Cryo-EM structure of lipid embedded human ABCA 7 at 3 . 6 Å resolution 1 2

Dysfunction of the ATP Binding Cassette (ABC) transporter ABCA7 alters cellular lipid homeostasis and is linked to Alzheimer’s Disease (AD) pathogenesis through poorly understood mechanisms. Here we determined the cryo-electron microscopy (cryo-EM) structure of human ABCA7 in a lipid environment at 3.6Å resolution that reveals an open conformation, despite bound nucleotides, and bilayer lipids traversing the transmembrane domain (TMD). We show that ATP hydrolysis in ABCA7 is modulated by its lipid environment as well as apolipoprotein (apo) A1 and apoE, the latter in an isoform dependent manner, and that apoA1 can directly bind ABCA7. Structural similarities between ABCA and ABCG family transporters suggest that TMD-nucleotide binding domain (NBD) pairs in both sets move as single rigid bodies to affect conformational transitions. Our data suggest these transitions in ABCA7 are influenced by TMD cavity lipids and apolipoprotein binding in addition to being coupled to ATP hydrolysis.


23
Dysfunction of the ATP Binding Cassette (ABC) transporter ABCA7 alters cellular lipid 24 homeostasis and is linked to Alzheimer's Disease (AD) pathogenesis through poorly understood 25 mechanisms. Here we determined the cryo-electron microscopy (cryo-EM) structure of human 26 ABCA7 in a lipid environment at 3.6Å resolution that reveals an open conformation, despite bound 27 nucleotides, and bilayer lipids traversing the transmembrane domain (TMD). We show that ATP 28 hydrolysis in ABCA7 is modulated by its lipid environment as well as apolipoprotein (apo) A1 29 and apoE, the latter in an isoform dependent manner, and that apoA1 can directly bind ABCA7. 30 Structural similarities between ABCA and ABCG family transporters suggest that TMD-31 nucleotide binding domain (NBD) pairs in both sets move as single rigid bodies to affect 32 conformational transitions. Our data suggest these transitions in ABCA7 are influenced by TMD 33 cavity lipids and apolipoprotein binding in addition to being coupled to ATP hydrolysis.

46
AD accounts for the overwhelming majority of dementias, affecting more than 5.5 million people 47 in the United States alone. In the absence of a cure, this number is expected to increase rapidly, 48 with devastating healthcare and socioeconomic consequences (Alzheimer's Association, 2020). 49 Genome wide association studies have identified ABCA7 as an important genetic risk factor for similarities with the C. elegans cell corpse engulfment protein (ced-7: 24% identical to ABCA7), 79 which regulates phagocytosis during programmed cell death (Wu and Horvitz, 1998

101
Phospholipids and apolipoproteins modulate ABCA7 in vitro activity 102 We expressed human ABCA7 harboring a C-terminal yellow fluorescent protein (YFP)-rho-ID4 (Ni-NTA) resin in the presence of detergent solubilized ABCA7. As seen in Figure 1D, ABCA7 137 is retained and subsequently co-elutes with apoA1, providing qualitative confirmation of direct 138 complex formation that we speculate holds true for apoE isoforms as well.

139
Cryo-EM structure of ABCA7 reveals lipid filled TMD cavity 140 The cryo-EM structure of nanodisc reconstituted ABCA7 was determined to 3.6Å resolution in L655 and T1646 from TM5 and TM11, respectively ( Figure 2B). TM1, TM2, and TM5 from 155 TMD1 contribute the majority of residues within 5 Å of the observed phospholipids compared to 156 only TM11 from TMD2 ( Figure 2C). This asymmetric arrangement with a smaller TMD2 interface arises from a rigid body shift of the TMD2-NBD2 pair of ABCA7 compared to the more symmetric 158 arrangement seen in the ABCA1 structure, as expanded upon below.

