Anaerobic fixed-target serial crystallography using sandwiched silicon nitride membranes

Anaerobic, room-temperature X-ray diffraction and absorption spectroscopy is performed using sandwiched silicon nitride membranes, prepared using 3D-printed tools, in an oxygen-free environment with a reducing agent. Using crystals of hemoglobin as a model system, it was demonstrated that this method ensures oxygen-free data collection.


Introduction
Serial synchrotron crystallography (SSX) is quickly becoming a popular approach for low-dose, room-temperature structural biology research (Stellato et al., 2014;Gati et al., 2014;Coe & Fromme, 2016;Diederichs & Wang, 2017).The development of fourth-generation storage-ring light sources (Eriksson et al., 2014) has enabled X-ray beams with high brilliance and micrometre-scale focusing; combined with advances in fastframe-rate X-ray detectors (Leonarski et al., 2020), dedicated SSX beamline endstations are now available for users.Current sample-delivery methods for SSX broadly fall into one of two categories: fixed-target methods, where crystal slurries are deposited onto a substrate that is rastered across the beam (Sherrell et al., 2015(Sherrell et al., , 2022;;Schulz et al., 2018;Zielinski et al., 2022), and flow systems, where crystals suspended in a medium are directly passed through the beam (Botha et al., 2015;Shilova et al., 2020;Nam, 2020;Knos ˇka et al., 2020;Ghosh et al., 2023).Still, challenges remain, from the amount of sample used to heating of the system, clogging of capillaries, burning of the material and sample sensitivity.One sample type with specific needs are metalloproteins, which can be estimated to make up a large part of proteomes.Anaerobic sample environments extend the range of redox chemistry that can be characterized by crystallographic means and are of use for studies of redox-active metalloproteins (Handing et al., 2018).However, the requirement for anaerobicity adds another layer of challenge and different solutions have been proposed, including the use of environmental chambers around the sample position (Fuller et al., 2017;Mehrabi et al., 2021) or sealing fixed-target chips under anaerobic conditions (Rabe et al., 2020).
Hemoglobin (Hb) is an iron-containing heme protein that is essential for oxygen transport.The protein is an assembly of four globular subunits, each containing a heme B (protoporphyrin IX group; Hsia, 1998).Equine hemoglobin was among the first protein structure to be solved at high resolution (Perutz et al., 1960).It is a useful model system for oxygen-sensitive metalloproteins due to its ubiquity and its high but reversible affinity for oxygen.Deoxyhemoglobin (DeoxyHb) is missing a sixth ligand in the coordination sphere of the ferrous heme iron (composed of the four heme nitrogens and a histidine) and thus is very sensitive to the presence of oxygen, oxyhemoglobin (OxyHb) has one O 2 molecule bound to a ferrous heme iron, and methemoglobin (MetHb), the oxidized form of Hb, has a heme iron that is in the +3 (ferric) oxidation state with a water bound in the sixth ligand position (Yang & Phillips, 1996).Both DeoxyHb and MetHb are used here as proofs of concept for our anaerobic system.
Here, we describe a method for anaerobic SSX at room temperature with silicon nitride membranes that can easily be implemented on another beamline.We have developed a fixed-target method that 'sandwiches' oxygen-sensitive crystals in between silicon nitride (SiN) chips (Coquelle et al., 2015;Shilova et al., 2020), protecting them from ambient oxygen.Using crystals of DeoxyHb, we demonstrate the practicality of this design for anaerobic SSX, and further verify using UV-Vis and X-ray spectroscopy that the hemoglobin crystals remain in the deoxy state.

Preparing microcrystals
Equine hemoglobin (Sigma, CAS 9047-09-0) stock solution at 20-25 mg ml À 1 was prepared in 10 mM HEPES pH 7.5 and crystallized using the stirred batch method (Sato-Tomita & Shibayama, 2017) by mixing in precipitant solution [26%(v/v) PEG 3350, 10 mM HEPES pH 7.5].At 20 � C, 500 ml precipitant was added to 250 ml protein solution.The mixture was stirred in small HPLC vials with rice-grain-sized stir bars on a magnetic stirrer for about 20 h, after which crystals of 5 � 5 � 5 mm in size were obtained.

3D-printed holder and accessories
Since handling this type of chip is challenging due to its small size and thinness, we have designed an assembling tool, a chip tray and a MX sample holder for easier manipulation.The MX sample holder was printed with Tough 1500 Resin (Formlabs, USA), while for the rest of the 3D-printed components Tough Resin V5 (Formlabs, USA) was used.The assembling tool went through a few iterations, as can be seen in Supplementary Fig. S1.The final iteration accounts for imperfections in chip size (some sides being longer or with 'horns' in the edges), avoids too high pressure with the lever that would destroy the chips and leaves one corner easily accessible for handling upon assembly (Fig. 1a).Using the chip tray (Fig. 1b), several chips can be prepared for quicker assembly.To ease sample mounting to the goniometer head at MX beamlines, the MX sample holder was designed to be easily attached to SPINE bases (Cipriani et al., 2006;Fig. 1c).The tightening screws allow extra pressure to be applied to the sides of the sandwiched chips during data collection, further ensuring airtightness.

