ity(p

A/p

F)

****0 1 2

0

200

400

600

800

Tbx20 W TTbx20 p.R311C

**

* *

Frequency (Hz)

AP

D90

dura

tion

(ms)

HUMAN iPSC-DERIVED VENTRICULAR MYOCYTES

50 ms

40m

V

0 mV

Tbx20 WT

Tbx20 p.R311CTbx20 (-)

Tbx20 (-) WT p.R311C0

100200300400500

Tbx20 (+)

*

* #

AP

D90

dura

tion

(ms)

100 ms

0.1

pA/p

F

hERG p.T152HfsX180

hERG p.T152HfsX180+p.R311C Tbx20

100 ms

50m

V

0 mV

hERG p.T152HfsX180hERG p.T152HfsX180+Tbx20 p.R311C

0 1 2 30

10

20

EpiEndo

Frequency (Hz)

AP

D90

prol

onga

tion

(%)

hERG p.T152HfsX180+Tbx20 p.R311Cvs

hERG p.T152HfsX180

n=10n=10

n=10

Fig. 5. (A) Dofetilide (1 μmol/L)-sensitive (IKr) tail currents obtained by digital subtraction in a noninfected hiPSC-derived cardiomyocytes. (B) IKr density in ahiPSC-derived cardiomyocyte infected or not with either WT or p.R311C Tbx20. Solid lines represent the fit of a Boltzmann equation. (C) Superimposed APsrecorded at 1 Hz in three hiPSC-derived cardiomyocytes infected or not with either WT or p.R311C Tbx20. (Top) The APD90 of each experimental group. (D)APD90 at different stimulation frequencies in cells infected with Tbx20 WT or p.R311C. (E) Simulated IKr traces (Top) and APs (Bottom) obtained at 0.1 Hz byusing the Grandi–Bers mathematical model of human ventricular endocardial cells by introducing the modifications produced by heterozygous p.T152HfsX180hERG alone or in combination with p.R311C Tbx20 on the IKr. (F) Percentage of APD90 prolongation in APs simulated at different frequencies in epicardial andendocardial cells. In B–D, points/bars represent mean ± SEM of ≥7 experiments in each group. *P < 0.05 vs. Tbx20(-); #P < 0.05 vs. Tbx20 WT.

Caballero et al. PNAS | Published online January 3, 2017 | E421

PHYS

IOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

June

4, 2

020

the inward rectifier K current (IK1) could not be recorded inHL-1 cells. Instead, the pacemaker current (If) predominatedat potentials between −150 and 0 mV but its current density–voltage relation was unaltered by either WT or p.R311C Tbx20(Fig. 6H) (n ≥ 6, P > 0.05). This result agrees with the absenceof a Tbx20 consensus binding site in the promoters of thegenes (HCN1, HCN3, and HCN4) encoding the channels un-derlying If in mice and humans (Table S4). Therefore, to assessthe role of Tbx20 in Kir2.1 functional regulation, the IK1 wasrecorded in enzymatically dissociated rat ventricular myocytesthat were transfected using a lentiviral construct (17). Fig. 6 Iand J show that the IK1 density was not modified at any of thevoltages tested in the presence of either WT or p.R311C Tbx20(n ≥ 7, P > 0.05).

Effects of p.R311C on the Expression of Cardiac Na and Ca Channels.SCN5A and CACNA1C genes codify for the α-subunit of the Na+

(Nav1.5) and L-type Ca2+ (Cav1.2) channels, respectively. Mouseand human SCN5A gene promoters lack a Tbx20 binding site (TableS4). Fig. 7 A and B confirm that the Na+ current (INa) density wasalso not modified by WT or p.R311C Tbx20 (n ≥ 8, P > 0.05).Furthermore, transfection of WT or p.R311C Tbx20 did not modifythe voltage dependence of the INa activation or inactivation or thecurrent kinetics (Fig. 7C and Table S3). Fig. 7 D and E show thatTbx20 WT or p.R311C did not modify the amplitude of the sus-tained influx of Na+ measured at the end of 500-ms depolarizations

