Cryo-Electron Microscopy and Biochemical Analysis Offer Insights Into the Effects of Acidic pH, Such as Occur During Acidosis, on the Complement Binding Properties of C-Reactive Protein

The pentraxin family of proteins includes C-reactive protein (CRP), a canonical marker for the acute phase inflammatory response. As compared to normal physiological conditions in human serum, under conditions associated with damage and inflammation, such as acidosis and the oxidative burst, CRP exhibits modulated biochemical properties that may have a structural basis. Here, we explore how pH and ligand binding affect the structure and biochemical properties of CRP. Cryo-electron microscopy was used to solve structures of CRP at pH 7.5 or pH 5 and in the presence or absence of the ligand phosphocholine (PCh), which yielded 7 new high-resolution structures of CRP, including pentameric and decameric complexes. Structures previously derived from crystallography were imperfect pentagons, as shown by the variable angles between each subunit, whereas pentameric CRP derived from cryoEM was found to have C5 symmetry, with subunits forming a regular pentagon with equal angles. This discrepancy indicates flexibility at the interfaces of monomers that may relate to activation of the complement system by the C1 complex. CRP also appears to readily decamerise in solution into dimers of pentamers, which obscures the postulated binding sites for C1. Subtle structural rearrangements were observed between the conditions tested, including a putative change in histidine protonation that may prime the disulphide bridges for reduction and enhanced ability to activate the immune system. Enzyme-linked immunosorbent assays showed that CRP had markedly increased association to the C1 complex and immunoglobulins under conditions associated with acidosis, whilst a reduction in the Ca2+ concentration lowered this pH-sensitivity for C1q, but not immunoglobulins, suggesting different modes of binding. These data suggest a model whereby a change in the ionic nature of CRP and immunological proteins can make it more adhesive to potential ligands without large structural rearrangements.

paper (Whatman, Merck, USA), before 10 µl of uranyl formate (2%) was added to stain for 10 seconds, before excess stain was removed with filter paper. Samples were left to air-dry before being imaged on an FEI Tecnai 12 Twin Transmission electron microscope (ThermoFisher, USA) with a LaB6 filament operating at 120 kV. Images were acquired with a defocus of ~0.5-1 µm on an Eagle 4k x 4K CCD camera (ThermoFisher).
CryoEM data analysis of CRP at pH 7.5 -PCh: All reconstructions were performed using Relion (1). Particles were extracted to produce a stack of 650,386 particles that were binned 4× with a box size of 64 pixels. The particles were subject to 2D classification with 200 classes and a 110 Å soft circular mask. Classes representing bad particles (e.g. contamination or noise) were discarded yielding a stack of 495,078 particles. To select pentameric CRP, 3D classification was utilized, using a low pass filtered (40 Å) map of PDB deposition 1B09 as a reference map and mask. Classes that resembled the tetrameric like side view of the decamer were discarded yielding a stack of 367,537 particles. These were then re-extracted to 2× binning with a box size of 128 pixels. These were subject to 2D classification with 100 classes and a 110 Å soft circular mask. Bad particles or decamers were discarded producing a stack of 322,949 particles. 3D classification was then used with 6 classes, using the low pass filtered map of 1B09 as previously described. Low resolution classes were discarded producing classes representing 256,289 particles. They were then refined with C5 symmetry, subject to Bayesian polishing and refined again. These particles were extracted without binning and a box size of 360 pixels. These were then refined with C5 symmetry imposed and thereafter subject to beam tilt, anisotropic magnification and per particle defocus estimation within Relion 3.1 (1). This was repeated twice for a total of three iterations. A further C5 refinement was carried out, and then the particles were polished as before. Duplicates were removed, as defined by particles within 100 Å of each other, yielding 254,184 particles. This particle set was used for final refinements with both C5 symmetry and C1 symmetry that reached 3.2 Å and 3.5 Å respectively with the latter using a loose mask. To extract decamers from the population of particles, unmasked 3D classification was carried out on the 4× binned 2D classes. Classes representing decamers were selected yielding a stack of 204,452 particles. These were extracted with a box size of 128 pixels and binned 2×. The particles underwent 2D classification with 100 classes and a 110 Å soft circular mask. Bad particles, contamination or pentamers were discarded producing a set of 204,354 particles. This was refined with C1 symmetry and then extracted without binning with a box size of 360 pixels. This was again refined with C1 symmetry and subsequently subject to three iterations of beam tilt, anisotropic magnification and per particle defocus respectively as described above. These particles were then refined with C1 symmetry and then polished. The polished particle set was then refined with C1 symmetry. Duplicates were then removed, as defined as particles within 140 Å of each other, yielding a particle stack of 185,953. This was refined with C1 symmetry resulting in a final map of 2.8 Å. Local resolution for both the pentameric and decameric maps was calculated using an implementation within Relion.

