Polymorphic nanobody crystals as long‐acting intravitreal therapy for wet age‐related macular degeneration

Abstract Wet age‐related macular degeneration (wet AMD) is the most common cause of blindness, and chronic intravitreal injection of anti‐vascular endothelial growth factor (VEGF) proteins has been the dominant therapeutic approach. Less intravitreal injection and a prolonged inter‐injection interval are the main drivers behind new wet AMD drug innovations. By rationally engineering the surface residues of a model anti‐VEGF nanobody, we obtained a series of anti‐VEGF nanobodies with identical protein structures and VEGF binding affinities, while drastically different crystallization propensities and crystal lattice structures. Among these nanobody crystals, the P212121 lattice appeared to be denser and released protein slower than the P1 lattice, while nanobody crystals embedding zinc coordination further slowed the protein release rate. The polymorphic protein crystals could be a potentially breakthrough strategy for chronic intravitreal administration of anti‐VEGF proteins.


| INTRODUCTION
Age-related macular degeneration (AMD) is a degenerative disease that affects the central retina in patients above 50 years of age, and the disease risks steeply increase to 20% after 70. 1 Accounting for 8.7% of blindness cases, AMD is considered the most common cause of visual impairment.By 2040, the projected number of AMD patients will approach $288 million, a heavy health care burden worldwide. 2t AMD, a malignant and advanced subtype of clinical AMD, causes 90% drastic visual impairment in AMD patients because of its rapid disease progression. 3 the development of wet AMD progresses, the new vessels gradually invade Bruch's membrane and the retinal pigment epithelium (RPE) layer and stimulate excessive secretion of vascular endothelial growth factor (VEGF) by RPE cells, which results in aberrant proliferation of the choriocapillaris, called choroidal neovascularization.10] Regardless of the molecular design, a central theme of anti-VEGF protein therapeutics is to achieve a prolonged injection interval and avoid frequent intravitreal injection. 11The near-future goal in industry is to lengthen the injection interval from the current $2 months (aflibercept, 2 mg dose) to 3-4 months (aflibercept, 8 mg dose phase III clinical trial) by drastically increasing the injection dose. 12,13However, with a strict limitation of injection volume (<100 μl) and needle size (30 G or smaller), the injectable protein dose in solution quickly reaches its ceiling due to inadequate protein stability and exponentially increased viscosity when the protein concentration is >100 mg/ml due to the intrinsically large, complex and flexible protein structure. 14Furthermore, considering the first-order drug elimination pharmacokinetics from vitreous humor into aqueous humor, [15][16][17] simply increasing the injection dose is not a highly effective strategy for prolonging the injection interval.Theoretically, the injection interval will only increase one elimination halflife (8-11 days) if the injection dose is doubled.
We engineered a series of crystallizable anti-VEGF proteins, which formed pseudo "polymorphic" crystal lattice structures.Some of the crystalline anti-VEGF proteins could act as potential long-acting intravitreal protein therapeutics for wet AMD, based on the following intrinsic characteristics: (1) As a thermodynamically more stable system, protein crystals are naturally more stable than protein solution, which has been proven experimentally. 18(2) High concentration protein crystal suspension is expected to have lower viscosity and thus better injectability than a protein solution with the same concentration. 19,20(3) More importantly, protein release will be hindered by the crystal structure; therefore, the overall intravitreal protein retention will be modulated by the crystal lattice as well as the first-order elimination, which could significantly increase the injection interval.
To modulate the protein crystal lattice energy through surface mutation, one of the strategies we employed was the "surface entropy reduction (SER)" hypothesis.2][23] The solvent-exposed residues with large hydrophilic side chains exposed to solvents, such as Lys, Glu, and Gln, possess high conformational entropy, which is more likely to hamper the crystallization of proteins by enhancing the surface entropy shield during protein molecule packing.Replacing amino acids with Ala, which has lower conformational entropy, can significantly promote protein crystallinity.[26] Therefore, the SER hypothesis provided an available perspective to induce different crystal lattice arrangements for fluctuations in crystal lattice entropy.On the other hand, the choice of crystal lattice is inextricably linked with the interaction among the contacting interface residues.Modification of the interaction by rational mutation at the contact interface would drive the protein molecules to occupy the new position in the crystal lattice, leading to different protein matrices.
The significantly lower viscosity, superior stability, and potential longacting release property make crystalline protein a potential breakthrough for novel wet AMD therapy.Here, we successfully changed the crystal lattice arrangement of therapeutic nanobody (termed mNb-WT) by rational mutations of surface residues, which remarkably reduced the dissolution rate of protein molecules from the crystal into the solvent.By modulating the solvent compositions, the dissolution rate could be delayed further.We believe that these findings will provide a new strategy for intravitreal sustained administration in wet AMD treatment.