to −20 mV (INaL) (n ≥ 6, P > 0.05). The human SCN2B pro-moter, which codifies for an Nav1.5 ancillary subunit, exhibitsthe Tbx20 binding site (Table S4). Luciferase assays demon-strated that Tbx20 WT and p.R311C were unable to modifyhuman SCN5A gene expression, whereas Tbx20 WT, but notTbx20 p.R311C, significantly increased the expression of hu-man SCN2B (Fig. 7 F and G).The mouse, but not the human, CACNA1C gene promoter

exhibits the Tbx20 binding site (Table S4). Therefore, Tbx20effects on the L-type Ca2+ current (ICaL) were tested in bothHL-1 cells and hiPSC-CMs. In HL-1 cells, ICaL was measured usingBa2+ as a charge carrier (IBa) (15, 18). Fig. S4 shows that Tbx20,both WT and mutated, significantly increased the IBa (n ≥ 9, P <0.05). Furthermore, p.R311C Tbx20 increased the IBa densitysimilar to Tbx20 WT (P > 0.05 vs. Tbx20 WT). However, neitherWT nor mutated Tbx20 affected the voltage dependence of acti-vation or inactivation of the channel (Fig. S4 and Table S3).Western blot analysis in HL-1 cells (Fig. S3) demonstrated thatboth WT and p.R311C Tbx20 significantly and similarly increaseCav1.2 expression (Fig. S4).Fig. 8A shows ICaL traces recorded in hiPSC-CMs infected or not

with the lentiviral constructs encoding for Tbx20, either WT or mu-tated. Neither WT nor p.R311C Tbx20 modified the ICaL density(Fig. 8A) at any of the voltages tested (Fig. 8B) (n ≥ 6, P > 0.05). Theluciferase assay done using the minimal human CACNA1C promoter

+60 mV-30 mV

-80 mV

2 s

2 pA

/pF

-50 -25 0 25 50 75

0

1

2

3 Tbx20 WT, n=6Tbx20 p.R311C, n=6

Tbx20 (-), n=9

Membrane potential (mV)

Dof

etili

de-re

sist

ant

curre

nt d

ensi

ty (p

A/p

F)

0.0

0.5

1.0

1.5

Tbx20(-)

Tbx20WT

Tbx20p.R311C

N=3N=3 N=3

Rel

ativ

e K

v7.1

expr

essi

on

0

1

2

3

TBX20(-)

TBX20WT

TBX20p.R311C

* *R

elat

ive

min

Kex

pres

sion

N=3

N=3 N=3

E

0.0

0.5

1.0

1.5

2.0

Tbx20 (+)

**

Human KCNQ1 promoter

Tbx20(-)

WT p.R311C SP1(+)

N=5 N=5 N=5

N=5

Nor

mal

ized

luci

fera

seac

tivity

B

0.0

0.5

1.0

1.5

2.0

Tbx20(-)

Tbx20WT

Tbx20p.R311C

* *

Human KCNE1 promoter

Nor

mal

ized

luci

fera

seac

tivity N=5

N=5 N=5

D

2 s

2 pA

/pF

Tbx20 WT

Tbx20 p.R311C

A

F

C

0.0

0.4

0.8

1.2

Tb2x0(-)

Tbx20WT

Tbx20p.R311C

Rel

ativ

e K

ir2.1

expr

essi

on

N=3 N=3N=3

HL-1 Cells

-8.0

-5.5

-3.0

-0.5

2.0

Tbx20 WT, n=6Tbx20 (-), n=8

If density (pA/pF)

-150 -110 -70-10

Tbx20 p.R311C, n=6

-30

Membrane potential (mV)

JI

-120 -80 -40 0 40

-20

-15

-10

-5

Tbx20 WT, n=11Tbx20 (-), n=12

IK1 density (pA/pF)

Tbx20 p.R311C, n=7

Membrane potential (mV)

