Báo cáo khoa học: Deletion of Phe508 in the first nucleotide-binding domain of the cystic fibrosis transmembrane conductance regulator increases its affinity for the heat shock cognate 70 chaperone docx

Here, we determined by surface plasmon resonance that: a F508del-murine mNBD1 binds Hsc70 with higher affinity KD, 2.6 nm than wild-type wt mNBD1 13.9 nm; b ATP and ADP dramatically reduc

of the cystic fibrosis transmembrane conductance

regulator increases its affinity for the heat shock cognate

70 chaperone

Toby S Scott-Ward1,2and Margarida D Amaral1,2

1 Universidade de Lisboa, Faculdade de Cieˆncias de Lisboa, BioFIG, Centre for Biodiversity, Functional and integrative Genomics, Portugal

2 Centro de Gene´tica Humana, Instituto Nacional de Sau´de Dr Ricardo Jorge, Lisboa, Portugal

Keywords

CFTR-interacting proteins; correctors;

mechanism of disease; small molecules;

surface plasmon resonance

Correspondence

M D Amaral, EMBL – European Molecular

Biology Laboratory, Meyerhofstrasse 1,

69117 Heidelberg, Germany

Fax: +49 6221 387 8306

Tel: +49 6221 387 8199

E-mail: mdamaral@fc.ul.pt

(Received 3 August 2009, revised 29

September 2009, accepted 1 October

2009)

doi:10.1111/j.1742-4658.2009.07421.x

The primary cause of cystic fibrosis (CF), the most frequent fatal genetic disease in Caucasians, is deletion of phenylalanine at position 508 (F508del), located in the first nucleotide-binding domain (NBD1) of the

CF transmembrane conductance regulator (CFTR) protein F508del-CFTR

is recognized by the endoplasmic reticulum quality control (ERQC), which targets it for proteasomal degradation, preventing this misfolded but par-tially functional Cl)channel from reaching the cell membrane We recently proposed that the ERQC proceeds along several checkpoints, the first of which, utilizing the chaperone heat shock cognate 70 (Hsc70), is the major one directing F508del-CFTR for proteolysis Therefore, a detailed charac-terization of the interaction occurring between F508del-CFTR and Hsc70

is critical to clarify the mechanism that senses misfolded F508del-CFTR

in vivo Here, we determined by surface plasmon resonance that: (a) F508del-murine (m)NBD1 binds Hsc70 with higher affinity (KD, 2.6 nm) than wild-type (wt) mNBD1 (13.9 nm); (b) ATP and ADP dramatically reduce NBD1–Hsc70 binding; (c) the F508del mutation increases by approximately six-fold the ATP concentration required to inhibit the NBD1–Hsc70 interaction (IC50; wt-mNBD1, 19.7 lm ATP); and (d) the small molecule CFTR corrector 4a (C4a), but not VRT-325 (V325; both rescuing F508del-CFTR traffic), significantly reduces F508del-mNBD1 binding to Hsc70, by  30% Altogether, these results provide a novel, robust quantitative characterization of Hsc70–NBD1 binding, bringing detailed insights into the molecular basis of CF Moreover, we show how this surface plasmon resonance assay helps to elucidate the mechanism of action of small corrective molecules, demonstrating its potential to validate additional therapeutic compounds for CF

Structured digital abstract

l MINT-7265886 : mNBD1 (uniprotkb: P26361 ) binds ( MI:0407 ) to Hsc70 (uniprotkb: P19120 )

by anti bait coimmunoprecipitation ( MI:0006 )

Abbreviations

Ab, antibody; C4a, corrector 4a; CF, cystic fibrosis; CFTR, cystic fibrosis transmembrane conductance regulator; ER, endoplasmic reticulum; ERQC, endoplasmic reticulum quality control; F508del, deletion of phenylalanine residue at position 508; h, human; Hsc70, heat shock cognate 70; Hsp70, heat shock protein 70; I172, inhibitor CFTRinh-172; LA, apo-a-lactalbumin; m, murine; NBD1, first nucleotide-binding domain; Red-LA, reduced apo-a-lactalbumin; SEM, standard error of the mean; SPR, surface plasmon resonance; V325, corrector VRT-325;

wt, wild-type.

Cystic fibrosis (CF) is a life-threatening genetic disease

caused by malfunction of CF transmembrane

conduc-tance regulator (CFTR) [1], a Cl)channel that plays a

central role in transepithelial ion transport [2] A single

amino acid deletion, of phenylalanine 508 (F508del), in

the first nucleotide-binding domain (NBD1) of CFTR

accounts for approximately 70% of CF chromosomes

worldwide [2] This mutation prevents the correct

fold-ing of CFTR, thus causfold-ing its retention by the

endo-plasmic reticulum (ER) quality control (ERQC) as an

immature intermediate that is rapidly degraded by the

ubiquitin–proteasome pathway [3] Like that of other

secretory proteins, the proper folding of CFTR is

highly dependent on molecular chaperones, such as

heat shock protein 70 (Hsp70) and its constitutively

expressed homologue, heat shock cognate 70 (Hsc70)

[4,5] Hsc70 and Hsp70 are single-polypeptide

cytoplas-mic proteins composed of an N-terminal ATPase

domain, a substrate-binding region, and a C-terminal

15 kDa ‘lid’ regulating binding affinity for ‘client’

pro-teins These chaperones bind short (approximately

seven residues) hydrophobic regions exposed in either

mutant or immature wild-type (wt) client proteins

[6,7] The binding of Hsc70 and Hsp70 to substrates is

tightly coupled to cycles of ATP binding and

hydroly-sis followed by ADP release [8]