| Theoretical evaluation
The clearance of protein solution in the eye, as the comparatively isolated environment, corresponded to the primary clearance kinetic equation (1).
where C t is the transient concentration of the protein drug in the vitreous humor of the eye and k is the primary clearance constant.When crystalline protein is introduced, the clearance process of the protein drug from the eye could be considered as two-compartment model, and it undergoes three steps: (1) the release of protein molecules from the crystal surface; (2) the diffusion of protein molecules from the crystal surface to the entire vitreous humor of the eye and reach equilibrium; and (3) the first-order clearance of protein molecules from the vitreous humor into the aqueous humor.
Step 2 was determined by the Noyes-Whitney dissolution equation (2) 27 : where C s is the equilibrium solubility of protein molecules in certain crystal lattice, k D is the diffusion rate constant of protein molecules in the vitreous humor, A is the surface area of the crystal, V is the volume of the intravitreal aqueous volume, and L is the thickness of the diffusion layer.Thus, the net intravitreal protein concentration (C) generated by the crystal suspension in the eye can be qualitatively rewritten as follows: Therefore, the differences in the protein crystal lattice of the same protein molecules could profoundly disturb the molecular clearance of the protein drug in the vitreous humor, through the modulation of the equilibrium solubility of crystalline of protein.

| Lattice arrangement of protein crystals formed by mNb-WT molecules
We dissected the molecular structure and crystal lattice arrangement of mNb-WT by single crystal X-ray diffraction.mNb-WT protein molecules were arranged in the P1 lattice inside the lamellar crystals with two protein molecules as one asymmetric unit (Figure 1a,b).Up to a 1.27 Å resolution (Table S1), we could clearly see the participation of water and sulfate ions during crystal packing in the solvent.
Water molecules mediate a strong hydrogen bond network to stabilize two protein molecules inside the asymmetric unit (interface a) (Figure 1c).Meanwhile, the sulfate ions also contributed to the protein stability in the crystal lattice, both by affecting the protein itself and by enhancing interactions between two protein molecules that belong to different asymmetric units (Figure 1d,e).In addition to sulfate ions, water still played an essential role in contributing to protein packing in the crystal lattice to various degrees at the contact interfaces among the asymmetric units, such as interface b (Figure 1f).
Because of the small size and compact structure, mNb-WT proteins were tightly arranged in the P1 lattice, and the Matthews coefficient was 1.76 Å 3 /Da, corresponding to $30% solvent content (Table 1).

| Rational design of nanobody mutants with intact VEGF binding affinities
To reduce the surface entropy of the entire protein according to the SER strategy, we analyzed the amino acids at the protein surface using the SER calculator web server and then determined the mutation candidates (K44 and E45) to improve the quality of the mNb-WT crystal. 28tant 1 (K44A mutation), mutant 2 (E45A mutation), and mutant No significant difference was shown in the VEGF binding ability of mutants and analogues compared with mNb-WT by ELISA (Figure 2b).
The differences in EC 50 were less than two-fold, indicating that mNb-WT and mutants/analogues showed equally potent binding capability to the VEGF-A 165 protein.The SEC comparison demonstrated that almost all the nanobodies were monomers in solvent; thus, the potential difference of the molecule packing in the crystal lattice reflects the performance of individual protein molecules rather than protein oligomers (Figure 2c).The secondary structure form detected by CD spectrum also showed no significant difference among mNb-WT and mutants/analogues (Figure 2d,e).
The overall structures predicted by the AlphaFold program for all of the nanobodies were highly similar to each other (Figure 2f).These results suggested that the nanobodies are all pharmaceutic analogues and can be considered the same protein molecules in drug administration.