IK1 in Rat Ventricular Myocytes

HG

+40 mV

-100 mV-40 mV

Tbx20 WT

50 ms

10 p

A/p

F

Tbx20 p.R311C

Fig. 6. (A) Traces of dofetilide-resistant current (IKs) recorded in IKs-predominant HL-1 cells transfected with WT or p.R311C Tbx20 by applying the pulseprotocol (Top). (B) Current density–voltage relationships for IKs recorded in HL-1 cells transfected or not with WT or p.R311C Tbx20. (C and D) Mean den-sitometric analysis of Kv7.1 (C) and minK (D) levels normalized to total protein. (E and F) Normalized luciferase activity in HL-1 cells expressing thepLightSwitch_Prom vector carrying the human KCNQ1 (E) or KCNE1 (F) promoters cotransfected or not with WT or p.R311C Tbx20. (G) Mean densitometricanalysis of Kir2.1 levels normalized to total protein. (H) Current density–voltage relationships for If recorded in HL-1 cells transfected or not with WT orp.R311C Tbx20. (I) IK1 traces recorded in two rat myocytes infected with WT and p.R311C Tbx20. (J) Mean current density–voltage curves for IK1 recorded in ratventricular myocytes infected or not with lentiviral constructs encoding WT and p.R311C Tbx20. Each point/bar represents mean ± SEM of n cells or N dishes ofcells in each group. *P < 0.05 vs. Tbx20 (-); **P < 0.01 vs. Tbx20 (-).

E422 | www.pnas.org/cgi/doi/10.1073/pnas.1612383114 Caballero et al.

Dow

nloa

ded

by g

uest

on

June

4, 2

020

confirmed that Tbx20 WT and p.R311C were unable to modify theexpression of the human CACNA1C gene (Fig. 8C), a result thatexplains the lack of Tbx20 effects over the human ICaL density.

Functional Analysis of the c.-66A>G Variation of KCNN3. The pro-band, sister II:6, and nephew III:1 also present a variation at the5′ UTR of the KCNN3 gene (NM_001204087.1:c.-66A>G) thatencodes the α-subunit of the small-conductance Ca2+-activatedK+ channel type 3 (SK3) (19). Luciferase experiments demon-strated that expression of mutated KCNN3 cannot be activatedby SP1 or Tbx20 (Fig. S5), whose binding sites are present in thehuman gene promoter (Table S4). Therefore, the variationcompletely abolished its transcription, thus leading to a KCNN3haploinsufficiency. However, because treatment of human mul-ticellular ventricular preparations with apamin (a selective SKblocker) does not modify the APD (20), the importance of thesechannels in repolarization seems to be negligible.

DiscussionHere we functionally describe the consequences of three variantsidentified in a Spanish family of African ancestry with LQTS.The TBX20 mutation selectively decreased the expression ofhERG channels, prolonging the AP in hiPSC-CMs. Conversely,the KCNH2 frameshift mutation did not modify IKr density. Ourresults strongly suggest that, in the adult heart, Tbx20 controlsthe expression of hERG channels, and thus TBX20 may beconsidered an LQTS-modifying gene.The p.T152HfsX180 hERG mutation was found in the proband

and in all of the affected relatives that were genotyped but in noneof the nonaffected family members. Therefore, the mutation wasconsidered pathogenic. Phenotypic manifestations in the familymatch most of the features of LQT2, and three of the proband’ssisters experienced seizures since they were children. Epilepsy hasbeen reported to be more common with LQT2 (39%) than withother subtypes (10%), possibly because KCNH2 is also expressed inthe brain (21). Interestingly, nephew III:2, who carries the TBX20but not the KCNH2 frameshift mutation, has also experiencedseizures. Additionally, one of the sisters died postpartum, which is aspecific trigger of symptoms in LQT2 (22). Functional analysis ofthe p.T152HfsX180 mutation demonstrated that this peptide of 332aa, of which only 152 correspond to the hERG sequence, exertschaperone-like effects on WT hERG channels in CHO cells and onIKr recorded in HL-1 cells. Indeed, transfection of p.T152HfsX180in HL-1 cells produces a “concentration-dependent” increase in IKr.As a consequence, current density generated by “heterozygous”transfection of WT and p.T152HfsX180 hERG channels was notdifferent from that generated by “homozygous” transfection of WThERG channels. This is a somewhat surprising result consideringthat, as expected, homozygous transfection of p.T152HfsX180 didnot generate current at all. We recently demonstrated that theNav1.5 N-terminal domain, by itself (the 132-aa peptide) (Nter),exerts a chaperone-like effect that increases INa and IK1 by en-hancing the expression of Nav1.5 and Kir2.1-Kir2.2 channels inCHO cells and in rat cardiomyocytes (17). We hypothesize that thep.T152HfsX180 peptide is able to increase membrane expression of

