Low-dose cryo-electron ptychography of proteins at sub-nanometer resolution

28 Cryo-transmission electron microscopy (cryo-EM) of frozen hydrated specimens is an efficient 29

for the structural analysis of purified biological molecules . However, for particles 30 smaller than 50kDa, single particle cryo-EM often fails due to the limited imaging contrast of 31 individual micrographs . For dose -resilient samples often studied in the physical sciences , 32 electron ptychography – a coherent diffractive imaging technique using 4D scanning 33 transmission electron microscopy (4D-STEM) – has recently demonstrated resolution down to 34 tens of picometers for thin specimens imaged at room temperature. 35 Here we appl ied ptychographic data analysis to frozen hydrated single protein particles, 36 reaching sub-nanometer resolution 3D reconstructions. We employed low-dose cryo-EM with 37 an aberration -corrected, convergent electron beam to collect 4D -STEM data for our 38 reconstructions. The high speed of the electron detector allowed us to record large datasets 39 of electron diffraction patterns with substantial overlaps between the interaction volumes of 40 adjacent scan position , from which the scattering potentials of the sample s were iteratively 41 reconstructed. The reconstructed micrographs show strong contrast enabling the 42 reconstruction of the structure of apoferritin protein at up to 5.8 Å resolution. Ptychography is 43 a promising tool to study the structure of smaller frozen hydrated protein particles, and, once 44 combined with electron tomography tilt series, bears the potential to provide unique insight 45 into the ultrastructure of vitrified biological tissue. 46 47 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 3 Main 48 Cryo-transmission electron microscopy (cryo -EM) has revolutionized life sciences and 49 pharmaceutical research in academia and industry. Proteins as single particles in thin vitrified, 50 aqueous layers can be structurally analyzed at near-atomic resolution by transmission electron 51 microscopy (TEM) in a matter of hours, if the proteins are available in sufficiently large 52 quantities and homogeneity, and if the protein particles have a molecular weight larger than 53 50 kDa 1. If these conditions are not met, cryo -EM often fails. Cryo-EM analysis of proteins 54 suffers from the low signal -to-noise ratio (SNR) in the images because proteins are weak-55 phase objects and highly fragile under the electron beam. Typically, the electron fluence must 56 be limited to below 20 e –/Å2, if a protein is to be imaged at high resolution. Data can be 57 recorded with slightly higher doses if fractionation of the electron dose is used, and recorded 58 frames are resolution weighted with dose -dependent frequency filters before averaging, to 59 obtain higher-contrast images 2. Cryo-EM furthermore suffers from particle movements under 60 the electron beam, which can be corrected partly during image processing through motion 61 correction in dose -fractionated "movies" 3–5. Finally, conventional cryo-EM suffers from the 62 oscillating contrast transfer function (CTF), which is dampened towards higher resolution from 63 the limited spatial and temporal coherence of the beam, the detector modulation transfer 64 function (MTF), and non-corrected specimen movements, among others. Phase plates, such 65 as Volta or laser phase plates 6,7 partly improve the CTF for low -resolution components, but 66 so far have not yet shown improvements in final resolution. 67 An alternative data acquisition scheme is scanning transmission electron microscopy (STEM), 68 which uses a focused electron probe that is scanned across the sample, while electron 69 detectors record the number of electrons scattered to a certain angle covered by the detector 70 pixel(s) as a function of the probe position 8. Bright-field (BF) STEM images provide only weak 71 phase contrast 9, and dark-field (DF) STEM yields high mass -thickness contrast, yet at low 72 dose efficiency, so that STEM was until recently of limited use in the life sciences . STEM Z-73 contrast imaging, employing annular detectors covering high scattering angles , is based on 74 Rutherford scattering, where the detector signal is approximately proportional to the square of 75 the atomic number and linear to the number of atoms within the interaction volume. Th is 76 linearity of the high -angle annular DF (HAADF) STEM signal has been exploited for mass 77 measurements of protein particles that had been freeze -dried on ultra -thin carbon film 78 supports 10,11. Early attempts of high -resolution low-dose aberration-corrected HAADF STEM 79 imaging showed only amplitude contrast and was found unsuitable for life sciences imaging 80 12. However, cryo-STEM tomography, combining BF and DF STEM with sample tilt series to 81 compute a 3D reconstruction of the vitrified specimen 13, has shown promising results for 82 thicker biological specimens up to 1 µm diameter. More recently, integrated differential phase-83 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 4 contrast (iDPC) STEM was applied to cryo -EM specimens, reaching 3.5 Å resolution for 84 protein 3D reconstructions from data recorded with a quadrant-detector 8,14. 85 Beyond monolithic or segmented detectors, the use of pixelated detectors in STEM allows to 86 record a 2D image of the diffracted probe for each 2D probe position, enabling advanced 87 analysis in this so-called 4D-STEM method 15. For example, electron dose-efficient recovery 88 of the specimen's scattering phase can be achieved by ptychographic processing of the 89 recorded diffraction patterns 16. For materials sciences specimens, this method has recently 90 established resolution records in the picometer range 17,18. Besides the inverse multi -slice 91 schemes used in that work, which consider multiple scattering events during reconstruction 92 but require comparably high electron doses, several other algorithms exist that employ the 93 single-scattering projection assumption. The latter can be divided into direct methods, such 94 as single -sideband ptychography 19,20, Wigner distribution deconvolution 21,22, or parallax 95 reconstructions from tilt-corrected BF STEM data 23–25, to provide phase information with real-96 time capabilities, or iterative methods, such as the computationally more demanding proximal 97 gradient algorithms (e.g., difference map 26 and ePIE 27). 98 4D-STEM with an aberration -free electron beam that is focused onto the specimen typically 99 scans the sample with a narrow Airy disk profile , limited to only a short vertical extension, 100 depending on the convergence semi -angle (CSA) used to form the probe 8. 