In our previously proposed model of the ERQC, the

folding states of CFTR are assessed within the ER at

multiple checkpoints [9,10], the first of these involving

the Hsc70⁄ Hsp70 machinery, the most critical one for

F508del-CFTR degradation Biochemical data suggest

that both immature wt-CFTR and F508del-CFTR

have Hsc70-binding sites, but the Hsc70 association

with F508del-CFTR is prolonged [11] However, there

is evidence that critical interaction sites for Hsc70

reside within the first half of CFTR, in particular

NBD1 [12,13] Indeed, analysis of the human (h) and

murine (m) CFTR-NBD1 sequences with the online

program limbo (http://switpc7.vub.ac.be/) performed

here predicts that they both contain at least three

hep-apeptide Hsp70-binding sites, with a high degree of

certainty (99%) Biochemical analyses of purified

NBD1 indicate that F508del reduces domain stability, and hence promotes aggregation [14,15] Molecular dynamics modelling studies suggest that F508del-NBD1 has more conformational freedom than the wild-type (wt), thus exposing its hydrophobic interior

to the solution and impairing its interdomain contacts [16,17] Collectively, these data implicate NBD1 as the most probable site for the interaction of CFTR with Hsc70

Several small molecules that improve CFTR folding, biogenesis and function were recently identified in high-throughput screens [18], such as corrector 4a (C4a) and corrector VRT-325 (V325), which promote trafficking of F508del-CFTR to the plasma membrane [19,20] However, the exact mechanism of action of these compounds and their putative binding sites on CFTR remain undefined

Here, we used a novel approach, surface plasmon resonance (SPR) [21], to quantify the interaction occur-ring between F508del-CFTR and Hsc70, versus that of the chaperone with wt-NBD1 Our data show that F508del-mNBD1 binds Hsc70 with approximately five-fold higher affinity than wt-mNBD1, and that both ATP and ADP dramatically reduce NBD1–Hsc70 binding Moreover, we also show that, in the presence

of a small molecule known to rescue the traffic of full-length F508del-mCFTR to the plasma membrane, the strength of the F508del-mNBD1–Hsc70 interaction is reduced by 30%

Results

Interaction of purified NBD1 and Hsc70 under the SPR conditions

To confirm whether NBD1 binds specifically to Hsc70 under conditions that would subsequently be used in SPR interaction studies, we immunoprecipitated Hsc70–NBD1 complexes in vitro from a mixture of the two purified proteins Electrophoresis of the purified F508del-mNBD1 and bovine Hsc70 prior to immuno-precipitation (Fig 1A) resolved single 30 and 73 kDa

l MINT-7265906 , MINT-7265964 , MINT-7265981 , MINT-7265951 : mNBD1 (uni-protkb: P26361 ) binds ( MI:0407 ) to Hsc70 (uniprotkb: P19120 ) by surface plasmon resonance ( MI:0107 )

l MINT-7265924 , MINT-7265939 : hNBD1 (uniprotkb: P13569 ) binds ( MI:0407 ) to Hsc70 (uni-protkb: P19120 ) by surface plasmon resonance ( MI:0107 )

l MINT-7265996 : Hsc70 (uniprotkb: P19120 ) binds ( MI:0407 ) to Apo-alpha-lactalbumin (uni-protkb: P00711 ) by surface plasmon resonance ( MI:0107 )

protein bands (left lane, both panels), which were

confirmed to be specific by the parallel western blot

analysis with L12B4 [antibody (Ab) against NBD1 of

CFTR] and SPA-815 (Ab against Hsc70), respectively

(right lane, both panels) The data in Fig 1B show

that purified Hsc70 is present in immunoprecipitated

complexes of either wt-mNBD1 (lanes 2, 3 and 5) or

F508del-mNBD1 (lane 4) incubated in SPR flow

buffer Moreover, the absence of Hsc70 in

immunopre-cipitates when BSA (Fig 1B, lane 1) was used instead

of NBD1, or in the absence of L12B4 (lane 7),

demon-strates that the binding of Hsc70 does not occur with

all protein substrates Increasing the concentration of

wt-hNBD1 by five-fold (Fig 1B, lane 3) did not

noticeably increase the amount of NBD1 or Hsc70

recovered (compare with lane 2), suggesting that the

L12B4-coated beads were already saturated with

NBD1 Addition of MgATP (2 mm) markedly reduced

the binding of Hsc70 to wt-mNBD1 (Fig 1B, lane 6),

consistent with the results of previous studies with

other Hsc70 substrates [8]

Analysis of stability of NBD1 under the SPR interaction conditions

We used intrinsic tryptophan fluorescence spectroscopy

to assess the structure and stability of NBD1 under assay conditions that would subsequently be used in SPR interaction studies A comparison of emission spectra obtained before (0 min) and after (10 min) incubation (Fig 2A) reveals that there was only a small decrease in the overall intensity of the fluores-cence emitted from F508del-mNBD1 and no shift in the peak wavelength Furthermore, the intensity of fluorescence emitted at 328 nm and 343 nm from diluted F508del-mNBD1 (Fig 2B) displayed a minor initial decrease, but was essentially stable during the

10 min incubation period (similar data were obtained with wt-mNBD1; not shown; n = 2) Previous studies indicate that, as compared with their folded forms, denatured wt-hNBD1 and F508del-hNBD1 display a dramatically reduced overall intensity of emitted fluo-rescence, and peak values are red-shifted to higher wavelengths [15] Hence, our data suggest that: (a) the wt-mNBD1 and F508del-mNBD1 used in this study have a structure consistent with a native fold [15,22]; and (b) the domain is stable under the conditions used

in SPR experiments

Assessment of optimal conditions for studying the Hsc70–NBD1 interaction by SPR

Before determining the effect of the F508del mutation

on the NBD1–Hsc70 interaction by SPR, we performed

0 5 6

0 2 4 6

Fluorescence (units) Buffer

0 min

10 min

343 nm

328 nm

F508del-m F508del-m

Fig 2 Spectroscopic analyses show that NBD1 is stable under the conditions used in SPR studies (A) Fluorescence emission spectra (300–400 nm) of F508del-mNBD1 (1 l M ) before (0 min) and after (10 min) incubation in buffer at 25 C (excitation wavelength of

295 nm) (B) Change in intrinsic tryptophan fluorescence emitted at

328 and 343 nm from F508del-mNBD1 (1 l M ) during incubation in buffer (25 C) Data are corrected for buffer fluorescence Similar results were obtained with other NBD1 preparations (F508del-mNBD1, n = 3; wt-(F508del-mNBD1, not shown, n = 2).