0 100 200 300 400 500-20

-15

-10

-5

0

Tbx20 WTTbx20 p.R311C

Tbx20 (-)

Time (ms)

I Na

(pA

/pF)

-100

-75

-50

-25

Tbx20 WT, n=14Tbx20 (-), n=20

INa density (pA

/pF)

-100

-60

-20

+20

Tbx20 p.R311C, n=8

Membrane potential (mV) CB

A

5 ms25 p

A/p

F

Tbx20 (-) Tbx20 WT Tbx20 p.R311C

+30 mV

-90 mV-120 mV

-2.0

-1.5

-1.0

-0.5

0.0

n=7 n=9 n=6

I Na,

Lde

nsity

(pA

/pF)

Tbx20WT

Tbx20p.R311C

Tbx20 (-)

-140 -100 -60 -200.00.20.40.60.81.0 Tbx20 (-), n=10

Tbx20 WT, n=6Tbx20 p.R311C

n=6

Membrane potential (mV)

Nor

mal

ized

I Na

0.0

0.5

1.0

1.5

Tbx20(-)

Tbx20WT

Tbx20p.R311C

Nor

mal

ized

luci

fera

seac

tivity

N=5 N=5 N=5

0.0

0.5

1.0

1.5**

Tbx20(-)

Tbx20WT

Tbx20p.R311C

Nor

mal

ized

luci

fera

seac

tivity

N=5

N=5N=5

Human SCN5A promoter Human SCN2B promoter

ED

GF

Fig. 7. (A) INa traces recorded in HL-1 cells transfected or not with WT orp.R311C Tbx20 by applying the pulse protocol (Top). (B and C) Currentdensity–voltage relationships (B) and steady-state inactivation (C) for INarecorded in the three experimental groups. (D and E) Superimposed INatraces (D) recorded in HL-1 cells transfected or not with WT or p.R311C Tbx20by applying 500-ms pulses from −120 to −20 mV and bar graph (E) showingthe mean INaL recorded at 500 ms. (F and G) Normalized luciferase activity inHL-1 cells expressing the pLightSwitch_Prom vector carrying the humanSCN5A (F) or SCN2B (G) promoters cotransfected or not with WT or p.R311CTbx20. Each point/bar represents the mean ± SEM of n cells or N dishes ofcells in each group. **P < 0.01 vs. Tbx20 (-).

-60 -40 -20 20 40 60

-12

-8

-4

Tbx20 WT, n=9Tbx20 (-), n=6

ICaL density (pA

/pF)

Tbx20 p.R311C, n=6

Membrane potential (mV)

Tbx20 (-) Tbx20 WT Tbx20 p.R311C

50 ms5 pA

/pF

+60 mV

-50 mV-30 mV-80 mV

0.0

0.4

0.8

1.2

Tbx20(-)

Tbx20WT

Tbx20p.R311C

Nor

mal

ized

luci

fera

seac

tivity

N=5N=5 N=5

Human CACNA1C promoter

A

B C

Fig. 8. (A) ICaL traces recorded in hiPSC-CMs infected or not with WT orp.R311C Tbx20 by applying the pulse protocol (Top). (B) Current density-voltage relationships for ICaL recorded in the three experimental groups.(C) Normalized luciferase activity in HL-1 cells expressing the pLightSwitch_Promvector carrying the human CACNA1C promoter cotransfected or not withWT or p.R311C Tbx20. Each point/bar represents the mean ± SEM of n cells orN dishes of cells in each group.

