Nucleotide inhibition of the pancreatic ATP-sensitive K+ channel explored with patch-clamp fluorometry

Pancreatic ATP-sensitive K+ channels (KATP) comprise four inward rectifier subunits (Kir6.2), each associated with a sulphonylurea receptor (SUR1). ATP/ADP binding to Kir6.2 shuts KATP. Mg-nucleotide binding to SUR1 stimulates KATP. In the absence of Mg2+, SUR1 increases the apparent affinity for nucleotide inhibition at Kir6.2 by an unknown mechanism. We simultaneously measured channel currents and nucleotide binding to Kir6.2. Fits to combined data sets suggest that KATP closes with only one nucleotide molecule bound. A Kir6.2 mutation (C166S) that increases channel activity did not affect nucleotide binding, but greatly perturbed the ability of bound nucleotide to inhibit KATP. Mutations at position K205 in SUR1 affected both nucleotide affinity and the ability of bound nucleotide to inhibit KATP. This suggests a dual role for SUR1 in KATP inhibition, both in directly contributing to nucleotide binding and in stabilising the nucleotide-bound closed state.


21
ATP-sensitive K+ channels (K ATP ) couple the metabolic state of a cell to its electrical activity (Ashcroft 22 and Rorsman, 2013). In pancreatic β-cells, closure of K ATP in response to glucose uptake trig-23 gers insulin secretion. As such, mutations in K ATP that affect its response to changes in cellular 24 metabolism cause diseases of insulin secretion, e.g. neonatal diabetes and persistent hyperinsu-  (Tucker et al., 1997). 37 In addition to nucleotide-dependent activation, SUR1 confers several other properties on the ulum and correct trafficking to the plasma membrane (Zerangue et al., 1999). 45 To date, the primary means of studying nucleotide-dependent effects on K ATP channel func-  Figure 1B). As we show here, this method is readily combined with patch-clamp electrophysiology 55 so that nucleotide binding and regulation of current can be measured simultaneously. This has 56 enabled us to quantitatively assess nucleotide binding to Kir6.2 and explore how this is coupled to 57 channel inhibition in both wild-type K ATP and K ATP carrying mutations that impair ATP inhibition.

59
Measuring nucleotide binding to Kir6.2. We previously used this FRET-based binding assay to 60 measure nucleotide binding to the second nucleotide-binding site of SUR1 (Puljung et al., 2019). 61 To measure binding to Kir6.2 in the complete K ATP complex (four full-length Kir6.2 subunits co-62 expressed with four full-length SUR1 subunits), we replaced a tryptophan at position 311 (W311) 63 that is 26 Å from the location of the inhibitory nucleotide-binding site on Kir6.2 with ANAP ( Figure   64 1C) such that each subunit is labelled with one ANAP molecule. We designate this construct Kir6.2*.  TNP-ATP (red, from PDB accession #5XW6) was docked into the nucleotide-binding site of Kir6.2 and positioned in NBS1 of SUR1 (green, from PDB accession #3AR7) by alignment as described in Materials and Methods. Distances from the native tryptophan at position 311 in Kir6.2 to the fluorescent moieties of the TNP-ATPs are displayed in Å. D. Theoretical FRET efficiency between ANAP and TNP-ATP as a function of distance, calculated from the Förster equation. The distances and corresponding FRET efficiencies between ANAP at position 311 and TNP-ATP bound to Kir6.2 (E Kir ) and SUR1 (E SUR ) are indicated. E. Spectral images acquired from an unroofed membrane expressing Kir6.2*-GFP + SUR1 and exposed to increasing concentrations of TNP-ATP. The y-dimension in each image represents distance. The x-dimension represents wavelength.
F. Line-averaged, background-subtracted spectra from E displayed with increasing concentrations of TNP-ATP coloured from purple to orange.     of W311 with ANAP did not affect inhibition by K ATP . Both subunits also showed similar sensitiv-91 ity to TNP-ATP, which inhibited with a higher apparent affinity relative to ATP (Figure 1- Figure   92 supplement 2B,C). quenching in Kir6.2*-G334D-GFP may also be due to an allosteric effect of the G334D mutation on 134 channel gating. We feel that this interpretation is unlikely, as G334D has been shown to have no 135 effect on the unliganded of K ATP (Proks et al., 2010). 136 Measuring current inhibition and nucleotide binding simultaneously. The apparent affin-137 ity of Kir6.2*-GFP + SUR1 for TNP-ATP in unroofed membranes was 25.6 µM ( Figure 1G and Table   138 1). This value is higher than the apparent affinity for nucleotide inhibition (6.2 µM) measured using   153 Strikingly, current inhibition occurred at a lower range of concentrations compared to nucleotide 154 binding ( Figure 2C,D). The apparent 50 for inhibition calculated from Hill fits was an order of 155 magnitude lower than the 50 for binding measured in the same patches ( Figure 2C, Table 2). 156 We considered several different gating models to explain this observation. In each model, we as- for the free parameters of our MWC-type model ( Figure 2F, Table 3). Based on these distributions,  vertical line. B. Current (left) and spectra (right) acquired from the same excised, inside-out patch exposed to TNP-ATP and coloured according to concentration. C. Concentration-response (n = 9) for TNP-ATP inhibition of Kir6 Table 3). This is a consequence of being so low that even in the MWC-   Table 3. E. Posterior probability distributions for the full MWC-type model fit to Kir6.2*-C166S-GFP + SUR1 or Kir6.2*-GFP + SUR1 (data from Figure 2F) overlaid on the prior probability distribution. small change in nucleotide affinity. This is an unexpected finding, as one might expect that an in-204 crease in would allosterically cause a decrease in the apparent affinity for inhibitory nucleotide 205 binding. To resolve this conflict, we again turned to PCF ( Figure 3C,D). Rundown was much slower 206 for Kir6.2*-C166S-GFP + SUR1, which may reflect the increased of this construct. Measuring 207 current inhibition in combination with nucleotide binding confirmed that whereas the apparent 208 nucleotide affinity was unchanged by the C166S mutation, current inhibition occurred at much 209 higher concentrations compared to binding ( Figure 3D). How can we explain this paradox? Fits of 210 the data with our MWC-type model ( Figure 3D,E) suggest that, in addition to the expected effect 211 on , the C166S mutation profoundly affects the ability of bound ligand to close the channel ( ) 212 without affecting ( Figure 3E, Table 3). We propose that, in addition to increasing the of the allosteric effect of SUR1 on nucleotide binding. To explore the effect of SUR1 more rigorously, we 245 again turned to PCF.