Details of TMD-ECD and lipid facing interfaces of ABCA7
160 What leads to the observed asymmetry in TMD arrangement in ABCA7 compared to apo ABCA1 161 and ABCA4? As seen in Figure 3, the electrostatic potential maps of cavity facing interfaces in 162 ABCA7 reveal marked differences between the two TMDs. The TMD1 interface is significantly 163 more electropositive than that of TMD2 ( Figure 3A  Our ABCA7 structure allows us to map known missense ABCA7 variants and also 180 compare known ABCA1 missense variants to analyze conservation in ABCA7. As shown in 181 Figure 4, many pathogenic missense ABCA7 mutants associated with higher AD risk, as well as 182 three 'protective' missense variants reported to reduce AD risk are found distributed over the entire 183 ABCA7 structure. Many of these overlap with residues comprising the TMD-ECD interfaces.  to be a temporary space to sequester phospholipids in ABCA1 or serve as a passage for 201 phospholipid transport to apolipoproteins, a postulate supported by our observations in ABCA7.

202
The observed ECD lipids are predominantly located within ECD1, which contributes the majority 203 of residues within 5Å of them.

204
Conserved rigid body motions define ABCA7 conformational transitions 205 Differences in local EM density quality in our ABCA7 maps ( Figure S2) suggest increased motion 206 in TMD2 and NBD2, while TMD1-NBD1 and the ECD appear to be more rigid. This is further 207 evidenced by a comparison of our ABCA7 structure with that of ABCA1. As shown in the 208 superposition in Figure figure 6D). The binding of 285 apolipoproteins to ABCA7 may rearrange the ECD and its TMD interactions in a way that 286 promotes TMD2-NBD2 to collapse into TMD1-NBD1 in a closed state (red arrows in figure 6D).

287
This would necessitate the exit of any TMD cavity resident lipids that could occur by the extraction 288 of a subset towards the ECD and/or laterally back into the surrounding lipid environment.  For the pulldown assay, 132 µg of 3C protease cleaved ABCA7 was mixed with pure his tagged 359 apoA1 using a 1:10 molar ratio in a buffer comprising 25 mM Hepes pH 7.5, 150 mM NaCl,   Cryo-electron microscopy data collection and processing. 405 Grids were clipped as per manufacturer guidelines and cryo-EM data was collected using a Titan 406 Krios electron microscope operating at 300kV and equipped with a Falcon 3EC direct electron 407 detector (Thermo Fisher Scientific.). Automated data collection was carried out using EPU 408 (Thermo Fisher Scientific) over multiple sessions in counting mode at a nominal magnification of 409 96,000x, corresponding to a calibrated pixel size of 0.895 Å. Image stacks comprising 60 frames 410 were collected at a defocus range of -0.6 to -2.6 µm and estimated dose rate of 1 electron/ Å 2 /frame 411 and further processed in Relion-3.1 (beta). Motion correction was done using Motioncor2 (Relion 412 implementation) (Zheng et al., 2017) and contrast transfer function (CTF) correction was 413 performed using Gctf (Zhang, 2016). A summary of the overall data processing scheme is 414 presented in Supplementary Figure S1C-E. In brief, 11802 micrographs were used for template 415 free picking of 6725108 particles, followed by particle extraction at a 3x binned pixel size of 2.685 416 Å/pix. The dataset was processed in two batches. After 2-3 rounds of 2D classification 1259324 417 particles from Set 1 and 1088487 particles from Set 2 were selected for independent 3D 418 classification steps (number of classes (K)=8 for both). The structure of human ABCA1 419 (EMDB6724) was used as a 3D reference for an initial 3D classification of a subset of the total 420 data to yield an initial sub-nanometer resolution map of ABCA7 that was used as a 3D reference 421 for the full datasets. After 1 round of 3D classification, both sets of data yielded a similar ensemble 422 of classes. A total of 113291 particles from similar looking classes (black boxes) were subjected 423 to an additional round of classification (K=3), ~80% of which fell into a high-resolution class that 424 yielded a 3.6 Å map after refinement and particle polishing steps. Similarly, 124114 particles from 425 a second set of two similar classes (red boxes in Figure S1D) were selected for subsequent 426 refinement, particle polishing, and post processing to yield a 3.2 Å map. All resolution estimates