X-ray diffraction studies on merohedrally twinned Δ1–62NtNBCe1-A crystals of the sodium/bicarbonate cotransporter

A truncated mutant missing the first 62 residues of the N-terminal, cytoplasmic domain of the sodium-bicarbonate NBCe1-A cotransporter crystallizes in space group P31 with pseudo-P3121 symmetry and a hemihedral twin fraction of 33.0%. Twinned fractions and twin-pair statistics over binned resolutions confirm that the calculated twin fraction is associated with hemihedral twinning and not to non-crystallographic symmetry.


Introduction
Bicarbonate (HCO 3 À ) transport in the kidney is critical for acid-base balance in the body. Every day, normal kidneys filter and reabsorb approximately 4 mol (250 g) of HCO 3 À . The proximal tubule is the site of the kidney that reabsorbs 80% of the filtered HCO 3 À load. Impaired bicarbonate reabsorption in the proximal tubule, diagnosed in patients as proximal renal tubular acidosis (pRTA or type-II RTA), is characterized by hyperchloremic metabolic acidosis with variable hypokalemia (Haque et al., 2012). Na + -coupled HCO 3 À transporters (NCBTs) are integral membrane proteins that are responsible for handling (either reabsorbing or secreting) Na + and HCO 3 À ions in tissues throughout the body. NBCe1-A is the major electrogenic NCBT transporter found at the basolateral membrane of the proximal tubule that mediates the crucial step in the transepithelial movement of HCO 3 À ions, thus maintaining blood pH. Studies in patients with isolated type-II RTA have led to the identification of 11 natural occurring mutations in NBCe1-A (Igarashi et al., 1999), two of which are Q29X and R298S in the cytoplasmic N-terminal domain (NtNBCe1-A). Most patients with NtNBCe1-A mutations have type-II RTA and in addition exhibit systemic manifestations including mental retardation, growth retardation and myriad ocular defects, e.g. band keratopathy, glaucoma and cataracts (Suzuki et al., 2012). These findings underlie the need to understand how NtNBCe1 may function in other tissues such as the corneal epithelium, lens epithelium and ocular ciliary epithelium, where it has recently been identified (Suzuki et al., 2012), and how mutations cause disease.
Protein crystals of the P3 subset of trigonal space groups are commonly associated with merohedral twinning (Yeates & Tsai, 2011;Yeates, 1997;Sawaya, 2007;Berthold et al., 2006). They are the most complicated case of twinning because not one but three sets of twin operators are possible, giving rise to apparent crystal point groups 321, 312 or 6 (Yeates & Tsai, 2011;Sawaya, 2007). Many programs to identify twinning have difficulty in distinguishing noncrystallographic symmetry (NCS) from merohedral twinning. Thus, twin fractions tend to be overestimated (Yeates & Tsai, 2011). In order to identify and calculate a conservative twin fraction, a partial twinning test for twodomain merohedral twinning (referred to as hemihedral twinning) can be implemented that evaluates the variable H (i.e. the difference between intensities of twin pairs divided by their sum) as a function of resolution (Yeates & Tsai, 2011) and this service is available at the NIH-UCLA twin server (Yeates & Fam, 1999). To complicate matters even more, NCS operators often also tend to coincide with crystallographic symmetry operators, resulting in space-group ambiguity (Sawaya, 2007). In the previous study of full-length NtNBCe1-A crystals that diffracted X-rays to 3.0 Å resolution (Gill & Boron, 2006a), the self-rotation function at the = 180 section demonstrated dyads in the ab* plane spaced approximately every 60 along starting at (30, 0 ), giving rise to space-group ambiguity between P3 1 21 and P3 1 with pseudo-P3 1 21 symmetry. Assuming that the dyads were not the expected crystallographic peaks of the P3 1 21 space group, they confirmed the presence of a twofold NCS element expected for a dimeric NtNBCe1-A as estimated in Gill & Boron (2006b) and accurately measured in Gill (2012). In this study, we characterize crystals of Á1-62NtNBCe1-A that diffract X-rays to 2.4 Å resolution, despite the detection of partial merohedral twinning, anisotropy and persistent space-group ambiguity.