4D-STEM can 101 also be performed with a strongly defocused electron probe (Figure 1), so that the convergent 102 electron beam illuminates the specimen with a disk large enough to expose an entire typical 103 single-particle protein in the vitrified cryo -EM specimen as one single scattering experiment. 104 The probe can then be scanned across the specimen surface at a step size in the nm range, 105 so that diffraction patterns are recorded from probe positions that share significant overlap in 106 the (x,y) sample plane . This overlap enables phasing of the recorded data for structure 107 reconstruction, and it eliminates the twin-image problem or mirrored reconstructions. 108 4D-STEM has recently gained significant momentum from the availability of ultrafast pixelated 109 electron detectors. Extending 4D -STEM to study beam -sensitive biological cryo -EM 110 specimens, however, faces several experimental and computational hurdles. For low-dose 111 experiments, scan strategies must be optimized for most efficient handling of specimen motion 112 and specimen charging. Data processing workflows have to be established that can deal with 113 the very low signal-to-noise ratio in the low-dose datasets and with the enormous quantity of 114 recorded data, reaching terabytes in size for one sample. Nevertheless, significant progress 115 in life sciences 4D -STEM has recently been demonstrated by the application of 4D -STEM 116 ptychography to frozen-hydrated viruses 16,28. 117 Here, we performed low-dose 4D-STEM studies with a defocused beam and ptychograph ic 118 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 5 data processing of frozen -hydrated single particle protein samples ( Fig. 1) and show that 119 structures of proteins at sub-nm resolution can be achieved from less particles than usually 120 imaged in cryo -EM. We implement cryo -electron ptychography with a n aberration corrected 121 cryo-EM instrument and fast hybrid pixel detector, and we discuss the current state of 122 automated data collection and data processing. 123 124 125 Fig. 1: Experimental setup of 4D-STEM. a, A vitrified life sciences sample is illuminated with a convergent electron beam. Electrons are recorded using a pixelated detector. The convergence semi angle (CSA) is in the range of a few mrad. The beam is focused to a plane above or below sample with a defocus in the micrometer range. b, An example ptychographic reconstruction of apoferritin . Scale bar: 50 nm. c, The sum of three 4D-STEM diffraction patterns. A typical diffraction pattern shows most of the electrons in the bright field (BF) disk of the radius corresponding to the CSA of 6.1 mrad, and some electrons are scattered to higher angles to the dark field (DF) region. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 6 Ptychography of cryo-EM samples 126 Ptychographic data were recorded with a probe-corrected electron microscope with a CSA of 127 4 to 6 mrad and a defocus of ~2 µm, resulting in a beam diameter on the sample of ~16 nm 128 for the 4 mrad CSA. Data were collected with a beam dwell time of 250 µs, and a STEM step 129 size of 2 nm, corresponding to 84% beam overlap for two adjacent circular probes. Using a 130 fluence of ~0.5 e -/Å2 per probe position, an average total electron fluence of ~35 e -/Å2 was 131 applied to the sample during one scan of 128x128 positions . The low electron dose and 132 parameters used in our experiments meant that on average less than 20% of the detector 133 pixels received an electron for one probe position , making the recorded diffraction patterns 134 very noisy. 135 A dataset of gold nanoparticles on a carbon film was collected before beginning experiments 136 to calibrate rotation angles and CSA, and a diffraction pattern with the specimen removed from 137 the beam path was recorded to serve as estimation for the probe -forming aperture in the 138 iterative reconstruction algorithm. Data collection was then performed with a setup as shown 139 in Fig. 2. The software SerialEM 29 was used to control the Titan Krios instrument, take real -140 space TEM images of the grid, and compile a table of X,Y coordinates, indicating the specimen 141 locations that have ice holes with suitable ice thickness and a high particle density. After this, 142 the instrument was switched to STEM mode with a convergent beam, and a custom Python 143 script was used for automated 4D -STEM data collection. Data collection was performed at 144 approximately 144 specimen locations per hour , each comprising a 128x128 position 4D -145 STEM scan, and resulted in approximately 310 GB of raw data per hour before compression. 146 Data collection was monitored in real-time by computing a parallax (or tilt-corrected BF-STEM) 147 reconstruction with the py4DSTEM software 30. This enabled fast, yet reliable, phase 148 reconstructions to guide acquisition, and provided the probe estimate necessary for the more 149 computationally-intensive ptychographi c reconstructions. Higher-resolution ptychography 150 reconstructions of the scattering potentials were calculated offline with py4DSTEM on a 4-151 GPU computer, obtaining up to 70 ptychographic reconstructions per hour. 152 153 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 7 154 155 Structural analysis of proteins 156 The cryo -electron ptychography pipeline was applied to apoferritin (AF) protein, the phi92 157 bacteriophage, and the tobacco mosaic virus (TMV). To test the quality of the AF protein and 158 its suitability for high -resolution data collection, apoferritin was first imaged by conventional 159 TEM (CTEM), before being studied using the 4D-STEM pipeline. 160 Conventional Cryo-EM imaging of Apoferritin 161 This spherical apoferritin protein particle of octahedral 24-fold symmetry has a diameter of 162 approx. 12 nm and a particle weight of 485 kDA. It shows a highly robust structural integrity 163 when prepared for cryo -EM imaging, making it the ideal test specimen for high -resolution 164 assays 31,32. Conventional TEM (CTEM), using dose -fractionated cryo-EM on a Titan Krios 165 equipped with a 300kV cold-FEG electron source and a Falcon4i direct electron detector was 166 Fig. 2: The ptychography data collection workflow. A python client controls the electron microscope via SerialEM and coordinates the TVIPS USG scan generator, the Dectris ELA camera recording, and the multi-GPU processing of recorded diffraction patterns. Dotted line: The parallax reconstruction is used to provide on- the-fly feedback about the data collection and serve as aberration estimate. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 8 used to record 9,846 dose-fractionated real-space images, from which 1'899'133 AF particles 167 were automatically selected and analyzed with the software cryoSPARC 33, resulting in a 3D 168 map from 411'705 particles at a resolution of 1.09Å ( Extended Data Figs. 2 and 3 ). Data 169 showed a ResLog B-factor of 28 Å2, indicating that the AF sample and the obtained cryo-EM 170 grid preparations were of outstanding quality and not limiting resolution for a structural analysis 171 (Supplementary Table 1). 172 Cryo-electron ptychography of Apoferritin, Phi92 bacteriophage, and TMV 173 4D-STEM was then used on the same AF specimen, but with a different, Cs probe corrected 174 Titan Krios, recording data with a CSA of 4 mrad. Online parallax reconstructions computed 175 during data collection provided instant feedback on the quality of the data collection. Offline 176 computed higher-resolution reconstructions were done by ptychography computation of the 177 sample scattering potentials, which are obtained when assuming an amplitude function of unity 178 and computing the resulting image only based on the reconstructed phase information 23. 179 Obtained 2D potential images were subjected to automated particle picking, resulting in 13’655 180 particles. Particle classification, alignment and 3D reconstruction using cryoSPARC 33 resulted 181 in a 3D map of the protein from 11’552 particles at a resolution of 5.8 Å ( Fig. 3a, Extended 182 Data Figs. 4 and 7, and Supplementary Table 1). 183 4D-STEM was also used to study the Phi92 bacteriophage, using a CSA of 5.1 mrad. Data 184 processing, using helical reconstruction for the bacteriophage's sheath was conducted using 185 cryoSPARC 33, while the capsid reconstruction was performed using RELION 34,35, resulting in 186 a reconstruction of the Phi92 sheath from 1’600 particles at 8.4 Å resolution, and of the Phi92 187 capsid from 182 particles at 12.1 Å resolution (Fig. 3b, Extended Data Figs. 5 and 7, and 188 Supplementary Table 1). 189 4D STEM was also applied to study the tobacco mosaic virus (TMV), using a CSA of 6.1 mrad. 190 Processing of the ptychography potential maps with RELION 34 resulted in a reconstruction 191 from 2’120 particles at 6.4 Å resolution ( Fig. 3c, Extended Data Figs. 6 and 7, and 192 Supplementary Table 1). 193 194 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 9 195 196 197 Fig. 3: Cryo-electron ptychography reconstructions of proteins. a, Apoferritin. b, Bacteriophage Phi92 tail protein. c, Tobacco mosaic virus (TMV). From left to right: Ptychographic image reconstruction (Scale bars are 30 nm); The same image Fourier transformed, showing circular amplitude oscillations (Scale bars are 0.5 nm-1); The 3D reconstruction of the protein; A cross-section of the 3D reconstruction for a and a bottom view of the 3D reconstructions for b and c. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 10

198 Semi-automated 4D-STEM was implemented as a cryo -electron ptychography pipeline and 199 applied to several biological test specimens. For reconstructing micrographs from recorded 200 datasets, the single -slice ptychography reconstruction implemented in the py4DSTEM 201 software was used. This algorithm refines the estimate for the Probe and the Object from the 202 measured intensities in the recorded diffraction patterns as amplitude constraints in Fourier 203 space, and by exploiting the knowledge about the position overlap in real space. The Probe 204 and Object are iteratively refined, using a stochastic gradient descent algorithm. In the 205 initiation of the projection algorithm, an empirically determined hyper-parameter tuning is also 206 integrated, which improves convergence 23. For details, see the Online Methods. 207 For CTEM imaging, the contrast transfer function (CTF) of the instrument is well characterized 208 and its correction provides access to the highest resolution, including correction of higher-209 order aberrations 35. Leidl et al. (2023) recently demonstrated that for thicker biological 210 specimens, multiple electron scattering needs to be taken into account if structural data 211 significantly beyond 1.2 Å resolution were to be achieved 36. 212 For 4D -STEM ptychography, the CTF of the iterative, gradient -based, regularized, 213 ptychography in the presence of noise as employed here , is not established . The CTF is 214 dependent on the 4D-STEM reconstruction algorithm as-well as acquisition parameters used. 215 Computed Fourier transforms of the 2D ptychography potential maps obtained in this work 216 show strong, ring -like amplitude oscillations for the background noise (Fig. 3). Our protein 217 reconstructions corroborate the assumption that the employed 4D STEM data collection 218 scheme and ptychography reconstruction algorithm transfers the specimen information to the 219 reconstructed potential maps via a purely positive transfer function without zero crossings, but 220 with varying amplitude, characterized by a contrast that is low for lowest spatial frequencies 221 terms, rising to a broad maximum at medium resolution, and falling off with amplitude 222 oscillations towards higher resolution 16,28,37. Here, we ignored the effect of any CTF during 223 image processing. Further work in analyzing, understanding and correcting the contrast in 224 cryo-electron ptychography will be required. 225 Electron ptychography has established resolution records for materials sciences specimens, 226 while the algorithms for signal reconstruction and our understanding of the mechanism of 227 contrast formation are still under active development. For life sciences samples, cryo-electron 228 ptychography promises to also give access to the highest resolution and contrast. However, 229 significant technical challenges remain that should be addressed to further improve the 230 resolution of the method. These include the implementation of dose fractionation during data 231 collection, which would further reduce the signal -to-noise ratio of the collected diffraction 232 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 11 pattern; reconstructed sub -frames should then be resolution -weighted and averaged, as 233 commonly done in real-space cryo-EM 2,3. Further required optimizations concern the level of 234 scan precision (noise in the scan coordinates), data collection speed (camera speed and data 235 transfer speed), and post -processing algorithms under the very low dose constraints for life 236 sciences specimens. 