CryoEM data analysis of CRP at pH 5 -PCh:
Particles were extracted with 4× binning and a box size of 64 pixels, from two data collection sessions, to give a combined total of 733,877 particles. These underwent 2D classification with 50 classes and a 110 Å soft circular mask. Bad classes and contamination were discarded, giving a stack of 617,674 particles. To filter out pentamers from the decamers in the population, masked 3D classification was utilized. Decamers were identified via a characteristic tetrameric like side view and discarded. This resulted in a particle set of 238,619 particles, which were thereafter extracted with 2× binning and a box size of 128 pixels. This stack underwent 2D classification with 50 classes and a 110 Å soft circular mask. Bad classes and decamers were discarded producing a set of 207,193 particles. This was refined with C5 symmetry imposed and reextracted without binning and a box size of 360 pixels. This stack of particles underwent 2D classification with 50 classes and a 110 Å soft circular mask. Good classes were selected yielding a total of 198,659 particles. These later went into 3D classification with 4 classes. The lowest resolution class was discarded resulting in a set of 171,942 particles. This was refined with C5 symmetry and thereafter underwent three iterations of beam tilt, anisotropic magnification and per particle defocus estimation as described above. The particles were then polished and refined using C5 symmetry. Duplicates were then removed as defined by particles that were closer than 100 Å, resulting in a final stack of 171,144 particles. This was then refined with C5 symmetry 3.3 Å and C1 refinement with a soft mask 4 Å. To select the decamers unmasked 3D classification with 6 classes was used on the particle set form the 4× binned 2D classification. Classes representing decamers were selected resulting in a stack of 362,064 particles. These were extracted with 2× binning and a box size of 128 Å and subjected to 2D classification with 50 classes and a circular soft mask of 130 Å. All the classes were of good enough quality so were all used in a subsequent refinement with C1 symmetry. These were extracted without binning and a box size of 360 pixels before 2D classification was repeated with 50 classes and a soft circular mask of 140 Å. Again, all classes were of good quality so were used in a refinement with C1 symmetry. This particle set was then subjected to three iterations of beam tilt, anisotropic magnification and per particle defocus, respectively, as described previously. Thereafter the particles were refined with C1 symmetry imposed. The particles were polished and duplicates were removed, as defined as particles closer than 140 Å, to give a final stack of 315,988 particles. This was refined with C1 symmetry to give a final map of 2.8 Å. Local resolution for both the pentameric and decameric maps was calculated using an implementation within Relion.