47

83

25

16

Hsc70

37 50 75

25

NBD1

1

Hsc70

IP: NBD1

1

–

–

1

1 –

1

5 –

1 1 –

1 1 –

1 1

2

1

1

– –

NBD1

A

B

Fig 1 In vitro interaction of NBD1 with Hsc70 confirmed by

immu-noprecipitation (IP) under conditions used in SPR studies.

(A) SDS ⁄ PAGE (P) and western blot (WB) analyses of purified

F508del-mNBD1 (F508del-NBD1; 5 lg) and bovine Hsc70 (1 lg)

with L12B4 and SPA-815, respectively, to confirm their specificity

(see Experimental procedures) (B) In vitro analysis of the

inter-action of Hsc70 with NBD1 by immunoprecipitation under

condi-tions equivalent to those used for SPR Protein G beads coated

with L12B4–NBD1 complexes were incubated with purified Hsc70

(1 l M ; 10 min at 25 C) and analysed by western blot using either

SPA-815 or L12B4 (see Experimental procedures) Lanes: 1, BSA

(3 l M ; without NBD1); 2, wt-hNBD1 (1 l M ); 3, wt-hNBD1 (5 l M ); 4,

F508del-mNBD1 (1 l M ); 5, wt-mNBD1; 6, wt-mNBD1 with MgATP

(2 m M ); 7, wt-hNBD1 with beads (without L12B4); 8, L12B4-coated

beads only (without NBD1 and Hsc70) Similar results were

obtained in three experiments (n = 3).

a series of control experiments to determine the

bind-ing specificity and optimal assay conditions To this

end, wt-hNBD1 and BSA were immobilized on sensor

chip surfaces, and the binding of L12B4 was

investi-gated The data in Fig 3A show that there was a

mea-surable and time-dependent interaction of Ab (15 nm)

with wt-hNBD1, but not with BSA-coated chip

surfaces (n = 3–5) These data indicate that even at a

low nanomolar concentration, there was potent

binding of L12B4 to immobilized NBD1 (23.7 ±

0.9 pmolÆnmol)1; n = 3), which is characteristic of

Ab–antigen interactions [23] The data also indicate

that we can use SPR to assay the binding of

interac-ting proteins to NBD1 To further optimize the

condi-tions for investigating the NBD1–Hsc70 interaction,

we tested sensor chips coated with NBD1 (on-chip

NBD1) Hsc70 (0.3 lm) bound specifically to sensor

surfaces coated with wt-mNBD1, F508del-mNBD1,

and wt-hNBD1 (Fig 3B; F508del-mNBD1, 2.5 ±

0.2 pmolÆnmol)1; wt-mNBD1, 2.9 ± 0.1 pmolÆnmol)1;

wt-hNBD1, 3.8 ± 0.2 pmolÆnmol)1; n = 3), with

mini-mal adsorption of Hsc70 onto BSA-coated surfaces

(< 0.2 pmolÆnmol)1) Hence, under these conditions,

the magnitude of Hsc70 binding to wt-mNBD1 was

substantially (approximately six-fold) lower than that

observed for L12B4

Then, we tested the reverse situation by immobiliz-ing Hsc70 on CM5 sensor chips (on-chip Hsc70), to investigate its ability to bind folded or unfolded pro-teins The immobilized chaperone was found to potently bind reduced apo-a-lactalbumin (Red-LA) (Fig 3C; 10 lm; 29.1 ± 3.8 pmolÆnmol)1; n = 3), an Hsc70 substrate [13], whereas minimal binding was detected with the nonreduced, folded form of the pro-tein (LA) (10 lm; 1.1 ± 0.7 pmolÆnmol)1; n = 3) As expected, virtually no binding of BSA (15 lm) to Hsc70-coated surfaces could be detected (< 0.1 pmo-lÆnmol)1; n = 20) under these conditions These data confirm that Hsc70 binding detected by SPR was only significantly detected for unfolded protein substrates Interestingly, when Red-LA was applied in the pres-ence of ATP (100 lm), the level of binding was reduced to  80% of that under nucleotide-free condi-tions (23.9 ± 1.9 pmolÆnmol)1; n = 3)

Next, we determined Hsc70 binding of all three NBD1 variants (0.5 lm; Fig 3D), and found that all bound potently to immobilized Hsc70 (F508del-mNBD1, 20.9 ± 2.0 pmolÆnmol)1; wt-mNBD1, 17.8 ± 2.3 pmolÆnmol)1; wt-hNBD1, 27.3 ± 3.8 pmolÆnmol)1; n = 4) Under these conditions, wt-hNBD1 displayed the highest level of Hsc70 binding, followed by F508del-mNBD1 Moreover, the addition of MgATP

0

20

40

0 3

6

B

–1 )

BSA

wt-h wt-m F508del-m

A

BSA

wt-h

0

20

40

0 20

40

D

BSA

F508del-m wt-h wt-m

wt-m + ATP

Time (s)

C

Time (s)

BSA

Red-LA

LA

Red-LA +ATP

Time (s) Time (s)

Fig 3 Hsc70 binds specifically to NBD1 and control proteins in SPR studies Charac-terization of the interaction of: (A) applied L12B4 (15 n M ) with covalently immobilized (‘on-chip’) wt-hNBD1; and (B) applied Hsc70 (0.3 l M ) with on-chip hNBD1, wt-mNBD1, or F508del-mNBD1 (see Experi-mental procedures) Hsc70 and L12B4 displayed minimal interaction with on-chip BSA [ , (A) and (B)] (C, D) The interaction

of applied Red-LA or LA (10 l M ) (C) and wt-hNBD1, wt-mMBD1 or F508del-mNBD1 (0.5 l M ) (D) with on-chip Hsc70 (On-Chip Hsc70; see Experimental procedures) BSA (15 l M ) showed minimal interaction with on-chip Hsc70 [–, (C) and (D), n = 15] Periods of protein application (and associa-tion) are indicated by the solid bars After protein application, dissociation was measured by injecting flow buffer over the protein-coated surface Similar results were obtained in additional experiments (n = 3 or

n = 4).