237 The frozen hydrated preparations of the proteins apoferritin, phi92 phage, and TMV studied 238 here by 4D -STEM and ptychography resulted in 3D reconstructions at sub -nm resolutions, 239 obtained from very few imaged particles. The ResLog B -factor analysis of the recorded data 240 (Online Methods and Supplementary Table 1) suggests that a larger 4D -STEM dataset of 241 the apoferritin protein might be able to reach 3.1 Å resolution. Nevertheless, this resolution 242 would still be worse than what is achievable by conventional cryo -EM. However, here we 243 applied 4D-STEM with a CSA of only 4 mrad to increase specimen height variation tolerance 244 and data collection efficiency ; a higher CSA would increase resolution. We also applied 4D -245 STEM without correcting for the resolution -limiting effects of specimen beam damage or 246 beam-induced particle motion, and we did not correct for the CTF of the ptychography method, 247 which would be required to fulfill the assumptions of equal noise contributions across all 248 resolution ranges for the maximum likelihood processing underlying the cryoSPARC and 249 RELION software packages. Dose fractionation and dose -dependent frequency weighting, 250 combined with correction for particle movement and a CTF correction that normalizes the 251 noise profile, would all increase the resolution performance of the 4D-STEM ptychography 252 method. 253 The strong contrast of cryo -electron ptychography evident in the reconstructed images , the 254 low numbers of particles needed to obtain sub-nm reconstructions, and the trustworthiness of 255 the projections obtained from the purely positive CTF suggests that 4D-STEM combined with 256 cryo-electron ptychography tomography bears great potential for 3D reconstructions of 257 biological tissue samples. For cellular samples, where the uniqueness of the sample does not 258 allow the averaging of signal from multiple particles, and where thicker slices of the sample 259 are often of interest, the cryo-electron ptychography is especially promising to gain insight into 260 the structural arrangement of smaller structures in cells and tissue. 261 262 Online content 263 Methods, additional references, Nature Portfolio reporting summaries, source data, 264 supplementary information, acknowledgements, peer review information, details of author 265 contributions and competing interests, and statements of data and code availability are 266 available at https://doi.org/….. 267 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 12

268 269 1. Henderson, R. The potential and limitations of neutrons, electrons and X -rays for microscopy of 270 unstained biological molecules. (1992). 271 2. Grant, T. & Grigorieff, N. Measuring the optimal exposure for single particle cryo-EM using a 2.6 272 A reconstruction of rotavirus VP6. Elife 4, e06980 (2015). 273 3. Brilot, A. F. et al. Beam-induced motion of vitrified specimen on holey carbon film. J. Struct. Biol. 274 177, 630–7 (2012). 275 4. Zhang, K. Gctf: Real-time CTF determination and correction. J. Struct. Biol. 193, 1–12 (2016). 276 5. Zheng, S. Q. et al. MotionCor2: anisotropic correction of beam-induced motion for improved cryo-277 electron microscopy. Nat. Methods 14, 331–332 (2017). 278 6. Danev, R. & Baumeister, W. Cryo-EM single particle analysis with the Volta phase plate. Elife 5, 279 (2016). 280 7. Schwartz, O. et al. Laser phase plate for transmission electron microscopy. Nat. Methods 16, 281 1016–1020 (2019). 282 8. Lazić, I. et al. Single-particle cryo-EM structures from iDPC-STEM at near-atomic resolution. Nat. 283

19, 1126–1136 (2022). 284 9. Evans, J. E. et al. Low-dose aberration corrected cryo-electron microscopy of organic specimens. 285 Ultramicroscopy 108, 1636–44 (2008). 286 10. Müller, S. A. & Engel, A. Mass Measurement in the Scanning Transmission Electron Microscope: 287 A Powerful Tool for Studying Membrane Proteins. J. Struct. Biol. 121, 219–30 (1998). 288 11. Engel, A. Biological application of scanning probe microscopes. Annu. Rev. Biophys. Biophys. 289 Chem. 20, 79–108 (1991). 290 12. Buban, J. P., Ramasse, Q., Gipson, B., Browning, N. D. & Stahlberg, H. High-resolution low-dose 291 scanning transmission electron microscopy. J. Electron Microsc. (Tokyo) 59, 103–12 (2010). 292 13. Wolf, S. G. & Elbaum, M. CryoSTEM tomography in biology. Methods Cell Biol. 152, 197–215 293 (2019). 294 14. Rose, H. H. Future Trends in Aberration -Corrected Electron Microscopy. Philos. Trans. Math. 295 Phys. Eng. Sci. 367, 3809–3823 (2009). 296 15. Ophus, C. Four -Dimensional Scanning Transmission Electron Microscopy (4D -STEM): From 297 Scanning Nanodiffraction to Ptychography and Beyond. Microsc. Microanal. 25, 563–582 (2019). 298 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 13 16. Zhou, L. et al. Low-dose phase retrieval of biological specimens using cryo -electron 299 ptychography. Nat Commun 11, 2773 (2020). 300 17. Diederichs, B., Herdegen, Z., Strauch, A., Filbir, F. & Müller -Caspary, K. Exact inversion of 301 partially coherent dynamical electron scattering for picometric structure retrieval. Nat. Commun. 302 15, 101 (2024). 303 18. Chen, Z. et al. Electron ptychography achieves atomic-resolution limits set by lattice vibrations. 304 Science 372, 826–831 (2021). 305 19. Rodenburg, J. M., McCallum, B. C. & Nellist, P. D. Experimental tests on double -resolution 306 coherent imaging via STEM. Ultramicroscopy 48, 304–314 (1993). 307 20. Strauch, A. et al. Live Processing of Momentum-Resolved STEM Data for First Moment Imaging 308 and Ptychography. Microsc. Microanal. 27, 1078–1092 (2021). 309 21. Rodenburg, J. M. & Bates, R. H. T. The theory of super -resolution electron microscopy via 310 Wigner-distribution deconvolution. Philos. Trans. R. Soc. Lond. Ser. Phys. Eng. Sci. 339, 521–311 553 (1997). 312 22. Bangun, A., Baumeister, P., Clausen, A., Weber, D. & Dunin -Borkowski, R. Wigner Distribution 313 Deconvolution Adaptation for Live Ptychography Reconstruction. Microsc. Microanal. 29, (2023). 314 23. Varnavides, G. et al. Iterative Phase Retrieval Algorithms for Scanning Transmission Electron 315 Microscopy. Preprint at https://doi.org/10.48550/arXiv.2309.05250 (2023). 316 24. Spoth, K. A., Nguyen, K. X., Muller, D. A. & Kourkoutis, L. F. Dose-Efficient Cryo-STEM Imaging 317 of Whole Cells Using the Electron Microscope Pixel Array Detector. Microsc. Microanal. 23, 804–318 805 (2017). 319 25. Yu, Y., Spoth, K., Muller, D. & Kourkoutis, L. Dose-efficient tcBF-STEM imaging with real-space 320 information beyond the scan sampling limit. Microsc. Microanal. 27, 758–760 (2021). 321 26. Elser, V. Phase retrieval by iterated projections. JOSA A 20, 40–55 (2003). 322 27. Maiden, A. M. & Rodenburg, J. M. An improved ptychographical phase retrieval algorithm for 323 diffractive imaging. Ultramicroscopy 109, 1256–1262 (2009). 324 28. O’Leary, C. M. et al. Contrast transfer and noise considerations in focused -probe electron 325 ptychography. Ultramicroscopy 221, 113189 (2021). 326 29. Mastronarde, D. N. SerialEM: A Program for Automated Tilt Series Acquisition on Tecnai 327 Microscopes Using Prediction of Specimen Position. Microsc. Microanal. 9, 1182–1183 (2003). 