Anaerobic chip assembly for X-ray crystallography
Before data collection, DeoxyHb crystals were freshly prepared inside a nitrogen-filled glovebox (Coy Laboratory Products).The average oxygen concentration in the glovebox was 2 p.p.m. during sample preparation and never went above 10 p.p.m..The aerobically grown hemoglobin microcrystals were transferred into the glovebox and chemically reduced by mixing sodium dithionite solution [50 mM sodium dithionate, 26%(w/v) PEG 3350, 10 mM HEPES pH 7.5] (Vojte ˇchovsky ét al., 1999) with hemoglobin microcrystals (5 and 20 ml, respectively) and incubated within the glovebox at room temperature for 2 h (during which the crystal slurry undergoes a colour change from brown to pink).Inside the glovebox, SiN chips (membrane size 2.5 � 2.5 mm, membrane thickness 1000 nm, frame size 5.0 � 5.0 mm, frame thickness 200 mm; Silson, United Kingdom) were loaded onto the assembling tool (Figs.2a and 2b), followed by the addition of 2 ml microcrystal slurry (Fig. 2c).To seal the chip, a small amount of superglue (Loctite Super Glue Brush On) was applied onto all four corners of the chip frame in an L-shape with a single hair from a small paintbrush (Fig. 2d).A second chip was then placed on the top (Fig. 2e) and pressed gently with the lever (Fig. 2f ) to secure it and to create a seal, as confirmed by a tiny amount of glue overflowing along all edges.The assembled chip was then placed in the MX sample holder (Fig. 1c) for data collection.

Anaerobic chip assembly for UV-Vis and X-ray absorption near-edge structure (XANES)
The purpose of the XANES experiment was to establish a time stamp to determine how long the chips can retain the deoxyhemoglobin form.A special holder was designed to fit six chip sandwiches into the beamline sample environment setup (Supplementary Figs.S2b and S2c).The chips were assembled in the same way as described in Section 2.3.
UV-Vis spectroscopy using the MX sample holder, with the chip prepared in the exact same way as for X-ray crystallographic data collection, was carried out with an Agilent Cary 60 UV-Vis spectrophotometer in the wavelength range 300-800 nm and followed over 2 h.The UV-Vis spectrophotometer was modified with an external optical cable with a probe leading to a black box with a holder for the MX sample holder.

Data collection
2.5.1.Crystallographic data collection.Data for MetHb and DeoxyHb were collected on the BioMAX beamline (Ursby et al., 2020).The prepared chip sandwich (Supplementary Fig. S2a) was mounted and aligned on the goniometer.Sample viewing, alignment and measurement were performed using the MXCuBE3 beamline-control software (Mueller et al., 2017).The chip was rastered through the X-ray beam in a regular snake-like pattern defined by a mesh grid drawn on the chip in MXCuBE3.The fast shutter was open during each row of the mesh.Rotation of the goniometer was blocked during data collection.Diffraction data were collected at room temperature with an X-ray beam size of 20 � 5 mm (full-width at half-maximum), through a 10 mm diameter aperture, at a photon energy of 12.6 keV and a flux of 2 � 10 12 photons s À 1 .10 000-20 000 images per chip were collected from five chips for both MetHb and DeoxyHb.The crystallographic data-acquisition process was split into several smaller mesh scans spaced apart.Consequently, instead of collecting 10 000-20 000 images on a single mesh, they were obtained through several smaller mesh scans.Diffraction data were recorded with an EIGER 16M hybrid pixel detector with an exposure time of 0.011 s.Each crystal received an average radiation dose of �40 kGy as calculated using RADDOSE-3D (Zeldin et al., 2013).For the DeoxyHb samples, it took 5-10 min to take the sample from the glovebox, put it inside the beamline hutch and perform data collection.
2.5.2.XANES data collection.XANES data were collected in fluorescence mode on the Balder beamline at MAX IV Laboratory using a silicon drift detector (SDD; X-PIPS 7-element SDD, Mirion Technologies, USA).The liquidnitrogen-cooled double-crystal (Si111) monochromator (FMB Oxford, United Kingdom) was calibrated to 7112 eV for the first inflexion point of an iron foil for Fe K-edge data collection.XANES spectra of the Fe K edge were recorded as continuous-fly scans (settings: acquisition time per point 10 ms, energy step 0.2 eV, range 7062-7212 eV, 23 s per scan) to investigate the oxidation state of the heme iron of hemoglobin in the SiN chips.The first chip was used to find and focus the beam; after this, every chip was collected once (each collection consists of five scans performed in direct sequence) with a time separation of 20 min between the chips.In addition to the chips with DeoxyHb crystals, a fully oxygenated crystal sample (a DeoxyHb sample exposed to air for 3 min) was measured.The best measurement strategy was a defocused beam (0.2 � 2 mm) with a manganese filter (Z À 1) in front of the fluorescence detector.The incoming flux at the sample was 3 � 10 12 photons s À 1 .With the given flux and beam parameters, the dose was estimated to be around 67 kGy per single scan (23 s).To obtain the normalized XANES spectrum presented in the figures, fluorescence counts were divided by ten (ionization chamber reading incoming flux) and normalized to unity edge step.