246
As Kir6.2*-GFP expression in the absence of SUR1 was not sufficient for PCF recordings, we 247 took a mutational approach to better understand the role of SUR1 in inhibitory nucleotide binding. ization mutation ( Figure 5E). However, in addition to direct effects of K205 on nucleotide binding, 263 we also observed a shift in for both mutations ( Figure 5E). This suggests a dual role for SUR1 in 264 K ATP inhibition, both in contributing to nucleotide binding and in stabilizing the nucleotide-bound 265 closed state.

267
We have developed a novel approach that allows for site-specific measurement of nucleotide bind-268 ing to K ATP and concomitant measurements of channel current. Performing these measurements 269 simultaneously allowed us to examine nucleotide regulation of K ATP function in great detail. We    which was specific to K ATP . FRET also provided the spatial sensitivity necessary to discriminate 285 between nucleotide binding directly to Kir6.2 and to the nucleotide-binding sites of SUR1. We as- . 291 The majority of this earlier work was performed using single-channel measurements of mutated 292 and/or concatenated channel subunits. In this study, we confirm these results using minimally 293 perturbed channels with nucleotide sensitivity similar to that of wild-type K ATP (Figure 1- Figure   294 supplement 2A). By using an MCMC approach to model fitting, we can also evaluate our models to 295 assess how well the derived parameters were determined by the data. MCMC fits provide a basis 296 for determining credible intervals for our parameter estimates. This allows for direct comparison 297 of values derived from wild-type and different mutant constructs.

298
Although we did not explicitly include the effects of PIP 2 on K ATP gating in our model formula-299 tions, we assumed that the effects of PIP 2 on were implicitly modelled in our parameter ; 300 i.e. rundown due to dissociation of PIP 2 manifests as a decrease in rather than a change in the 301 number of channels. Although we were able to extract identifiable parameter estimates for , 302 and , our estimates of for each model we considered were appreciably less well constrained 303 than for the other parameters. We expect that this uncertainty arises from measuring a hetero-

311
Previous studies suggest that, whereas K ATP closure occurs via a concerted mechanism, individ-312 ual nucleotide binding events at Kir6.2 are not equivalent (Markworth et al., 2000). Earlier attempts 313 to determine the stoichiometry of inhibitory nucleotide binding to Kir6 to be closed when just one molecule of nucleotide is bound.

322
It has been proposed that there is direct negative cooperativity between binding events at dif-323 ferent subunits on Kir6.2 (Wang et al., 2007). We fit our data to an extended MWC-type model    (Yan et al., 2007; Lin et al., 2015). eRF1-E55D was included to increase 433 efficiency of ANAP incorporation (Schmied et al., 2014). Experiments were carried out 2-4 days  (Li et al., 2000). Thus, we did not consider this site 438 for further experimentation.