328 30. Savitzky, B. H. et al. py4DSTEM: A Software Package for Four -Dimensional Scanning 329 Transmission Electron Microscopy Data Analysis. Microsc. Microanal. 27, 712–743 (2021). 330 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 14 31. Yip, K. M., Fischer, N., Paknia, E., Chari, A. & Stark, H. Atomic -resolution protein structure 331 determination by cryo-EM. Nature 587, 157–161 (2020). 332 32. Nakane, T. et al. Single-particle cryo-EM at atomic resolution. Nature 587, 152–156 (2020). 333 33. Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. cryoSPARC: algorithms for rapid 334 unsupervised cryo-EM structure determination. Nat. Methods 14, 290–296 (2017). 335 34. Scheres, S. H. RELION: implementation of a Bayesian approach to cryo -EM structure 336 determination. J. Struct. Biol. 180, 519–30 (2012). 337 35. Zivanov, J., Nakane, T. & Scheres, S. H. W. Estimation of high-order aberrations and anisotropic 338 magnification from cryo-EM data sets in RELION-3.1. IUCrJ 7, 253–267 (2020). 339 36. Leidl, M. L., Sachse, C. & Müller -Caspary, K. Dynamical scattering in ice-embedded proteins in 340 conventional and scanning transmission electron microscopy. IUCrJ 10, 475–486 (2023). 341 37. Yang, H., Pennycook, T. J. & Nellist, P. D. Efficient phase contrast imaging in STEM using a 342 pixelated detector. Part II: optimisation of imaging conditions. Ultramicroscopy 151, 232 –239 343 (2015). 344 38. Danev, R., Yanagisawa, H. & Kikkawa, M. Cryo -Electron Microscopy Methodology: Current 345 Aspects and Future Directions. Trends Biochem. Sci. 44, 837–848 (2019). 346 39. Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot. Acta 347 Crystallogr. D Biol. Crystallogr. 66, 486–501 (2010). 348 40. Yamashita, K., Palmer, C. M., Burnley, T. & Murshudov, G. N. Cryo-EM single-particle structure 349 refinement and map calculation using Servalcat. Acta Crystallogr. Sect. Struct. Biol. 77, 1282–350 1291 (2021). 351 41. Williams, C. J. et al. MolProbity: More and better reference data for improved all -atom structure 352 validation. Protein Sci. 27, 293–315 (2018). 353 42. Rosenthal, P. B. & Henderson, R. Optimal determination of particle orientation, absolute hand, 354 and contrast loss in single-particle electron cryomicroscopy. J. Mol. Biol. 333, 721–45 (2003). 355 356 357 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 15 Online Content 358 359

360 Protein expression and purification 361 For apoferritin, the plasmid encoding the mouse heavy chain apoferritin was graciously 362 donated by Masahide Kikkawa 38. The plasmid was transformed into E. coli BL21 (DE3) cells. A 363 saturated overnight culture was inoculated into 500 mL LB media containing Kanamycin , and 364 grown at 37 ºC and induced with 1 mM IPTG for 3 hours after reaching an OD600 of 0.6. The pellet 365 was harvested at 5000 x g and resuspended in 20 mL Buffer A (300 mM NaCl, 20 mM HEPES 7.5, 366 10 mM EDTA), lysed by sonication and then centrifuged for 30 minutes at 20000 x g at 4 ºC. The 367 supernatant was collected and heated at 70 ºC for 10 minutes in a 50 mL tube on a heatblock until 368 milky. The supernatant was mixed by inversion and heated for 5 more minutes. The mixture was 369 centrifuged for 30 minutes at 20000 x g at 4 ºC. The supernatant (~20mL) was precipitated with 370 6.65 g of ammonium sulfate, stirring at 4 ºC for 10 minutes. This corresponds to ~50 -55% 371 ammonium sulfate saturation at 20 ºC. The mixture was centrifuged for 30 minutes at 20000 x g at 372 4 ºC. The pellet was then resuspended in 2 mL of Buffer A and dialyzed with a 12 -14 kDa 373 membrane in 2 L of Buffer A but containing only 2 mM EDTA. The sample was further purified by 374 size exclusion chromatography and loaded directly onto a Superose 6 10/300 GL column with a 375 mobile phase of 150 mM NaCl, 10 mM HEPES 7.5. Peak fractions that corresponded to 24 -mer-376 sized apoferritin were pooled and reinjected for a second round of size exclusion chromatography. 377 Peak fractions were concentrated to 8 mg/mL, using a 100 kDa concentrator and frozen at -80 ºC. 378 The bacteriophage Phi92 sample, initially obtained by Petr Leiman (Texas University Medical 379 Branch, TX, USA), was a kind gift from Davide Demurtas (EPFL, Lausanne, Switzerland). 380 The tobacco mosaic virus (TMV) sample, initially obtained by Ruben Diaz-Avalos (La Jolla Institute 381 for Immunology, CA, USA), was a kind gift from Philippe Ringler (University of Basel, Switzerland). 382 Real-space cryo-EM data collection and processing 383 3 µL aliquots of the apoferritin at 80 mg/ml were applied onto Ultrafoil gold 1.2/1.3 grids, and 384 plunge frozen in liquid ethane with a Leica GP2 plunger (4 °C, 100% rel. humidity, 30 s waiting 385 time, 3 s blotting time). 386 For conventional (real-space) cryo-EM, data were collected in automated manner using EPU 387 v.3.0 on a cold-FEG fringe-free Thermo Fisher Scientific (TFS) Titan Krios G4 TEM, operating 388 at 300 kV with aberration-free image shift (AFIS), and recording data with a Falcon IV (TFS) 389 electron counting direct detection camera in electron counting mode. The data were collected 390 at a nominal magnification of 250 kx (pixel size is 0.3 Å at the specimen level) and a total dose 391 of 40 (e -/Å2) for each exposure. The data from the 1.5 s exposures were stored as electron 392 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 16 event recording (EER) files. The data from 3 × 3 neighboring ice holes in the carbon films were 393 collected using beam and/or image shifting, while compensating for the additional coma 394 aberration (AFIS). The data were collected with the nominal defocus range of −0.3 to −0.9 µm. 395 With an average throughput of 922 images per hour, a total number of 9'846 movies were 396 collected within one 12-h session. 397 In total, 9'846 movies were imported into cryoSPARC v4.4.1 (eer_upsamp_factor=2, 398 eer_num_fractions=80). Upsampled EER movies were drift-corrected and dose-weighted with 399 Patch Motion Correction (res_max_align=4, output_fcrop_factor=1/2), and the contrast 400 transfer function (CTF) was estimated with Patch CTF Estimation (amp_contrast=0.07, 401 res_max_align=3, df_search_min=1000, df_search_max=20000). 402 After rejection of bad images based on CTF Fit resolution (2 - 5 Å) and Relative Ice Thickness 403 (1 - 1.2) parameters, 8'609 images were used for further processing. Initial particle picking 404 was performed on a subset of 50 images with Blob Picker, using ring blobs of 110 -120 Å 405 diameters, NCC score of 0.3, Power Score of 261–712, and a minimum relative separation 406 distance (diameter) of 0.5. Picked particles were extracted with a box size of 720 pixels, 407 Fourier cropped to 120 pixels (binning 6x6 times) and subjected to the streaming reference -408 free 2D classification. 