Data processing, model building and refinement
Diffraction data were indexed, integrated, merged and converted to MTZ format using CrystFEL 0.10.1 (White, 2019;White et al., 2012).The indexing rates were 43.3% and 53.6% for MetHb and DeoxyHb, respectively.Data truncation, phasing and structure refinement were performed in CCP4 Cloud (Krissinel et al., 2022;Agirre et al., 2023).Highresolution structures of DeoxyHb (PDB entry 1ibe; Wilson et al., 1996) and MetHb (PDB entry 6sva; Mikolajek et al., 2023) were used as models for molecular replacement with Phaser (McCoy et al., 2007).The structures were refined by one round of rigid-body refinement using REFMAC5 (Nicholls et al., 2018;Murshudov et al., 1997Murshudov et al., , 2011) ) followed by several rounds of restrained refinement.Model building was performed in Coot (Emsley et al., 2010;Krissinel & Henrick, 2004) and all structural representations were generated in PyMOL (Schro ¨dinger).Room-temperature SSX structures of MetHb and DeoxyHb were obtained at 1.95 and 1.85 A ˚resolution, respectively.Data-collection and refinement statistics are presented in Table 1.The DeoxyHb chips were also processed and refined separately to investigate them individually (datacollection and refinement statistics are presented in Supplementary Table S1).

Structural differences between deoxy and met forms
The �-subunits of DeoxyHb have a water molecule in the heme pocket about 3 A ˚(3.1 A ˚in this case) from the iron and hydrogen-bonded to the distal histidine (Fig. 3).In MetHb the �-and �-heme groups have a water molecule as the sixth ligand, approximately 2 A ˚(2.2 A ˚in this case) from the iron (Fig. 4).When hemoglobin changes from the deoxy form to the oxy or met form there is no significant change in the inner region of the �1�1/�2�2 dimer.Visible changes are present in the outer region of the dimer and in the dimer-interface area.The interaction between �1�2 and �2�1 moves towards the centre of the molecule.Residues in contact in the �� interfaces are between �B (Glu30/A-Phe36/A) and �H (Phe122/ B-Lys132/B), �G (Ser102/A-Leu113/A) and �G (Gly107/B-Phe118/B), and �H (Phe117/A-Lys127/A) and �B (Arg30/B-Pro36/B).The most noticeable movement is at the �-heme, where a water molecule is bound in MetHb compared with the empty �-heme in DeoxyHb.The �-heme of MetHb moves downwards to the proximal side of the heme pocket (Supplementary Fig. S3).Both structures are in agreement with published structures (Baldwin & Chothia, 1979;Liddington et al., 1992;Wilson et al., 1996).All five chips with DeoxyHb that were also independently processed exhibited the presence of the deoxy state (Supplementary Table S1 and Supplementary Figs.S12-S15), with the exception of chip 3 that contained a bubble following the preparation, resulting in too few indexed images for a separate analysis.

XANES data
From the preliminary test of the chip sandwiches with an X-ray beam at the Balder beamline, we learned that the Fe K edge shifts after long exposure (60 s).The position of the Fe K edge shifted to higher energy after the first scan (30 s), and by the second XANES scan the edge had already approached the position for fully oxygenated hemoglobin.This fact was used in a strategy to find out how long the chip seal lasts under ambient conditions in air by waiting 20 min between starting to scan each of the consecutive chip sandwiches, i.e.only the first scan of each chip was compared (10 s to reach the edge, 23 s full scan).XANES data for the six individual chip sandwiches mounted in the chip holder made for the Balder beamline showed that two of the chip sandwiches (chips 3 and 4) had intact seals when the measurement started (i.e. were still in the deoxy state; Supplementary Fig. S4), as their absorption edges were not shifted to higher energy as observed on O 2 binding.This gave us a time stamp for the seal of at least 40 min (the time between the start of scan 1 of chip 2 and of chip 4).Fig. 5 shows a comparison of the edge position of chip 4 upon the first scan and the second scan with the deoxy and oxy states, respectively.The conclusion from this comparison is that only the first scan of chip 4 (black dashed line) had an edge position corresponding to the hemoglobin deoxy state (blue line), while the second scan (grey dashed line) had already shifted to the position of the oxygenated state (red line).