(100 lm) dramatically reduced the binding of

wt-mNBD1 to Hsc70 (Fig 1A; 3.7 ± 1.3 pmolÆnmol)1;

n= 3) These data show that we can use SPR to

accu-rately measure the specific binding of NBD1 to

immobi-lized Hsc70 The substantially higher binding of NBD1

to immobilized Hsc70 over that of Hsc70 chaperone

binding to immobilised NBD1 can be attributed to the

ability of Hsc70 to better survive the acidic, low-salt SPR

immobilization conditions than NBD1 [24] Hence, in

subsequent experiments, we decided to immobilize Hsc70

and quantify the binding of wt- and F508del-mNBD1

Assessment of the impact of F508del on the

NBD1–Hsc70 interaction

To determine the strength of the NBD1–Hsc70

inter-action, we thus immobilized the chaperone and

quantified the binding of increasing concentrations of

wt-mNBD1 and F508del-mNBD1 (Fig 4A,B) Using

kinetic analyses, we determined that rates of

associa-tion (ka) and dissociation (Fig 4C, kd) of wt-mNBD1

and Hsc70 are extremely slow [ka, 3660 ± 596 m)1Æs)1;

kd, (5.2 ± 1.4)· 10)5s)1; n= 3] In contrast,

F508del-mNBD1 bound Hsc70 with a comparable

association rate, but had a five-fold lower dissociation

rate [ka, 4030 ± 655 m)1Æs)1; kd, (1.0 ± 0.2)·

10)5s)1; n = 3) Using these constants, we calculated

that Hsc70 bound F508del-mNBD1 with five-fold higher affinity than wt-mNBD1 [dissociation constant (KD): F508del-mNBD1, 2.6 ± 0.5 nm; wt-mNBD1, 13.9 ± 0.8 nm; n = 3] These data indicate that when Phe508 is deleted, there is a significant increase in the real-time affinity of mNBD1 for Hsc70 (P < 0.01) Analysis of the dose–response data (Fig 4D) revealed that the maximum amount of F508del-mNBD1 bound

to Hsc70 at saturating concentration appears to be lower than that of wt-mNBD1 (Bmaxapp: F508del-mNBD1, 73.1 ± 3.9 pmolÆnmol)1; wt-mNBD1, 106.6 ± 6.6 pmolÆ nmol)1; n = 3) Moreover, apparent dissociation con-stants determined directly from the dose–response data (Eqn 1) indicate that F508del-mNBD1 bound Hsc70 with at least three-fold higher affinity than wt-mNBD1 (KDapp: F508del-mNBD1, 23.2 ± 1.7 nm; wt-mNBD1, 70.2 ± 4.7 nm; n = 3)

Analysis of the effect of adenine nucleotides on the binding of NBD1 to Hsc70

To further explore the effect of adenine nucleotides on the NBD1–Hsc70 interaction, we quantified the effect

of ATP and ADP on the binding of wt-mNBD1 and F508del-mNBD1 to immobilized Hsc70 As shown by the solid lines in Fig 5A,B, increasing concentrations

of ATP caused a dramatic reduction in the binding of

96 98 100 102

0 25 50 75 100 0 50 100

0 50 100

Time (s)

Time (s)

wt-m

F508del-m

0.02

0.1 0.2

0.05

0.5

0.02 0.1 0.2

0.05 0.5

F508del-m

wt-m

F508del-m

wt-m

Time (s)

Fig 4 F508del increases the affinity of

NBD1 for on-chip Hsc70 (A, B) The

inter-action of increasing concentrations of

(A) wt-mNBD1 and (B) F508del-mNBD1 with

on-chip Hsc70 (C) Dissociation of

Hsc70-bound wt-mNBD1 and F508

del-mNBD1 (0.1 and 0.2 l M ) shown on a

highly expanded scale Binding was

normalized independently for each

dissocia-tion curve to an initial maximum value (100)

at the start of each dissociation phase

(920 s) First-order fits ( ) to the data

denote relative dissociation rates (D) The

change in amount of NBD1 bound by

on-chip Hsc70 at equilibrium with increasing

concentrations of wt-mNBD1 (d) and

F508del-mNBD1 (s; mean ± SEM; n = 3).

Other details as in legend to Fig 3.

wt-mNBD1 and F508del-mNBD1 (0.5 lm) to Hsc70.

In contrast, the introduction of a control molecule,

NADP (500 lm), did not significantly reduce the

bind-ing of either wt-mNBD1 or F508del-mNBD1 (0.5 lm)

to immobilized Hsc70 (Fig 5C; wt-mNBD1, P = 0.45;

F508del-mNBD1, P = 0.46; n = 3) Interestingly, the

binding of wt-mNBD1 and F508del-mNBD1 to Hsc70

was also reduced by the addition of ADP (500 lm),

although, at this concentration, the degree of

inhibi-tion was less than for ATP (P = 0.01; n = 3)

Fig-ure 5C, displaying a comparison of the summarized

data for wt-mNBD1 and F508del-mNBD1, shows that

the interaction of wt-mNBD1 with Hsc70 was most

potently inhibited by increasing ATP concentration

(IC50, 19.7 ± 2.1 lm; n = 3) For F508del-mNBD1,

the ability of ATP to inhibit the Hsc70 interaction was

significantly reduced (IC50, 111 ± 4.8 lm; P = 0.001;

n= 3) Collectively, our data suggest that the NBD1– Hsc70 interaction is inversely dependent on the ATP concentration and that F508del stabilizes this interaction