409 Two best 2D class averages were selected as templates for the Template Picker, using a 410 particle diameter of 126 Å and an angular sampling of 10 degrees. In total, 1'899'133 particles 411 were picked. Picked particles were inspected with Inspect Particle Picks, and the best 412 1'104'665 particles were extracted with a box size of 720 pixels, Fourier cropped to 120 pixels, 413 and subjected to the 3 rounds of reference -free 2D classification and Heterogeneous 414 Refinement. A small subset of the 5'005 particles was used for the ab-initio structure building 415 with 3 classes. 416 The best 544'845 particles were re-extracted with a box size of 800 pixels and further refined 417 with Homogeneous Refinement to the resolution of 1.21 Å (B -factor = 23 Å 2). Custom 418 parameters of Homogeneous Refinement include d the application of octahedral (O) 419 symmetry, per-particle defocus optimization using a Defocus Search Range of 4000 Å, per -420 group CTF parameter optimization with fitting the beam tilt, beam trefoil, spherical aberration, 421 anisotropic magnification, and correction for the curvature of the Ewald sphere. 422 Refined particles were split into 9 optical groups with Exposure Group Utilities using Action 423 “split”. Grouped particles were further refined with Homogeneous Refinement to the resolution 424 of 1.15 Å (B -factor = 19 Å 2). Custom parameters of Homogeneous Refinement include 425 application of octahedral (O) symmetry, an initial low -pass filter of 15 Å, a dynamic mask 426 threshold of 0.3, a dynamic mask near value of 4 and far value of 8, defocus refinement, global 427 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 17 CTF refinement, and correction for the curvature of the Ewald sphere. 428 Refined particles were used as an input to Local Motion correction using a maximum alignment 429 resolution of 2 Å and a B-factor of 20 Å2 during alignment. Overlapping particles were removed 430 with a minimum separation distance of 120 Å. The remaining 411'705 particles were subjected 431 to Reference Based Motion Correction with default settings and further refined with 432 Homogeneous Refinement (defocus refinement, global CTF refinement, and correction for the 433 curvature of the Ewald sphere) to the resolution of 1.09 Å (B-factor = 18 Å2). 434 An atomic model for apoferritin was re -built manually from the previously deposited model 435 (PDB ID 8J5A) with COOT v0.9.6 39, iterated with rounds of refinement with Servalcat 40 and 436 Molprobity v4.5.2 41. After each round of refinement, side-chains were manually inspected and 437 adjusted in COOT with Molprobity. 438 4D-STEM data collection 439 Ptychographic data were recorded with a probe -corrected TFS Titan Krios, equipped with a 440 cold FEG electron source, TVIPS Universal Scan Generator (USG), Falcon 4i camera (TFS), 441 and a Dectris ELA hybrid pixel detector. The microscope was operated in STEM mode with a 442 CSA of 4 to 6 mrad and a defocus of ~2 µm, resulting for the 4 mrad CSA in a beam diameter 443 on the sample of ~16 nm. Data were collected with a beam dwell time of 250 µs, and a STEM 444 step size of 2 nm, corresponding to 84% beam overlap for two adjacent circular probes. Using 445 a fluence of ~0.5 e -/Å2 per probe position, an average total electron fluence of ~35 e -/Å2 was 446 applied to the sample during one scan. To avoid the physical gaps between the segments of 447 the ELA detector, data were recorded on only one of the detector segments of 256x256 pixels. 448 The STEM camera length was set so that the edge of that 256x256 px square was at 6.58 449 mrad. The low dose regime and parameters used in our experiments resulted in less than 20% 450 of the detector pixels , on average, recording an electron at all. Dose fractionation, not 451 employed here, could have been implemented by repeatedly scanning the same specimen 452 location several times with an even lower dose, which would have further reduced the dose 453 per pattern. 454 To improve convergence of the ptychographic reconstruction, especially under low -dose 455 conditions, we optimized the alignment of the STEM probe before data acquisition , using the 456 Probe Corrector S -CORR software by TFS on a standard gold cross -grating sample . 457 Optimization of the aberrations up to twice the CSA that was later used for data collection , 458 here to a semi-angle of 8 mrad, took approximately 10 minutes. We then collected a calibration 459 dataset before beginning experiments to aid the optimization in the py4DSTEM software. 460 Finally, a vacuum image of the probe was recorded in diffraction space to serve as estimate 461 for the probe-forming aperture in the iterative reconstruction algorithm. 462 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 18 For data collection, the software SerialEM 29 was used to control the Titan Krios instrument 463 and record a grid atlas at low magnification in image mode. Suitable grid squares on that atlas 464 were manually selected, and for each of these, SerialEM was used still in conventional TEM 465 mode with parallel illumination, to record grid square maps. These maps were then used to 466 semi-automatically compile a table of X,Y coordinates, indicating the specimen locations 467 where ice holes had suitable ice thickness and a high particle density. After this, the instrument 468 was switched to STEM mode with a convergent beam, and a home -built Python script was 469 used for automated 4D-STEM data collection, performing the following steps: 470 1. Controlling the microscope via SerialEM commands to position the instrument 471 specimen stage. 472 2. Preparing the ELA detector for capturing frames. 473 3. Activating the TVIPS scan generator for acquisition. 474 4. Compressing and saving the recorded diffraction patterns for further analysis. 475 Data collection was performed at approx. 144 specimen locations per hour and resulted in 476 approx. 310 GB of raw data per hour before compression. Data collection could be further 477 accelerated by recording larger ptychography scans at each stage position, and by further 478 accelerating the camera recording and data storage speeds, which all would have increased 479 the amount of data recorded. 480 Data analysis 481 Data collection was monitored in real -time by computing a parallax reconstruction 23 with the 482 py4DSTEM software 30. In addition, a higher resolution ptychography reconstruction was 483 calculated offline with py4DSTEM, with the following steps: 484 1. Data loading and flat-field correction: Electron diffraction patterns from one scan in 485 h5 format were loaded, flat-field normalized, and hot/dead pixel corrected by replacing 486 values of identified non -functional pixels with median values computed from their 487 neighbors. Scan metadata such as scan step size were stored separately. 