Caballero et al. PNAS | Published online January 3, 2017 | E423

PHYS

IOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

June

4, 2

020

hERG channels. The molecular determinants, and the proteinsinvolved in this effect, merit further analysis. The question now iswhether this KCNH2 frameshift mutation is, by itself, responsiblefor the LQT phenotype of the family.The proband and relatives II:3, II:6, and III:2 harbor a mu-

tation in the TBX20 gene. Mutations in Tbx20 have been pre-viously described and lead to defects in cardiac septation,valvulogenesis, and chamber growth (23). Indeed, Tbx20 is nec-essary for proper organogenesis, because it carries strong tran-scriptional activation and repression domains and physicallyinteracts with other transcription factors involved in cardiacdevelopment (4). It is noteworthy that the R311 residue lies inthe transcriptional activator domain but none of the mutationcarriers presented any structural cardiac defect.Functional analysis demonstrated that Tbx20 does not directly

control the expression of the channels that underlie INa, If, IK1,and IKs. Regarding the IKs, the results demonstrated that bothWT and mutated Tbx20 increased minK expression. Therefore,in HL-1 cells, an IKs augmentation would have been expected,because minK increases Kv7.1 conductance (24, 25). However,simultaneously, minK acts as an endocytic chaperone favoringthe internalization of the Kv7.1–minK complexes expressed inthe membrane (26), an effect that would decrease IKs density.Therefore, the balance between these two opposite actions couldexplain why the minK augmentation was not accompanied by achange in IKs density. Our results confirm previous data dem-onstrating that, in mice, Tbx20 increases ICaL (6). However,Tbx20 did not increase the expression of human CACNA1C,because the canonical Tbx20 binding site is not present in thegene promoter. Accordingly, Tbx20, either WT or mutated, didnot modify the ICaL in hiPSC-CMs. Our results demonstratedthat Tbx20 significantly increases the expression of hERG andthus IKr in HL-1 cells and hiPSC-CMs. Conversely, in adult flies,neuromancer (the invertebrate ortholog of Tbx20) negativelyregulates the expression of the invertebrate homolog of the ERGchannel (eag-like K+ channel) (5). Moreover, functional analysisdeveloped in HL-1 cells and hiPSC-CMs strongly suggested thatthe p.R311C mutation specifically disables the protranscriptionalactivity of Tbx20 on the KCNH2 gene. Therefore, we proposethat in the human adult myocardium, this Tbx20 mutation leadsto a prolongation of ventricular repolarization.Results in flies and mice demonstrated that Tbx20 is a key

determinant of adult cardiac function (5, 6). Indeed, heart-spe-cific knockdown of the gene that encodes neuromancer in flies(nmr-2) interferes with cardiac performance and disrupts con-tractile myofibrillar patterning (5). In adult mice, heterozygousloss of TBX20 leads to dilated cardiomyopathy (27), and theconditional homozygous loss of Tbx20 results in severe cardio-myopathy with associated arrhythmias and death (6). It has beenproposed that, in the adult heart, Tbx20 is the pivotal element ofa transcriptional cohort (also constituted by Mef2A, Tead1, Esrr,and Creb1) that fine-tunes expression of continuously requiredproteins in response to the current myocyte state, availability ofresources, and contractile requirements (6). Therefore, evenwhen Tbx20 mutant carriers apparently exhibit a mild phenotypeunder basal conditions, we cannot rule out that their myocar-dium adapts poorly to more demanding situations (e.g., sympa-thetic tone increase or even hormone- or drug-induced decreaseof the repolarization reserve), because probably the p.R311Cmutation affects Tbx20 ability to coordinate adaptive responsesof the transcriptional cohort. Therefore, the simultaneous pres-ence of KCNH2 and TBX20 mutations probably contributes tothe LQTS phenotype in this family.As in other families (28), expressivity of the LQTS phenotype