The effect of small molecule compounds on the NBD1–Hsc70 interaction

Small molecules recently identified in high-throughput screens have been proposed to affect CFTR-NBD1 conformation [25,26] Here, we tested by SPR whether and how these molecules affect the NBD1–Hsc70 interaction The data in Fig 6A show there was a signifi-cant reduction in the Hsc70 binding of F508del-mNBD1 (1 lm) upon acute application of C4a, which was not

0 25 50

0 25 50

0

25

50

[Compound] (µ M ) Time (s)

C4a

Con

I172

Incubation Incubation

+ ATP

C4a V325 I172

*

* V325

F508del-m (1.0 µ M ) F508del-m (1.0 µ M )

Fig 6 C4a alters the binding of F508del-NBD1 to on-chip Hsc70 (A) Interaction of F508del-mNBD1 (1 l M ) with on-chip Hsc70 in the absence ( , Con) and presence (50 l M ) of: C4a (–), V325 ( ), or I172 (gray ) (B) Change in amount of F508del-mNBD1 bound by on-chip Hsc70 in the presence of I172 ( ; 50 l M ) or increasing concentrations of V325 (s) and C4a (d; mean ± SEM; n = 3) (C) Binding of F508del-mNBD1 to Hsc70 ( h; Con ) following incubation of NBD1 (1 l M ) with C4a ( ) or V325 ( ; 50 l M ) for 30 min at 16 C in SPR buf-fer alone (Incubation) or in SPR bufbuf-fer plus 100 l M MgATP (Incubation + ATP; mean ± SEM; n = 3) Other details as in the legend to Fig 3.

0 10 20 30

0

10

20

30

1 10 100 1000

0 10 20 30

103

23

203

wt-m

F508del-m

3

53

503 103 23

203

3

53 wt-m F508del-m

ATP ADP NADP

Fig 5 Adenine nucleotides reduce the binding of NBD1 to on-chip Hsc70 The interaction of (A) wt-mNBD1 (0.5 l M ) and (B) F508del-mNBD1 (0.5 l M ) with on-chip Hsc70 in the presence of NADP (– –, 500 l M ), MgADP ( , 500 l M ) or increasing concentrations of MgATP (–) (C) The change in binding of wt-mNBD1 (d) and F508del-mNBD1 (s) to on-chip Hsc70 in the presence of NADP (D, ; 500 l M ), MgADP (j,h; 500 l M ), or increasing concentrations of MgATP (d,s) (mean ± SEM; n = 3) Other details as in the legends to Figs 3 and 4.

observed for V325 or inhibitor CFTRinh-172 (I172;

n= 3) We quantified the Hsc70 binding of

F508del-mNBD1 in the presence of increasing

concen-trations of C4a and V325 (Fig 6B) The data show that

only at the higher concentration tested (50 lm) could

C4a, but not V325 or I172, significantly reduce the

Hsc70 binding of F508del-mNBD1, by 30% (control,

32.2 ± 2.2 pmolÆnmol)1; C4a, 23.4 ± 1.9 pmolÆnmol)1;

P= 0.04; n = 3) To better understand the mechanism

of action of these correctors, we also preincubated

F508del-NBD1 with either C4a or V325 (50 lm, 30 min,

16C) prior to application, and then quantified NBD1

binding to Hsc70 As shown in Fig 6C, again only C4a,

and not V325, significantly reduced the binding of

F508del-mNBD1 to Hsc70 (C4a, P = 0.04; V325,

P= 0.58; n = 3) Comparison of the data with those in

Fig 6B indicates that the magnitude of inhibition

( 30%) was comparable to that observed with acute

compound application However, following

preincuba-tion of NBD1 with C4a or V325 at increased MgATP

concentration (106 lm), we observed no effect of these

corrector compounds on the NBD1–Hsc70 interaction

(Fig 6C; P = 0.74–0.95; n = 3)

Discussion

Various components of the ERQC recognize aberrant

conformations of secretory proteins and target them

for proteasomal degradation so as to avoid clogging of

the secretory pathway In the case of the CFTR protein

bearing F508del, the major CF-causing mutation, it has

been shown that the molecular chaperone Hsc70 plays

a major role in this disposal mechanism [11,12] In the

present study, we have used an SPR approach to

inves-tigate how deleting F508 from NBD1 of mCFTR alters

the interaction of the domain with the molecular

chap-erone Hsc70, and whether small molecule correctors of

CFTR folding can affect this critical interaction

Both wt-NBD1 and F508del-NBD1 bind Hsc70

Our data indicate that Hsc70 can interact with both

wt-NBD1 and F508del-NBD1, in agreement with

pre-vious studies [12] showing that both wt-CFTR and

F508del-CFTR can associate with this chaperone via

NBD1 The strength of NBD1–Hsc70 binding determined

here by SPR is substantially higher (low nanomolar KD)

than the majority of previous non-SPR assessments of

Hsc70⁄ Hsp70–substrate interactions (nanomolar to

micromolar KD values [27–29]) This is likely to reflect

the fact that we have used a potent Hsc70 substrate

(isolated NBD1 of CFTR) Moreover, the conditions

under which the experiments were performed favour

high-affinity interaction of Hsc70 with a client protein Nonetheless, although we could detect interaction of Hsc70 with Red-LA, this was not the case for LA In addition, previous measurements of the affinity of Hsc70–substrate binding have almost exclusively been the result of peptide, and not protein, binding by Hsp70⁄ Hsc70 [6–8,30]