488 2. Organizing 4D-STEM data: The dataset was then reorganized into a four-dimensional 489 array based on the scanning coordinates. Binning of diffraction patterns was not 490 performed. 491 3. Extraction of key parameters: The py4DSTEM package was then used to compute 492 the average diffraction pattern, identify the brightfield disk, and calculate the reciprocal 493 space pixel size. 494 4. Parallax reconstruction: A parallax reconstruction was computed to give a first 495 reconstruction image and to estimate probe defocus. 496 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 19 5. Running ptychographic reconstruction: Starting from the parallax reconstruction, a 497 ptychographic reconstruction was initiated and iterated until convergence , typically 498 reached in 30 iterations , as monitored by an L2 norm of the difference between 499 modeled and measured diffraction intensities as error metric for single -slice 500 ptychography. The probe function in the ptychography reconstruction was initialized 501 based on provided aberrations (C1 in our case) for the phase, and the known 502 amplitudes from the vacuum measurements of the probe, which were recorded for 503 each data collection session. T he probe's Fourier amplitude was replaced after each 504 iteration with the amplitude from the vacuum probe measurements. The phase of the 505 probe was also regularized after each update step by fitting a low -order surface 506 expansion (up to radial and angular order 3) and removing residuals 23. 507 6. Image storage: Finally, the reconstructed micrograph was cropped if necessary and 508 saved for subsequent analysis. 509 The workload was distributed across four GPUs, each processing a single dataset. This 510 parallel approach allowed the processing of 70 datasets per hour with a 4-GPU computer. 511 4D-STEM ptychography structure determination of proteins 512 For apoferritin , ninety cryo -electron ptychography datasets (i.e., 4D -STEM scans) were 513 acquired using SerialEM. A grid atlas was initially acquired using the Falcon4i camera, from 514 which suitable grid squares were manually identified to compile grid square maps and identify 515 suitable acquisition locations. These locations had an ice thickness between 50 and 100 nm. 516 At these locations, automated ptychography scans with a CSA of 4 mrad were started , using 517 an average total electron fluence of ~35 e-/Å2 per scan. 518 Ptychography reconstructions of the sample potential maps, which are obtained when 519 assuming an amplitude function of unity and computing the resulting image only based on the 520 reconstructed phase information , were obtained at a pixel size of 1.5 Å/px. These potential 521 maps were imported as MRC image files into cryoSPARC 33. Micrographs were subjected to 522 automated particle picking, resulting in 38'425 particles ( Extended Data Fig. 4). 2D 523 classification identified 10 classes. The class averages were used as templates to refine the 524 automated particle picking, resulting in 13'665 particles. After another 2D classification and 525 selection of the best classes, a total of 11'552 particles were used for ab -initio 3D structure 526 determination. No further correction for any remaining contrast transfer function was applied. 527 The resulting 3D map was analyzed by gsFSC with the 0.143 cutoff criterion, showing a final 528 resolution of 5.8 Å of the map (Fig. 3a, and Extended Data Figs. 4 and 7 , and 529 Supplementary Table 1). 530 For the Phi92 bacteriophage, eighty-three 4D-STEM datasets were acquired using the same 531 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 20 methodology as before, but at a CSA of 5.1 mrad. The micrographs were recorded at a total 532 electron dose of approximately 49 e-/Ų. The pixel size of the final micrographs was 1.44 Å/px. 533 The 2D ptychographic micrographs obtained were imported as MRC image files into both 534 cryoSPARC 33 and RELION 34,35 for subsequent analysis. The helical reconstruction process 535 of the bacteriophage's sheath was conducted using cryoSPARC, while the capsid 536 reconstruction was performed using RELION. 537 Initial manual particle picking of the bacteriophage sheath in cryoSPARC yielded 2'420 538 particles. Following 2D classification and selection of the best classes, 1'600 particles were 539 utilized for ab -initio 3D reconstruction. The helical parameters employed in this process 540 included a helical rise of 35.30 Å and a twist of 26.3 degrees. The final processing stage 541 involved non-uniform refinement (NU-Refine) with a Z fraction set at 50%, resulting in an 8.4 542 Å resolution map of the Phi92 sheath. 543 Concurrently, manual particle picking of the capsid in RELION resulted in the identification of 544 356 particles. After a similar 2D classification process, 182 particles were selected for the final 545 3D reconstruction. This approach yielded a 12.1 Å resolution map of the Phi92 capsid ( Fig. 546 3b, Extended Data Figs. 5 and 7, and Supplementary Table 1). 547 For the tobacco mosaic virus (TMV), forty-three 4D-STEM datasets were acquired, following 548 the same protocol as before, but with a CSA of 6.1 mrad. The datasets were acquired at a 549 total dose of approximately 32 e-/Ų per micrograph, and with a pixel size of 1.15 Å/px. 550 These 2D ptychographic micrographs were imported as MRC image files into the RELION 551 software. The initial phase of manual particle picking of TMV in RELION resulted in 3'400 552 particles. Subsequent 2D classification and the selective process for the best classes reduced 553 the number to 2'120 particles, which were then used for the ab-initio 3D reconstruction. Using 554 helical parameters for the processing of a helical rise of 1.38 Å and a twist of 22.036 degrees 555 resulted in a 6.4 Å resolution map of the TMV ( Fig. 3c, Extended Data Fig. 6 and 7, and 556 Supplementary Table 1). 557 ResLog B-factor estimation 558 The information content of the recorded data from conventional TEM and 4D -STEM 559 ptychography imaging was evaluated by calculation of the FSC0.143, the Guinier-plot amplitude 560 falloff B-factor, and the Reconstruction "ResLog" B-factor 31,42. With the latter, the number “#” 561 of particles required to reach a resolution of "d" Angstrom, can be calculated as: 562 # = ( 1 𝑁asym ) ⋅ [(〈𝑆〉 〈𝑁〉) 2 ⋅ 30 𝜋 𝑁𝑒 ⋅ 𝜎𝑒 ⋅ 𝑑] ⋅ 𝑒 𝐵 2∙𝑑2 563 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 21 with 𝑁asym being the number of asymmetric units in the particles, 〈𝑆〉 〈𝑁〉 being the signal-to-noise 564 ratio of the electrons in the recorded data, 𝜎𝑒 = 0.004 Å2 being the elastic cross -section of 565 carbon, 𝑁𝑒 being the used electron dose, 𝑑 being the real-space resolution in Å, and 𝐵 being 566 the ResLog B-factor in Å2. 