in this family ranged from the mild phenotype of the proband tothe high symptomatic phenotype of sister II:1. Besides de-mographic variables such as gender and age, variable expressivitymay be attributed to the concurrence of additional genetic

modifiers (29), including the presence of two or more variants,either in the same gene (compound heterozygosity) or in dif-ferent genes (digenic heterozygosity), and the presence of non-synonymous single-nucleotide polymorphism (28, 29). All suchconditions converge in this Spanish family. The proband andsister II:6 present digenic heterozygosity (the KCNH2 frameshiftand the TBX20 mutations), whereas sister II:1 presents a com-pound heterozygosity (the frameshift and the variant in KCNH2).Interestingly, nephew III:2, who has seizures, carries the benignp.Q1068R hERG variant and the TBX20 mutation.We are aware of the potential limitations of this study, in-

cluding that a better experimental approach would have been toanalyze in hiPSC-CMs the effects produced by WT and mutatedTbx20 over all of the cardiac currents responsible for the APmorphology. It would have been even better to analyze the ef-fects produced by Tbx20 over cardiomyocytes differentiatedfrom hiPSCs derived from each family member. The latter wouldhave allowed directly testing the impact of the variants (and theirconnection) in a constant genetic background, weighting theirrespective involvement in the phenotypic expression of theLQTS. Despite such a limitation, the results strongly suggest thatTBX20 is a potential LQTS modifier gene because WT Tbx20increases hERG channel expression. The data show also thatsome mutations, such as p.R311C, can disable Tbx20 protran-scriptional activity over the KCNH2 gene. Therefore, the puta-tive effects of Tbx20 variants on penetrance, expressivity, andoutcome among LQTS patients merit further analysis.

MethodsClinical Evaluation. Patients were evaluated by the Arrhythmia Unit of theHospital Universitario La Paz. The study was approved by the InvestigationCommittee of the hospital and conforms to the principles outlined in theDeclaration of Helsinki. Each participant gave written informed consent.

DNA Sequencing. Genomic DNA was sequenced by means of a HaloPlexcustom panel including coding regions and untranslated boundaries of the 82genes listed in Table S1. Sequencing using the Ion Torrent Personal GenomeMachine was carried out at NIMGenetics. Variants identified in KCNH2,TBX20, and KCNN3 were confirmed by the Sanger method.

Cell Culture and Transfection. HL-1 and Chinese hamster ovary cells weretransiently transfected by using Lipofectamine 2000 and FuGENE X-tremeGENE,respectively, and cultured as described (15, 17, 18, 30).

Rat Ventricular Myocyte Isolation. Animal studies were approved by theCommittee on the Use and Care of Animals at Complutense University andconformed to the guidelines from Directive 2010/63/EU of the EuropeanParliament on the protection of animals used for scientific purposes. Singleventricular myocytes isolated from male Sprague–Dawley rats by enzymaticdissociation (17) were infected with lentiviral constructs encoding for humanTbx20 WT or p.R311C.

Patch Clamping. Currents were recorded using the whole-cell patch-clamptechnique (15, 17, 18, 30). Series resistance was compensated manually usingthe compensation unit of the Axopatch amplifier; ≥80% compensation wasachieved. No significant voltage errors (<5 mV) due to series resistance wereexpected with the micropipettes used.

IKr, ICaL, and AP Recordings in hiPSC-CMs. Enriched and mature DF19-9-11ThiPSC-CMs were generated as described elsewhere (13) and infected with thelentiviral constructs coding WT or p.R311C Tbx20. Currents were recorded at21 to 23 °C, and APs were recorded at 35 °C using the whole-cell patch-clamp technique.

Western Blot Analysis. Cav1.2, Kir2.1, Kv7.1, minK, hERG, MiRP1, and Tbx20proteins were detected in HL-1 cells transfected or not with Tbx20 WT orp.R311C by Western blot following previously described procedures (17, 18).

Luciferase Gene Expression Reporter Assay. Luciferase reporter assays wereconducted in HL-1 cells transfected with pLightSwitch_Prom luciferase expression

E424 | www.pnas.org/cgi/doi/10.1073/pnas.1612383114 Caballero et al.