Expression, purification and crystallizations
The expression and purification of full-length Nt (residues 1-362) have previously been described in Gill & Boron (2006b) and subsequently refined in Gill (2012). Crystal conditions have previously been described in Gill & Boron (2006a). To generate better X-ray diffracting crystals, a truncated mutant lacking the first 62 residues (Á1-62NtNBCe1-A) was recombinantly expressed and purified using similar procedures and conditions as the full-length Nt. Á1-62NtNBCe1-A was crystallized by hanging-drop vapor-diffusion methods (McPherson et al., 1995). Each drop had a total volume of 4 ml and consisted of equal parts of well solution and a 30 mg ml À1 NtNBCe1 stock solution. The volume of the well was 0.5 ml. The mother solution in the well consisted of 35%(v/v) saturated ammonium sulfate solution in 150 mM sodium citrate pH 6.5, similar to that used for the full-length Nt (Gill & Boron, 2006a), yielding wedgeshaped crystals as shown in Fig. 1.

X-ray data collection and processing
The procedures for data collection from Nt crystals have been described previously (Gill & Boron, 2006a). For Á1-62NtNBCe1-A crystals, data were collected on the X29 beamline using a Quantum 3 Â 3 detector (Area Systems Detector Corporation, San Diego, California, USA) at the National Synchrotron Light Source at Brookhaven National Laboratory, Long Island, New York, USA. At the synchrotron, a Á1-62NtNBCe1-A crystal was flash-cooled by swiping the crystal in bicycle oil (Tri-Flow superior lubricant with Teflon; Sherwin-Williams Consumer Group, Cleveland, Ohio, USA) and immediately placing the crystal in a cryostream. Data were collected initially to 2.6 Å resolution and then subsequently to 2.4 Å resolution after annealing the crystal a few times (i.e. thawing the flash-cooled crystal in mother liquor for $3 min and then again flashcooling in the cryostream). The crystal was oscillated through 60 (assuming a P3 space group) and data collection then continued with an oscillation angle of 0.8 . Each frame was exposed for 3 s. The crystal-to-detector distance was set to 325 mm. Data were processed using the program HKL-2000 (Otwinowski et al., 2003). Merohedral twinning was validated by calculating the variable H as a function of resolution using the NIH twinning server (http://nihserver.mbi. ucla.edu/Twinning; Yeates & Fam, 1999). Anisotropic corrections were applied to the Nt and the Á1-62NtNBCe1-A data sets using the server at http://services.mbi.ucla.edu/anisoscale (Strong et al., 2006). After anisotropic scaling, isotropic B factors of À27.24 Å 2 for the previously collected Nt data set and À20.94 Å 2 for the data set from Á1-62NtNBCe1-A were applied to restore the magnitude of the high-resolution reflections as described in Strong et al. (2006).

Crystal parameters
Data collected to 3.0 Å resolution from the full-length Nt crystal have been characterized in Gill & Boron (2006a). These reported NtNBCe1-A molecules crystallized in a trigonal space group, yielding equal possibilities of space group P3 1 or P3 1 21. For either space group, the crystal had unit-cell parameters a = b = 51.7, c = 200.6 Å , where the smaller dimensions roughly reflect the diameter of the dimer near neutral pH. Using a monomer of the N-terminal domain of family member AE1 (NtAE1; Zhang et al., 2000) as a probe in molecular-replacement methods, the crystal-packing solution in either space group yielded an arm-swapped dimer in character with NtAE1. For the P3 1 possibility, the entire dimer is contained in the asymmetric unit. For P3 1 21, the twofold symmetry of the dimer coincides with a crystallographic axis. The space group was to be sorted out by model refinement. However, this ultimately was not possible owing to a poor data-to-parameter ratio. Subsequently, the annealed Á1-62NtNBCe1-A crystal diffracted X-rays to 2.4 Å resolution. Data reduction indicated the same apparent space-group ambiguities as the full-length Nt except that the unit-cell parameters were a = b = 54.0, c = 398.4 Å , approximately doubled along the c axis. In a P3 space group, a calculated Matthews coefficient (V M ; Kantardjieff & Rupp, 2003;Matthews, 1968) of 2.39 Å 3 Da À1 (monomer molecular weight 35 073 Da) indicates four monomer molecules in the asymmetric unit; in a P321 space group, the V M indicates two monomer molecules in the asymmetric unit.