567 This can be simplified to: 568 # = const⋅ 𝑒 𝐵 2∙𝑑2. 569 The ResLog B-factor in Å2 can be calculated, based on the relation: 570 ln(#)= 𝐵 ⋅ 1 2⋅𝑑2 + offset, with offset= ln(const). 571 When computing the resolution 𝑑 of a reconstruction with smaller subsets of # number of 572 particles and plotting ln(#) as a function of 1 2⋅𝑑2, the ResLog B -factor becomes apparent as 573 the slope of a line fitted through the experimental data points. 574 The predicted theoretically achievable resolution when expanding the dataset to more 575 particles could then be obtained with: 576 𝑑 = √ 𝐵 2∙(ln(#)−offset) . 577 The computed FSC0.143, Guinier-plot B-factors, ResLog B-factors, and the predicted resolution 578 for 1'000'000 particles are given in Supplementary Table 1. 579 580 Data availability 581 For the apoferritin analysis by real-space cryo-transmission electron microscopy, the raw cryo-582 EM images were deposited to the Electron Microscopy Public Image Archive (EMPIAR) with 583 accession code EMPIAR-XXX. The cryo-EM map was deposited to the Electron Microscopy 584 Data Bank (EMDB) with accession code EMDB -19436. The atomic model was deposited to 585 the Protein Data Bank (PDB) with accession code PDB-8RQB. 586 For the apoferritin analysis by cryo -electron ptychography, raw data are available EMPIAR 587 XXX and the reconstructed volume at EMD-19425. 588 For the Phi92 analysis by cryo-electron ptychography, raw data are available at EMPIAR XXX. 589 The reconstructed sheath volume is at EMDB-19430 and the reconstructed capsid volume at 590 EMD-19423. 591 For the TMV analysis by cryo -electron ptychography, raw data are available EMPIAR XXX 592 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 22 and the reconstructed volume at EMD-19413. 593 Code availability 594 The py4DSTEM package is available at https://github.com/py4dstem/py4DSTEM. The d ata 595 processing Jupyter Notebook utilized here is available at https://github.com/LBEM-CH/4D-596 BioSTEM. 597

598 We acknowledge fruitful discussions with Duncan Alexander, Cécile Hébert, and Marco 599 Cantoni, EPFL Lausanne. 600 Author Contributions 601 B.K. designed the data acquisition and processing workflow, and integrated the components 602 of the experimental setup. B.K., I.M., and R.C.G. collected ptychography data. I.M. performed 603 the single particle analysis of the ptychographically acquired data. S.R., G.V., and C.O., are 604 the developers of the py4DSTEM software and provided essential support in the processing 605 of ptychographic data. K.L. purified and prepared the Apoferritin sample. S.N. and A.M. 606 collected TEM images and ran Single Particle Analysis (SPA) on the Apoferritin sample. 607 M.L.L., C.S. and K. M.-C. provided expertise and support on diffractive imaging. H.S. designed 608 the experiments, provided resources, and oversaw data analysis and interpretation. B.K. and 609 H.S. wrote the manuscript with input from all authors. 610 Funding 611 This work was supported by the Swiss National Science Foundation, grant 200021_200628 612 and by the European Union (ERC 4D -BioSTEM, No. 101118656). Views and opinions 613 expressed are, however, those of the authors only and do not necessarily reflect those of the 614 European Union or the European Research Council Executive Agency. Neither the European 615 Union nor the granting authority can be held responsible for them. 616 Work at the Molecular Foundry was supported by the Office of Science, Office of Basic Energy 617 Sciences, of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231. SMR 618 and CO acknowledge support from the U.S. Department of Energy Early Career Research 619 Program. GV acknowledges support from the Miller Institute for Basic Research in Science, 620 Competing Interest 621 The authors declare no competing interests. 622 623 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 23 Extended Data 624 625 Extended Data Fig. 1: The 4D-STEM electron probe from a micrograph recorded at 6.1 626 mrad CSA and -758 nm defocus. a , Amplitude and phase of the reconstructed real -space 627 electron probe on the sample. b, Amplitude and phase of the reconstructed Fourier probe on 628 the detector. 629 630 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 24 Extended Data Fig. 2: Real-space cryo -transmission electron microscopy structural 631 analysis of Apoferritin at 1.09 Å resolution. 632 633 634 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 25 Extended Data Fig. 3: Real-space cryo-transmission electron microscopy of apoferritin 635 and its data processing details. 636 637 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 26 Extended Data Fig. 4: Ptychography structure determination of Apoferritin, using 4D -638 STEM data. The atomic model of Apoferritin used for the fitting was PDB 8J5A. 639 640 641 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 27 Extended Data Fig. 5: Ptychography structure determination of the bacteriophage 642 Phi92, using 4D-STEM data. 643 644 645 646 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 28 Extended Data Fig. 6: Ptychography structure determination of tobacco mosaic virus 647 (TMV), using 4D-STEM data. 648 649 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 29 Extended Data Fig. 7: FSC curves belonging to each individual dataset. 650 651 652 653 654 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint Küçükoğlu et al. Cryo-electron ptychography 30 655 Supplementary Information 656 657 Apoferritin Apoferritin Phi92 Sheath Phi92 Capsid TMV Microscope TFS Titan Krios, CFEG, Falcon 4i TFS Titan Krios, CFEG, Probe corrector, TVIPS USG, Dectris ELA detector Voltage [kV] 300 Nominal Magnification 250'000 ––– Electron exposure [e-/A2] 40 35 49 32 CSA [mrad] ––– 4.0 5.1 6.1 Detector outer angle [mrad] ––– 6.58 6.85 8.61 Step size [Å] ––– 20.0 Probe positions per micrograph ––– 128 x 128 Micrographs 9'846 90 83 43 Defocus [µm] 0.5 - 0.9 1.4 - 2.3 1.3 - 2.1 0.8 - 2.0 Pixel size [Å] 0.3 1.50 1.44 1.15 Initial particle numbers 1'899'133 13’665 2’420 356 3’400 Final particle numbers 411'705 11'552 1’600 182 2’120 Symmetry imposed O O Helical I Helical Guinier Plot B-factor [Å2] 18 642 350 ––– 283 FSC threshold 0.143 Map resolution [Å] 1.09 5.8 8.4 12.1 6.4 EMPIAR ID TBD TBD TBD TBD TBD EMDB ID EMD- 19436 EMD- 19425 EMD- 19430 EMD- 19432 EMD- 19413 PDB ID 8RQB ––– ––– ––– ––– ResLog B-factor [Å2] 28 174 739 ––– 358 Predicted resolution if expanding to 1'000'000 particles [Å] 1.05 3.11 6.08 ––– 4.38 658 Supplementary Table 1 : Cryo-EM and cryo -electron ptychography data collection and map 659 statistics 660 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.12.579607doi: bioRxiv preprint