CryoEM data analysis of CRP at pH 7.5 +PCh:
Particles were extracted with 4× binning and a box size of 64 pixels to give a total of 1,596,930 particles. The particles were subject to 2D classification with 200 classes and a 110 Å soft circular mask. Bad classes and contamination were discarded yielding a particle stack of 1,461,767 particles. These were subject to unmasked 3D classification using a low pass filtered map of 1B09 as a reference with 6 classes. A 3D class corresponding to a pentamer was selected to filter out this subpopulation of 414,369 particles. These were then extracted to 2× binning with a box size of 128 Å and subject to another iteration of 2D classification with 100 classes and a 110 Å soft circular mask. Bad classes, contamination and decamers were discarded, resulting in 397,433 particles. These particles were then used in an unmasked 3D classification run, using a low pass filtered map of 1B09 as before with 2 classes. The highest resolution class was selected, corresponding to 269,217 particles, and then refined with C5 symmetry imposed. The particles were then extracted without binning with a box size of 360 pixels. They were then refined with C5 symmetry imposed and then subject to polishing within Relion. The particles were refined again and then duplicates were removed, as defined by particles within 140 Å of each other, yielding 259,507 particles. Next, the C5-symmetric refinement was repeated, as well as a separate refinement using C1 symmetry coupled with a loose mask, yielding maps with resolutions of 3.3 Å and 3.5 Å respectively. To select decamers, a class representing a well-defined decamer was selected from the 4× binned unmasked 3D class averaging job previously described. This corresponded to 393,732 particles. These were reextracted with 2× binning and a box size 128 pixels. The particles underwent 2D classification with 100 classes and a 130 Å soft circular. Contamination, poorly aligned particles and pentamers were discarded, giving a total of 352,317 particles. This stack was used in a 3D classification job with 2 classes, using the decameric class from the previous 3D classification job as a reference map. The highest resolution class, containing 175,177 particles, was then extracted without binning and a box size of 360 pixels. This was refined without symmetry and polished. The polished particles set was refined before duplicates were removed, as defined by particles within 140 Å of each other, yielding 167,957 particles. This was refined as before yielding a map of 3.3 Å. Local resolution for both the pentameric and decameric maps was calculated using an implementation within Relion.
CryoEM data analysis of CRP at pH 5 +PCh: Given the suboptimal particle distribution for this condition, three separate data collections were required to acquire sufficient numbers of particles for high-resolution analysis. In the case of the decameric population of particles, all three data collections were processed independently until post particle polishing. For the first and second data collection, henceforth known as optics group 1 and 2, the same in-silico purification strategy was used, whilst keeping the data sets separate and independent. First, particles were extracted to 4× binning with a box size of 64 pixels. Particles underwent 2D classification with 100 classes and a 110 Å soft circular mask. Classes representing contamination or poorly aligned particles were discarded. These were used as an input for an unmasked 3D classification job with 4 classes, using a low pass filtered map of the PDB entry 1B09 as a reference map. A single class representing a decamer was selected. These were extracted with 2× binning with a box size of 128 pixels and were subject to further 2D classification with 100 classes and a 130 Å soft circular mask. Contamination, poorly aligned particles or classes representing pentamers were discarded. These were extracted with no binning and a box size of 360 pixels. These underwent another round of 2D classification, in an identical manner as described above, and well aligned decamers were selected giving a total of 99,026 and 119,214 particles for optics group 1 and 2, respectively. These were separately refined with no symmetry imposed and then subject to 3 iterations of a sequence of beam tilt, anisotropic magnification and per particle defocus estimation. For the third data collection, optics group 3, particles were extracted with 4× binning and subject to 2D class averaging as described above for optics groups 1 and 2. Particles were then subject to two rounds unmasked 3D classification using the same settings as for optics group 1 and 2. Contamination and low-resolution particles were discarded after each iteration, resulting in a stack of 1,086,837 particles. They were then re-extracted with 2× binning and a box size of 128 pixels and subject to 2D class averaging classification with 100 classes and a 130 Å soft circular mask. Both pentameric and decameric particles were selected, but contamination or poorly aligned particles were discarded yielding 983,286 particles. As before, 3D classification was used to discard particles comprising lower resolution classes and then the resultant particles underwent 2 iterations of 2D class averaging with