Resolution and self-rotation functions
Annealing the Á1-62NtNBCe1-A crystal apparently improved the X-ray diffraction from 2.6 to 2.4 Å resolution. Data re-collection and reduction suggest that both versions have the same possible space groups within the trigonal set and with similar unit-cell parameters. The = 180 section of the self-rotation functions after annealing reveals that the highest set of dyads have a 5 offset from the crystallographic b* axis, indicating they arise from NCS. With two dimers calculated to be in the asymmetric unit from a necessary P3, P3 1 or P3 2 space group, the = 180 section suggests two overlapping twofold elements whose NCS axes (presumably from two biological dimers) are parallel to one another.

Merohedral twinning
The Á1-62NtNBCe1-A crystal before annealing had an estimated hemihedral twin fraction of 43.6% with twin operator h; Àh À k; Àl (or the symmetry-related operator k; h; Àl) in reciprocal space. The crystal after annealing was calculated to have a hemihedral twinning fraction of 33.0%, again with twin operator h; Àh À k; Àl. The values for are calculated for the entire resolution range (34.1-2.41 Å ) and are in agreement with the refined calculated using PHENIX (Adams et al., 2010). The data-collection statistics are shown in Table  1. Twin fraction and twin-pair intensity statistic H values over binned resolutions are illustrated in Fig. 2. Here, a conservative resolution limit of 2.6 Å is used in the calculations for the Á1-62NtNBCe1-A crystal. As illustrated, partial twinning is observed as the estimates of remain significantly above zero and hH 2 i tends to be similar out to high resolution (green and red curves). In the case of NCS but no twinning, the similarity between twin pairs (to a degree that depends on ) tends to drop off strongly as a function of resolution, usually by 3 Å (T. O. Yeates, personal communication). This is exemplified by the full-length Nt crystal (Gill & Boron, 2006a). The overall (44.8-3 Å ) twinning fraction appears to be 33.1% in the data set comprising the first 120 of oscillations (increasing to 42.4% with 359 of oscillations). However, looking at the split shells out to higher resolution, the apparent twin fraction is significantly overestimated owing to NCS, which is deduced by the drop off in and the sharp increase in hH 2 i (purple curves).

Anisotropy
Annealing also reduced the anisotropy of the NtNBCe1-A and Á1-62NtNBCe1-A crystals. Full-length NtNBCe1-A crystals initially did not diffract X-rays. After annealing, a crystal diffracted X-rays to 3.0 Å resolution. The collected data had strong anisotropy (ÁB of 40.54 Å 2 ). The data from a Á1-62NtNBCe1-A crystal before annealing also had strong anisotropy (ÁB of 38.82 Å 2 ). The same Á1-62NtNBCe1-A crystal after annealing shows that the anisotropy is mild (ÁB of 21.52 Å 2 ). No truncations were necessary.

Molecular packing
Inspection of systematic absences did not help to reduce the spacegroup possibilities, presumably as a consequence of the twinning. However, molecular-replacement searches using Phaser (McCoy et al., 2007) yielded the highest peaks in P3 1 with a Z-score of 15. Searches in P3 and P3 2 did not yield solutions. The Z-score of searches in all trigonal space groups are shown in Table 2  18.6 (0.9) Twin fraction (%) 33 † R merge = P hkl P i jI i ðhklÞ À hIðhklÞij= P hkl P i I i ðhklÞ, where I i (hkl) is the ith measurement and hI(hkl)i is the mean of all measurements of the intensity of the reflection with Miller indices hkl.