Dow

nloa

ded

by g

uest

on

June

4, 2

020

reporter vectors carrying the minimal promoters of human SCN5A, SCN2B,CACNA1C, KCNQ1, KCNE1, KCNH2, or KCNN3 (15, 18).

Tbx20 Silencing. For analysis of Tbx20 silencing, HL-1 cells were infected withlentivirus-encoding shRNA Tbx20 or scrambled shRNA (17).

Statistical Analysis. Results are expressed as mean ± SEM. Unpaired t test orone-way ANOVA followed by Newman–Keuls test was used where appro-priate. In small-size samples (n < 15), statistical significance was confirmed byusing nonparametric tests. Comparisons between categorical variables weredone using Z test. To take into account repeated sample assessments, data

were analyzed with multilevel mixed-effects models. A value of P < 0.05 wasconsidered significant. Additional details are presented in SI Methods.

ACKNOWLEDGMENTS. We thank Paloma Vaquero, Sandra Sacristán, LorenaOndo, and Ainara Albadalejo for their invaluable technical assistance.This work was supported by grants from Comunidad Autónoma deMadrid (S2010/BMD-2374: ITACA); Ministerio de Economía y Competitividad(SAF2014-58769-P); Instituto de Salud Carlos III (PI16/00398, CB16/11/00303,and CB16/11/00504); ERA-Net for Research on Rare Diseases (AC14/00029);Mutua Madrileña and BBVA Foundations; National Heart, Lung, and BloodInstitute of the US National Institutes of Health (R01-HL122352 to J.J.); andLeducq Foundation (to J.J.).

1. Nakano Y, ShimizuW (2016) Genetics of long-QT syndrome. J Hum Genet 61(1):51–55.2. Havakuk O, Viskin S (2016) A tale of 2 diseases: The history of long-QT syndrome and

Brugada syndrome. J Am Coll Cardiol 67(1):100–108.3. Keating MT, Sanguinetti MC (2001) Molecular and cellular mechanisms of cardiac

arrhythmias. Cell 104(4):569–580.4. Sakabe NJ, et al. (2012) Dual transcriptional activator and repressor roles of TBX20

regulate adult cardiac structure and function. Hum Mol Genet 21(10):2194–2204.5. Qian L, et al. (2008) Transcription factor neuromancer/TBX20 is required for cardiac

function in Drosophila with implications for human heart disease. Proc Natl Acad SciUSA 105(50):19833–19838.

6. Shen T, et al. (2011) Tbx20 regulates a genetic program essential to adult mousecardiomyocyte function. J Clin Invest 121(12):4640–4654.

7. Kapa S, et al. (2009) Genetic testing for long-QT syndrome: Distinguishing pathogenicmutations from benign variants. Circulation 120(18):1752–1760.

8. Anson BD, et al. (2004) Molecular and functional characterization of common poly-morphisms in HERG (KCNH2) potassium channels. Am J Physiol Heart Circ Physiol286(6):H2434–H2441.

9. Spector PS, Curran ME, Zou A, Keating MT, Sanguinetti MC (1996) Fast inactivationcauses rectification of the IKr channel. J Gen Physiol 107(5):611–619.

10. Tamargo J, Caballero R, Gómez R, Valenzuela C, Delpón E (2004) Pharmacology ofcardiac potassium channels. Cardiovasc Res 62(1):9–33.

11. Amorós I, et al. (2011) Functional effects of a missense mutation in HERG associatedwith type 2 long QT syndrome. Heart Rhythm 8(3):463–470.

12. Nerbonne JM (2000) Molecular basis of functional voltage-gated K+ channel diversityin the mammalian myocardium. J Physiol 525(Pt 2):285–298.

13. Herron TJ, et al. (2016) Extracellular matrix-mediated maturation of human pluripotentstem cell-derived cardiac monolayer structure and electrophysiological function. CircArrhythm Electrophysiol 9(4):e003638.