Previous studies have used SPR to quantify the binding of nonchaperone proteins (the cytoskeletal, PDZ-anchoring protein NHERF⁄ EBP50 and the related, PDZ-based scaffold protein Shank2) to C-ter-minal CFTR peptides and, in one case (the Ca2+⁄ lipid-dependent annexin A5), isolated NBD1 Although the proteins studied are unrelated to Hsc70⁄ Hsp70, they were also found to bind CFTR fragments with high affinities (KD; C-terminus–EBP50, 22 nm; C-ter-minus–Shank2, 56 nm [31]; NBD1–annexin A5, 4 nm [32]) However, in contrast to our findings, Trouve

et al [32] found that ATP increased the binding of NBD1 to annexin A5, and that deleting Phe508 had

no effect on the annexin A5–NBD1 interaction Inter-estingly, they demonstrated that CPX, a potential NBD1 small molecule ligand [33], inhibited binding Overall, the data suggest that Hsc70 interacts with NBD1 at alternative sites to annexin A5 and with different characteristics

The enhanced binding of Hsc70 to wt-hNBD1 rela-tive to wt-mNBD1 that we find here may be due to variations in amino acids between these forms of NBD1 that help to stabilize this domain against inter-action with Hsc70 Interestingly, when some of the variant residues in NBD1 of F508del-mCFTR are substituted for residues at corresponding positions in F508del-hCFTR, they act as revertants of the folding and trafficking defects [14] This is the case for Thr539 (Ile539 in humans), a so-called revertant of F508del-hCFTR [34], and also Ser429 (Phe429 in humans), recently shown to contribute to rescue of the traffick-ing defect of F508del-hCFTR [35] The presence of these ‘profolding’ residues in mNBD1 is a probable explanation for the recently reported attenuated processing⁄ trafficking defect of F508del-mCFTR in comparison with F508del-hCFTR [36]

F508del increases the affinity of CFTR NBD1 for Hsc70

Our data demonstrate that deleting Phe508 from mNBD1 increases five-fold the affinity of the domain for the Hsc70 chaperone However, our data also indi-cate that Hsc70 binds  30% more wt-mNBD1 than F508del-mNBD1 at saturation One explanation for this reduction is that the deletion of Phe508 increases

the affinity of NBD1 for not only Hsc70 but also itself.

Indeed, as F508del-NBD1 is more prone to aggregation

than wt-NBD1 [14,15], the aggregated, mutant form

would thus be incapable of binding to Hsc70 This

would effectively reduce the average concentration of

the free monomeric form of F508del-mNBD1 available

to bind Hsc70 and hence the maximum binding

In a recent study, it was found that, in vivo, both

wt-CFTR and F508del-CFTR bind equal amounts of

the Hsp70 chaperone [37] Hence, the differences in

binding caused by the F508del mutation observed here

might be due to the in vitro nature of the SPR

approach versus in vivo complexity Within the cell,

additional factors influence Hsc70–CFTR binding, e.g

cobinding and competition of Hsc70 cochaperones,

such as the E3-ubiquitin ligase CHIP, which promotes

the fast dissociation of the Hsc70–CFTR complex,

targeting CFTR for proteasomal degradation [38–40]

Moreover, some Hsc70⁄ Hsp70 interaction sites are

likely to occur only in the context of the native

confor-mation of the full-length protein [5], as proposed in a

recent structural model of CFTR, where the surface of

NBD1 containing Phe508 mediates an interdomain

contact with intracellular cytoplasmic loop 4 of

mem-brane-spanning domain 2 [16] Nevertheless, our

find-ings are consistent with the reported prolonged

association of Hsc70 with F508del-CFTR relative to

that with wt-CFTR [11], which appears to constitute

the first ERQC checkpoint occurring in vivo [9,12]

Our data reported here thus suggest that the absence

of Phe508 increases the accessibility of Hsc70 to one

or more binding sites on NBD1

Binding sites on NBD1 promoting Hsc70

interaction

A critical issue in this field is the location of the

func-tionally important Hsc70-binding site(s) on NBD1; in

particular, whether removal of Phe508 creates a novel

Hsc70-binding site in NBD1 Although it is known

that Hsc70 and Hsp70 bind short, hydrophobic peptide

pockets exposed on substrate proteins, the exact

pri-mary sequence of these peptides is variable [6,30]

Accordingly, the NBD1 proteins employed in this

study (Thr389–Gly673) contain many of these short,

hydrophobic sequences, including the region around

Phe508, constituting potential Hsc70-binding sites

Analysis of hNBD1 and mNBD1 sequences with the

limbo program, which predicts likely binding sites of

the Hsp70⁄ Hsc70 homologue, DnaK, identified three

novel regions (Ser466–Leu472, Leu568–Pro574, and

Asp614–Gln621), all three of which are distant from

Phe508 Qu and Thomas [14] localized a putative

Hsc70-binding site to Gly545–Ala561 [13], a hydropho-bic pocket that is partially exposed in the crystal struc-ture of NBD1 and that is also distant from Phe508 However, the residue limits of the NBD1 used in this study (Gly404–Ser589) were different from those used

in the present study

Nevertheless, even if no additional binding sites occur in F508del-mNDB1, they may be more accessi-ble to Hsc70, in comparison with wt-mCFTR Sup-porting this notion, biochemical analyses of purified NBD1 indicate that F508del promotes aggregation, a property known to result from exposure of hydro-phobic residues [14,15] Moreover, molecular dynamics modelling studies suggest that F508del-NBD1 exposes its hydrophobic interior to the solution more often than wt-NBD1 [16,17]