Figure 2
H versus resolution twinning test. The binned twinning-test calculations are shown for the Á1-62NtNBCe1-A crystal and the full-length Nt crystal for twin operator 2 along a, b. As described in the text, the trends demonstrate that the calculated twin fractions are associated with hemihedral twinning and not NCS (green and red curves). In contrast, the binned twinning test for the full-length Nt crystal (purple curves) suggests little to no twinning despite an apparent twin-fraction calculation of 33.1% over the entire unbinned resolution range. Note that the Á1-62NtNBCe1-A crystal before annealing has an apparent higher twin fraction of the crystal compared with that after annealing, either indicating a dramatic physical change in the lattice after annealing or a non-uniform crystal lattice. similar mechanism for dimerization, i.e. two interlocking domains at the C-terminal end. The two arm-swapped dimers have their twofold axes parallel to each other, both aligned with the crystallographic a axis.

Purification and other challenges with the Nt
The easiest way to overcome merohedrally twinned crystals is to find another space group whose point-group symmetry does not lend itself to this sort of twinning or to screen many crystals in the problematic crystal system in the hope of finding a crystal with a single domain. These possibilities did not lend themselves to Á1-62NtNBCe1-A, and only a very limited number of crystals grew to sufficient size for either Nt or Á1-62NtNBCe1-A. Growing reproducible Nt and Á1-62NtNBCe1-A crystals has been challenging for a number of reasons. Isolated molecules often lead to peculiar discrete waves of precipitation (Gill & Boron, 2006b). Precipitation can be overcome by keeping the protein in a taut conformation to prevent specific and nonspecific self-associations, and is achieved by strict adherence to acidic to neutral pH during purification (Gill, 2012). Protein solutions are then stable as intact dimers for a week at room temperature or 277 K and no longer have to be stored by unquantifiable ammonium-sulfate precipitations. For longer-term storage, although a number of additives have been screened, cooling still induces a high polydispersity as judged by dynamic light-scattering measurements. Fresh protein must be purified for crystallization trials beyond these time limits. Moreover, even with a few crystals in hand, most cryoprotectants (traditional oils, PEGs, salts, alcohols) used in low-temperature X-ray data collection induced highly mosaic lattices, a problem fortuitously overcome by brief exposure to commercial bicycle oil, rapid cooling and annealing procedures as outlined here.

Conclusions
We have determined the space group of Á1-62NtNBCe1-A crystals in order to implement viable refinement procedures on Nt crystals containing disease-causing mutations. Calculating twin fractions over small incremental resolution bins was useful to identify that the values arose from actual twinning, unlike those seemingly from fulllength Nt crystals. Excluding the highest resolution shells, which appear to give artificially low values for the twin fraction, possibly owing to the larger measurement errors in this range, the trend demonstrates that the percentage twin fraction has a value in the low 30s after annealing and in the low 40s before annealing, assuming that the change in behavior can be attributed to the crystal treatment and not to variations in the crystal lattice. With structure determination of the Á1-62NtNBCe1-A now possible, the role of the Nt in maintaining the interstitial pH and the effect of N-terminal truncation mutations can be detailed. From our earlier biophysical studies and low-resolution structure of full-length Nt, we predict that truncations at the extreme Nt eliminate part of a gate regulating access into a substrate tunnel that traverses the Nt and enters into the transmembrane domain, also causing loss of pH sensitivity, dimer-dimer self-associations or patch formation in the membrane, regulation of bicarbonate uptake into the Nt and possibly affecting other proteinprotein interactions such as those needed for efficient trafficking to the membrane.