14. Plant LD, Xiong D, Dai H, Goldstein SA (2014) Individual IKs channels at the surface ofmammalian cells contain two KCNE1 accessory subunits. Proc Natl Acad Sci USA111(14):E1438–E1446.

15. Barana A, et al. (2014) Chronic atrial fibrillation increases microRNA-21 in humanatrial myocytes decreasing L-type calcium current. Circ Arrhythm Electrophysiol 7(5):861–868.

16. Luo X, et al. (2008) Genomic structure, transcriptional control, and tissue distributionof HERG1 and KCNQ1 genes. Am J Physiol Heart Circ Physiol 294(3):H1371–H1380.

17. Matamoros M, et al. (2016) Nav1.5 N-terminal domain binding to α1-syntrophin in-creases membrane density of human Kir2.1, Kir2.2 and Nav1.5 channels. CardiovascRes 110(2):279–290.

18. Pérez-Hernández M, et al. (2016) Pitx2c increases in atrial myocytes from chronic atrialfibrillation patients enhancing IKs and decreasing ICa,L. Cardiovasc Res 109(3):431–441.

19. Adelman JP, Maylie J, Sah P (2012) Small-conductance Ca2+-activated K+ channels:Form and function. Annu Rev Physiol 74:245–269.

20. Nagy N, et al. (2009) Does small-conductance calcium-activated potassium channelcontribute to cardiac repolarization? J Mol Cell Cardiol 47(5):656–663.

21. Johnson JN, et al. (2009) Identification of a possible pathogenic link between con-genital long QT syndrome and epilepsy. Neurology 72(3):224–231.

22. Khositseth A, Tester DJ, Will ML, Bell CM, Ackerman MJ (2004) Identification of acommon genetic substrate underlying postpartum cardiac events in congenital longQT syndrome. Heart Rhythm 1(1):60–64.

23. Kirk EP, et al. (2007) Mutations in cardiac T-box factor gene TBX20 are associated withdiverse cardiac pathologies, including defects of septation and valvulogenesis andcardiomyopathy. Am J Hum Genet 81(2):280–291.

24. Sanguinetti MC, et al. (1996) Coassembly of K(V)LQT1 and minK (IsK) proteins to formcardiac I(Ks) potassium channel. Nature 384(6604):80–83.

25. Sesti F, Goldstein SA (1998) Single-channel characteristics of wild-type IKs channelsand channels formed with two minK mutants that cause long QT syndrome. J GenPhysiol 112(6):651–663.

26. Xu X, et al. (2009) MinK-dependent internalization of the IKs potassium channel.Cardiovasc Res 82(3):430–438.

27. Stennard FA, Harvey RP (2005) T-box transcription factors and their roles in regulatoryhierarchies in the developing heart. Development 132(22):4897–4910.

28. Priori SG, Napolitano C, Schwartz PJ (1999) Low penetrance in the long-QT syndrome:Clinical impact. Circulation 99(4):529–533.

29. Napolitano C, Novelli V, Francis MD, Priori SG (2015) Genetic modulators of thephenotype in the long QT syndrome: State of the art and clinical impact. Curr OpinGenet Dev 33(1):17–24.

30. Caballero R, et al. (2010) Flecainide increases Kir2.1 currents by interacting with cys-teine 311, decreasing the polyamine-induced rectification. Proc Natl Acad Sci USA107(35):15631–15636.

31. Tuteja D, et al. (2010) Cardiac small conductance Ca2+-activated K+ channel subunitsform heteromultimers via the coiled-coil domains in the C termini of the channels.Circ Res 107(7):851–859.

32. González de la Fuente M, et al. (2013) Chronic atrial fibrillation up-regulates β1-adrenoceptors affecting repolarizing currents and action potential duration.Cardiovasc Res 97(2):379–388.

33. Shimizu W, et al. (2003) Epinephrine unmasks latent mutation carriers with LQT1form of congenital long-QT syndrome. J Am Coll Cardiol 41(4):633–642.

Caballero et al. PNAS | Published online January 3, 2017 | E425

PHYS

IOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

June

4, 2

020