Effect of ATP on the Hsc70–NBD1 interaction Adenine nucleotides are of critical importance to the binding of Hsc70 to CFTR [8] Consistently, our data show that ATP dramatically reduces the ability of mNBD1 to bind Hsc70 However, ATP can mediate this effect in two distinct ways First, the binding and hydrolysis of ATP at the nucleotide-binding domain of Hsc70 is known to accelerate its binding and release of substrates, reducing the affinity of their interaction Once ADP occupies the nucleotide-binding site of Hsc70, it converts the chaperone to a high-affinity binding form [8] Second, ATP, a native ligand of CFTR-NBD1, may bind to this domain, stabilizing its structure [14,41] and possibly hindering one (or more) Hsc70-binding site(s) Several lines of evidence argue that the effect of ATP on the NBD1–Hsc70 interaction observed here is mediated predominantly via NBD1 and not by Hsc70 First, although ATP reduced NBD1 binding to Hsc70, it did not alter the associa-tion–dissociation profile, suggesting that it did not alter the kinetics of the interaction Second, ADP, rather than enhancing NBD1–Hsc70 binding, as pre-dicted for an Hsc70-mediated effect, substantially reduced binding Third, the IC50 values for ATP inhi-bition of wt-mNBD1 and F508del-mNBD1 binding to Hsc70 ( 20 and  110 lm, respectively, this study) are comparable to the previously reported apparent dissociation constants for ATP and wt-hNBD1 and F508del-hNBD1 ( 90 lm [42]) but substantially higher than the dissociation constant for ATP and Hsc70, which is in the order of  0.7 lm [43] Finally, ATP caused a reduction of only  20% in the Hsc70 binding of denatured lactalbumin, itself not predicted

to interact with ATP, whereas Hsc70 binding to wt-mNBD1 was reduced by  80% Our data also

show that the ATP concentration required to inhibit

the mNBD1–Hsc70 interaction is dramatically

increased by deleting Phe508 Less optimal ATP

bind-ing to F508del-NBD1 would be consistent with the

gating defect of F508del-CFTR, characterized by long

interburst intervals [44], according to the current

model of CFTR channel gating [45] However,

regard-ing the isolated domain, ATP binds wt-NBD1 and

F508del-NBD1 with equivalent affinity, arguing

against such an effect However, it remains plausible

that the F508del-induced exposure of Hsc70-binding

sites on NBD1 diminishes the ability of ATP binding

to stabilize this domain against its Hsc70 interaction

Altogether, interpretation of these data would be

con-sistent with ATP promoting ordered homodimerization

of NBD1, a process that conceivably occludes the

chaperone-binding site to reduce Hsc70 binding It

has, in fact, been reported that wt-NBD1 can form

homodimers under certain conditions [46,47] Thus, a

reduced dimerization ability of F508del-NBD1 relative

to wt-NBD1 would explain why higher ATP

concen-trations are required to inhibit its binding to Hsc70

Effect of small molecules on the interaction of

Hsc70 with F508del-NBD1

Here, we also used SPR to determine whether C4a and

V325 have an effect on the NBD1–Hsc70 interaction

Our data demonstrate that C4a did indeed reduce

F508del-mNBD1 binding to Hsc70 This suggests that

C4a binds directly to F508del-mNBD1, either at the

same binding site, or by allosteric stabilization of

NBD1, to influence its folding and thus decrease the

affinity of the domain for the Hsc70 chaperone This

finding is consistent with the results of recent in vivo

studies on full-length F508del-hCFTR [20,26]

However, it should be noted that substantially higher

concentrations of C4a were required to affect the

NBD1–Hsc70 interaction than those needed to correct

cell surface expression of CFTR, although, in vivo, a

more complex mechanism of action may occur than

the simple disruption of an NBD1–Hsc70 complex A

plausible explanation, nevertheless, is that C4a binds

directly to a site on monomeric NBD1 that stabilizes

homodimerization, similarly to the ATP effect (see

above) This could likewise occlude the Hsc70-binding

site(s) and thus reduce Hsc70 binding Strikingly, we

observed a significantly reduced effect of C4a at high

ATP concentrations

Our data demonstrate that SPR provides a powerful

approach to quantifying the Hsc70–NBD1 interaction

and the impact of F508del (or other NBD1 mutations)

and corrector compounds on this folding-sensitive

association In particular, our data show that ATP dra-matically reduces the ability of mNBD1 to bind Hsc70, and indicate that it is mostly this ability of ATP to displace Hsc70 from NBD1 binding that is impaired by F508del C4a significantly inhibits the F508del-mNBD1 interaction with Hsp70, suggesting direct binding As this effect is abolished at high ATP concentrations, ATP and C4a may compete for the same F508del-mNBD1 binding site⁄ surface We conclude that this SPR approach constitutes a useful assay for the determina-tion of whether and how correctors affect the Hsc70– F508del-NBD1 interaction, which will improve our understanding of the mechanism of action of small molecules with therapeutic potential for CF, a critical step in bringing them to the clinical setting Future stud-ies should focus on the quantification of the effect of these correctors on the NBD1–Hsc70 interaction in the presence of other relevant CFTR domains (e.g intra-cellular cytoplasmic loop 4 and⁄ or nucleotide-binding domain 2) at controlled MgATP concentrations

Experimental procedures

Reagents All three forms of NBD1 used in this study (wt-hNBD1, wt-mNBD1, and F508del-mNBD1; Thr389–Gly673) were prepared using the same protocol as previously described [22] Briefly, the proteins were expressed in BL21 (DE3 strain) Escherichia coli from pSmt3 vectors, purified by nickel-affinity and size-exclusion chromatography, and stored at )80 C [10 mgÆmL)1NBD1, 20 mm Tris, 150 mm NaCl, 5 mm MgCl2, 2 mm ATP, 2 mm b-mercaptoethanol, 12.5% (v⁄ v) glycerol, pH 7.6] Postpurification, each preparation of NBD1 protein (wt-hNBD1, wt-mNBD1, and F508del-mNBD1) was analysed by intrinsic tryptophan fluorescence and CD spectroscopy to confirm that it carried

a native fold (personal communication: data at http://www cftrfolding.org/reagentRequestshWT.asp) Because F508del-hNBD1 CFTR is highly insoluble and prone to aggregation during purification [14], it was not available for use in this study Hence, wt-mNBD1 and F508del-mNBD1 were used

to characterize the impact of the F508del mutation on CFTR–Hsc70 interactions Purified bovine Hsc70 protein 751) and monoclonal rat Ab against Hsc70 (SPA-815) were obtained from Assay Designs (Ann Arbor, MI, USA) Monoclonal mouse Ab against NBD1 (L12B4) was obtained from Chemicon (Temecula, CA, USA) Biacore materials were obtained from GE Healthcare (Milwaukee,