a 130 Å soft circular mask. This resulted in a population of decamers and pentamers consisting of 858,265 particles. Decamers were selected yielding a stack of 471,385 particles which underwent 3D classification with 2 classes using a low pass filtered version of our pH 7.5 cryoEM derived decameric model as a reference map. The highest resolution class, comprising 378,258 particles, was selected and extracted without binning and a box size of 360 pixels. This was then refined without symmetry and subject to particles polishing. Subsequently, the particles were refined as before and beam tilt, anisotropic magnification and per particle defocus were estimated as described in previous conditions. This was then refined without symmetry. Subsequently, all three optics groups were combined, subject to 3D class averaging with 2 classes and the highest resolution class was selected and duplicates were removed, as defined by particles within 140 Å of each other, producing a final combined particle stack of 323,169 particles. This was refined without symmetry to yield a 3 Å map. To separate out the pentamers, 2D classes representing pentamers were selected from the 2D classification job at 2× binning for optics group 1. For optics group 2, a 3D class representing a pentamer was taken from the 4× binned data, subsequently extracted with 2× binning and underwent 2D classification selecting for pentamers. Pentamers from optics group 3 were selected from the second iteration of 2D classification in the 2× binned data. This resulted in three independent stacks of 10,007, 49,289 and 385,487 particles for optics groups 1, 2 and 3, respectively. Particles were processed together hereafter. Particles were subjected to 2 iterations of 2D classification with 100 classes and a soft mask of 150 Å. Contamination or poorly aligned particles were discarded after each iteration, yielding a stack of 174,402 particles. The particles were underwent unmasked 3D classification, with 2 classes using a low pass filtered map of 1B09 as a reference. After, two iterations of 2D classification were used as before to further clean the particle set. Refinements were attempted several times during this cleaning process, with none reaching below ~8 Å in resolution. Instead, the best pentameric 2D classes from the final iteration of 2D classification were selected, then duplicates were removed, as defined by particles within 100 Å of each other, yielding 23,302 particles.

Figure S12. Negative stain EM of CRP under pH and calcium concentrations tested in ELISAs.
Individual pentamers dominate (orange squares), but decamers are also present (blue squares). Scale bar = 100 nm for all panels. Figure S13. Analysis of additional density proximal to His95, Cys97 and Gly113. Density (black mesh) and chain A of each high-resolution model for all conditions. A water molecule modelled into each density can hydrogen bond to deprotonated (pH 7.5) and protonated (pH 5) histidine residue His95 and the backbone amine of glycine Gly113. Carbon, nitrogen, oxygen, sulphur and hydrogen atoms are coloured light purple, blue, red, yellow and white, respectively.

Figure S14. Locations of CRP residues involved in C1q binding and activation.
Locations of residues previously known to be important in C1q binding and complement activation include His38, Lys57, Arg58, Glu88, Asp112, Lys114, Asp169, Thr173, Tyr175 and Leu176 (3)(4)(5). These residues are shown on the C5-symmetric pentameric model at pH 7.5 (A). Model is in light blue, and the carbon atoms of highlighted residues are coloured orange, with the oxygen and nitrogen atoms coloured red and blue, respectively. B) Locations of the same residues on the decameric model at pH 7.5. Model is coloured light purple with the carbon atoms of highlighted residues and heteroatoms coloured as in A.
For clarity, Lys57, Arg58 and Glu88 have not been shown. Supplementary Tables   Table S1. Angles between monomers within high-resolution pentameric CRP structures as measured between Asp60 residues, in cryoEM derived atomic models presented in this study. Dec and Pent denote decameric and pentameric CRP structures, respectively. Red and blue signify larger and smaller angles, respectively. Angles were measured using UCSF Chimera.  Table S3. Cross correlation and RMSD comparisons between each high-resolution model and map resolved from cryoEM data presented in this study. Cross correlation and RMSD values were calculated using chimera. Pentamers (Pent) were compared to chains A-E of decamers (Dec). Red signifies greater deviation between maps/models, whereas blue represents greater similarities. Cross correlation and RMSD values were calculated using UCSF Chimera. Table S4. RMSD between C5-symmetric pentameric models resolved from cryoEM data presented in this study with a subset of crystal structures. Red signifies greater deviation between models, whereas blue represents greater similarities. RMSD values were calculated using UCSF Chimera.  Table S5. Comparison of EM derived and X-ray crystallographic protein models. X-ray crystal structures are denoted by their PDB codes and the cryoEM derived models presented in this study. Oligomeric state, pH and ligands present are shown. Red and blue signify more acidic and more basic pH, respectively.