UV-Vis data
Chips were prepared as described in Section 2.3.The samples in both chips measured stayed pink, and neither chip changed from the deoxy to the met form in the first 30 min according to the spectra (Fig. 6 and Supplementary Fig. S11).The second chip was also measured 1 h 30 min and 2 h 10 min after the start of the experiment (Supplementary Fig. S11).These later time points show a change from the deoxy form, which could also be seen by eye as the sample colour changed from pink to brown.

Discussion
As demonstrated by a combination of XANES, UV-Vis absorption spectroscopy and X-ray crystallography experiments, the described method slows down reoxidation of the hemoglobin to a level that allows data collection within a reasonable time frame for oxygen-sensitive samples.The XANES data showed that the hemoglobin crystals stayed in the deoxy form for at least 40 min (see Fig. 5 and Supplementary Fig. S4).The edge-scan data revealed a relatively low success rate of the chip sandwiches in the Balder holder, which was primarily attributed to the design of the multichip holder.Specifically, the middle row positions (chips 3 and 4) experi-enced pressure from both sides of the SiN chip through a twoscrew mechanism, as depicted in Supplementary Fig. S2.Conversely, the upper and lower rows of the holder applied pressure to the chip from only one side (one screw).This asymmetrical pressure distribution on the SiN seals seems to have resulted in leaks (chips 2, 5 and 6) or an increased susceptibility to breakage.Notably, only chips 3 and 4 successfully preserved the deoxy form.In the light of these observations, a new holder for crystallographic collection was developed to maintain an even pressure on all four sides of a single chip, as illustrated in Fig. 1(c).The viability of the newly designed MX sample holder was confirmed using UV-Vis spectroscopic analysis, which demonstrated the sustained retention of the deoxy state over more than 30 min (Fig. 6 and Supplementary Fig. S11).To establish the consistency of the method, crystallographic data were collected from a total of five chips, and all four chips that gave sufficiently complete data exhibited the presence of the deoxy state (Supplementary Table S1 and Supplementary Figs.S12-S15).Based on the XANES data, it became evident that X-ray exposure causes the iron edge to shift to a higher energy, although it is not clear whether this is because the chip seal breaks or whether it is due to some other effect.The reason for this observed edge shift may be either binding of O 2 to the ferrous iron (OxyHb) or oxidation of iron to the ferric state (MetHb).The doses of the first XANES scan and of the crystallographic measurements were similar (70 and 40 kGy, respectively), and both showed the presence of the deoxy form.This dose is lower than that recently determined to be a safe dose for SSX experiments (around 380 kGy; de la Mora et al., 2020), but while this was the case for lysozyme, we have studied a heme protein that may be expected to be more sensitive to radiation damage.It is advisable to keep the dose low and independently verify the oxidation state through spectroscopy if possible, as other samples might have a different sensitivity to X-ray radiation.The assembling tool for chip-sandwich preparation simplifies the process, facilitates visualization and decreases the risk of crystal or chip damage.CAD models of the tools presented are made available in the supporting information.Furthermore, SiN chips do not cause strong diffraction that can damage detectors, but are single use only.Using the MX sample holder and the protocol presented, it is feasible to collect data from oxygen-sensitive samples and the method can be implemented at many MX beamlines that have access to a glovebox and an MX compatible goniometer.

Figure 2
Figure 2 Procedure for anaerobic chip assembly.(a) Assembling tool with chip tray.(b) The first chip is loaded into the assembling tool.(c) 2 ml of sample slurry is gently loaded onto the chip.(d) Small drops of superglue are added to all four corners of the chip.(e) A second chip is placed on top to create a sealed chip sandwich.( f ) Chips are gently pressed with the lever of the assembling tool to secure the seal between them.

Figure 4
Figure 4 View of the final 2mF o À DF c electron-density map for the heme pockets of MetHb: (a) the � subunit and (b) the � subunit.Density is contoured at 1.1�.The distance between His58/63 and the heme iron is 4.1 A ˚, while the distance between the water molecule and the heme iron is 2.1 A Figure

Figure 5
Figure 5The Fe K-edge XANES of hemoglobin crystals in the deoxy state (blue, single scan) and in the fully oxygenated state (red) compared with crystals in SiN chip 4: first scan (black dashed line) and second scan (grey dashed line).Inset, a close-up view of the edge position from the area marked with a grey rectangle in the main figure.

Figure 6 UV-
Figure 6UV-Vis spectra of a chip sandwich (chip 1) with DeoxyHb in the MX sample holder (the spectra have been normalized in Origin 2018b).

Table 1
Data-collection and refinement statistics.Values in parentheses are for the highest resolution shell.