WI, USA) All other chemicals, proteins (BSA and bovine apo-a-lactalbumin) and reagents were purchased from Sigma Aldrich (St Louis, MO, USA) or BDH (Poole, UK), and were of research grade or higher (‡ 99% purity)

Biochemical analysis of purified NBD1 and Hsc70

Coimmunoprecipitation of purified NBD1 (1 lm, unless

otherwise stated) and Hsc70 (1 lm) in SPR flow buffer

[150 mm KCl, 2 mm MgCl2, 0.1% (v⁄ v) Triton X-100,

0.1% (v⁄ v) dimethylsulfoxide, 1 mm b-mercaptoethanol,

20 mm Hepes, pH 7.0] was performed using L12B4, as

pre-viously described [39] SDS⁄ PAGE and western blot using

SPA-815 (Hsc70) and L12B4 (NBD1) were also performed

as previously described [39] Protein quantification was

performed with a modified Lowry method

Spectroscopic analysis of NBD1

Wild-type and F508del-mNBD1 (1 lm) were diluted in

NaCl⁄ Pi (150 mm NaCl, 20 mm Na2PO4, 1 mm

b-mercap-toethanol, pH 7.4) and, following excitation at 295 nm, the

emitted intrinsic tryptophan fluorescence was measured at

25C as previously described [15] The intensity of

fluores-cence emitted from NBD1 at 328 and 343 nm was corrected

for background fluorescence from buffer at these two

wave-lengths

SPR

Interaction analyses were performed in SPR flow buffer

at constant temperature (25C), using a Biacore 2000

system (GE Healthcare) as previously described [6]

Ligand proteins (20 lgÆmL)1 in sodium acetate; 10 mm,

pH 5.0) were covalently immobilized (‘on-chip’) on the

surface of carboxymethyl-dextran (CM5) sensor chips,

according to the manufacturer’s instructions (estimated

final concentrations of immobilized proteins on sensor

chip surface: 0.8–1.5 mm) The binding and dissociation

of free analyte (‘applied’) proteins at a constant flow rate

(30 lLÆmin)1, 180 lL) was then measured The surface of

the chip was regenerated between sample applications

with sequential injections of HCl (10 mm, 20 lL) and

NaOH (10 mm, 20 lL) LA (20 mgÆmL)1) was reduced by

incubating in SPR flow buffer containing dithiothreitol

(20 mm, 60 min, 4C) as previously described [48]

Red-LA was then diluted to 10 lm (145 lgÆmL)1) and applied

immediately

The effect of small molecules (C4a, V325, I172; 10 mm,

100% dimethylsulfoxide) was determined by diluting these

compounds to different concentrations (as indicated in

the figure legends) in flow buffer containing BSA

(0.2 mgÆmL)1), and applying them either immediately (acute

effect) or after 30 min of incubation at 16C (incubated

effect) The final dimethylsulfoxide concentration was

adjusted to 0.5% Protein-coated CM5 chips were used for

2 weeks, or until nonspecific binding increased (‡ 5%) All

experiments were performed in parallel with an inactivated,

or blank, flow cell not coated with protein

Data analysis All SPR sensograms were corrected for buffer-induced refractive index changes at an uncoated reference surface, analysed using biaevaluation software (biaeval; v 3.2;

GE Healthcare), and displayed in sigmaplot (v 10; Systat, San Jose, CA, USA) as pmoles of interacting protein (e.g NBD1) bound per nmole of immobilized protein (e.g Hsc70) Molar concentrations of the proteins were calcu-lated from their measured concentration (mgÆmL)1), using their molecular masses as determined from amino acid com-position (wt-NBD1, 31 976 Da; F508del-NBD1, 31 829 Da; wt-hNBD1, 31 969 Da; bovine Hsc70, 71 241 Da)

The kinetics of interaction (association, ka, and dissocia-tion, kd, rates) were determined from each set of dose– response data by global fitting of the association and dissociation phases of all binding curves in that dataset (biaeval) The dissociation constant (KD) for each dose– response set was then determined (kd⁄ ka), and the values were averaged (mean KD) To determine the apparent maximal binding (Bmaxapp, pmolÆnmol)1) and dissociation constant (KDapp, lm) of mNBD1 directly from the data (Eqn 1), the amount of mNBD1 bound at equilibrium (Beq, pmolÆnmol)1) was determined from kinetic analysis and plotted against [mNBD1] (sigmaplot) The [MgATP] required to inhibit binding by 50% [IC50, lm; Eqn (2)] was determined by plotting the amount of mNBD1 bound at

320 s against [MgATP]

Beq¼Bmax

app ½NBD1

KDappþ ½NBD1 ð1Þ

B¼ Bminþ Bmax Bmin

1þ 10ðIC50½ATPÞ ð2Þ where: Beqis binding at equilibrium; Bmaxappand KDappare the apparent maximum binding and dissociation constants, respectively; B is binding (at 320 s); Bmin and Bmax are minimum and maximum binding, respectively; and [NBD1] and [ATP] are the concentrations of NBD1 and ATP, respectively

Statistical analysis Unless otherwise stated, data are presented as the mean ± standard error of the mean (SEM) (n‡ 3) Mean data for NBD1 dose responses were calculated as follows: (a) global kinetic analysis of each dose–response set generated a kd and ka value for each [NBD1]; (b) these values were averaged; and (c) the averaged values were used to calculate a single KD (kd⁄ ka) The mean KD was then determined by averaging KD values from repeat dose– response experiments (n = 3) For mean KDapp and Bmaxapp values, the KDapp and Bmaxapp were determined directly for each dose–response dataset as described, and the