458
The HA tag plus a short linker (YAYMEKGITDLAYPYDVPDY) was inserted in the extracellular region

462
We consistently observed a lower molecular weight band as well. This band must correspond to 463 an N-terminally truncated Kir6.2 product, as the apparent molecular weight shifted with addition 464 of the C-terminal GFP tag. Based on the molecular weight, we predict that the truncated protein 465 product initiated from a start codon in the first transmembrane domain. Therefore, we believe 466 it is unlikely that this protein would form functional channels or traffic to the plasma membrane.   respectively. Images were captured on an EMCCD camera (ImagEM; Hamamatsu Photonics; Wel-484 wyn Garden City, UK) binned at 2 x 2 pixels and analysed using Python. A median filter with a 485 box size of 32 x 32 pixels was applied to improve the signal-to-noise ratio by reducing background 486 fluorescence. 487 We examined the surface expression of our ANAP-labelled constructs using confocal microscopy 488 (Figure 1-Figure supplement 1A,B). When Kir6.2-W311 TAG -GFP was co-transfected with SUR1 along 489 with pANAP and eRF1-E55D in the presence of ANAP, the ANAP and GFP fluorescence were co-490 localized at the plasma membrane. When wild-type Kir6.2-GFP was transfected under the same 491 conditions, only GFP fluorescence was observed at the plasma membrane. ANAP fluorescence 492 was diffuse and confined to the cytoplasm or intracellular structures. Thus, the plasma-membrane Surface expression assays. 495 We measured surface expression of HA-tagged Kir6.2 subunits using an approach outlined by 496 Zerangue et al. (Zerangue et al., 1999; Puljung et al., 2019). Cells were plated on 19 mm cov-497 erslips coated with poly-l-lysine and transfected as described above. Following incubation, cells 498 were rinsed with PBS before fixation with 10% formalin for 30 minutes at room temperature. Af-499 ter washing again, cells were blocked with 1% BSA in PBS for 30 minutes at 4 ∘ C before a 1-hour 500 incubation at 4 ∘ C with a 1:1000 dilution (in PBS) of rat anti-HA monoclonal antibodies. Cells were 501 then washed 5 times on ice with 1% BSA in PBS followed by a 30-minute incubation at 4 ∘ C with a

564
For kinetic measurements, the solution changer and camera were controlled using pClamp 565 10 software coupled to a Digidata 1322A digitizer. Each fragment of unroofed membrane was 566 exposed three times to the same test concentration of nucleotide. Spectra were acquired every 567 three seconds. These technical replicates were averaged and presented as a single experiment.

568
Bleaching was corrected by fitting the ANAP intensity of the last ten spectra acquired during each 569 nucleotide-free solution wash to equation 1.

571
The tip of the patch pipette was centred on the slit of the spectrometer immediately after patch 572 excision. Currents were measured as described above. Fluorescence emission spectra from the 573 excised patch were acquired concurrently with current measurements, both during test solution 574 application as well as nucleotide-free solution. Background subtraction was slightly imperfect due 575 to the exclusion of TNP-ATP from volume of the glass of the pipette, resulting in spectra that have 576 negative intensities at the TNP-ATP peak at high nucleotide concentrations. However, this over-577 subtraction does not affect the size of the ANAP peak, which we used to quantify nucleotide bind-  Two variations on the full MWC model were also considered, and diagrammatic formulations are shown in Figure 2 -Figure supplement 1 so that all priors are flat apart from L, which is weakly informative with 99% of its density falling 612 between unliganded open probabilities of 0.01 and 0.99, and 85% falling between 0.1 and 0.9.

613
Each model was run with 4 independent chains for 10,000 iterations each after a burn-in period 614 of 20,000 iterations, saving every 10th sample for a total of 4,000 samples per model.  (Gronau et al., 2017), and leave-one-out cross-validation (LOO-CV) was performed using the loo 619 package (Vehtari et al., 2017). 620 Docking.

621
Computational docking of TNP-ATP into the nucleotide binding site of Kir6.2 was performed using   Table 3. Table 3: Fitted parameters for the MWC-type models. , and their associated quantiles are reported as log 10 values to maintain consistency of the accuracy they are reported at.      Pairwise correlation plots of , and from the full MWC-type model fit to Kir6.2*-GFP co-expressed with wild-type SUR1, SUR1-K205A, and SUR1-K205E. B. Pairwise correlation plots of and from the full MWC-type as above with fixed to 0.8 (Trapp et al., 1998). for the other three constructs). As the two fits were very similar, the dashed curve mostly overlays the solid curve. The most notable differences between the fits are that the negative cooperativity model allows for non-sigmoidal curves, and the single-binding model predicts much larger pedestals of current at saturating concentrations of TNP-ATP than either of the other two models.