| Pseudo "polymorphic" crystal lattice formed by mNb-WT and its mutants/analogues
With more mutations than mutants 1-7, analogues 1 and 2 showed similar crystal lattice and crystal morphology as mNb-WT (Figure 3a,b and Figure 3f,g).Apparently, more conformational changes have emerged compared with mNb-WT molecules in the crystal lattice as the results of the sequence differences (Figure 3c,h).The distinct wiggling of the whole side chains of several surface amino acids, even the turn of the terminal group of some side chains, contributed to the variation in the molecular interaction within the crystal lattice of analogues 1 and 2. The sulfate ion participation was similar to that of mNb-WT (Figure 3d,e and Figure 3i,j), while the arrangements of sulfate ion sites in analogues 1 and 2 were more compact than mNb-WT, indicating stronger interactions among side chains at sulfate ion sites and a higher density of crystals after mutation.This could also be confirmed by crystal lattice parameters, in which the Matthews coefficient of analogues 1 and 2 changed to 1.59 and 1.58 Å 3 /Da, respectively, and the solvent content was further reduced to 22.79 and 22.30% (Table 1).
We dissected the crystal lattice arrangement of mutant 2 using the micro-electron diffraction (microED) technique after treating cluster crystals by cryo-FIB milling.We were surprised that mutant 2 was arranged in the P2 1 2 1 2 1 lattice instead of the P1 group, with only one protein molecule as the asymmetric unit, although the crystal morphology was still lamellae, similar to mNb-WT (Figure 4a,b).Many surface residues underwent dramatic conformational changes induced by only one residue mutation (Figure 4c).Unfortunately, we could not capture the electronic density map of water, thus unable to determine the participation of sulfate ions in the crystal lattice of mutant 2.However, if we overlaid mutant 2 molecules with molecule A of the mNb-WT asymmetric unit, the terminal amino group of the K54 side chain showed an orientation dramatically opposed to that of mNb-WT to form a hydrogen bond with the S32 residue (Figure 4d); however, all conformations of the S53, K54, G55, and G103 residues, which bond with the sulfate ion of molecule B in the mNb-WT asymmetric unit, matched that of molecular conformation in the mutant 2 crystal (Figure 4e).These findings implied that the latter case was more likely if sulfate ions were involved in mutant 2 protein packing.Regardless of the sulfate ions' participation in protein lattice formation, the solvent content of 30% to 29% reduction suggested that the protein molecule of mutant 2 was denser than that of mNb-WT (Table 1).

| Zinc coordination enabled a densely packed pseudo "polymorphic" crystal
Unexpectedly, mutant 7, replacing Q14 with N14, resulted in a radical change to the protein arrangement in the crystal lattice.The space group of mutant 7 turned into a P2 1 2 1 2 1 lattice from the P1 lattice in mNb-WT with only one protein molecule in the asymmetric unit (Figure 5a).The crystal morphology of mutant 7 was prismatic crystals instead of lamellae (Figure 5b).Unlike mNb-WT crystals,  With just one carbon atom shorter than the Q14 side chain, the conformation of many surface residues of the mutant 7 (i.e., Q14N) protein was completely changed in the crystal lattice (Figure 5d).
Through alignment with mNb-WT, the interaction among residues in mNb-WT at the same position as the first zinc binding site was found to be much looser, whereby D60 chose to interact with K58 in a more sprawling form and the R108 residue was forced to interact with water molecules to stabilize the side chain (Figure 5e).At the second zinc binding site position, the situation was very similar (Figure 5f).
However, at the same position of the sulfate ion binding site of mNb-WT, this did not occur.The sulfate ions were replaced by water molecules to connect these residues together in the same manner (Figure 5g,h).Although the interaction may be weaker than before, it indeed stabilized the residues to a similar extent.
Unexpectedly, the solvent content of the mutant 7 crystal was 21.92%, the minimum among the protein crystals, indicating the most compact molecular arrangement (Table 1).

| Protein release kinetics from crystals were controlled by the crystal lattice
To verify whether the crystal arrangement could affect the protein release kinetics from the crystals, we analyzed the trend of the twodimensional projected area of different crystals in the PBS (pH = 7.4) to reveal the release rate of each crystal.Limited by the tiny number of crystals of each protein, we photographed the dissolution process of each crystal in 10 μl PBS buffer at 2 min intervals.Then, we measured the two-dimensional projected area of all of the tested crystals and recorded the percentage of area reduction at each time point.
The images of the dissolution process of different crystals are shown in Figure 6a.Obviously, the size reduction rate of the protein crystals was significantly decreased to various degrees by different surface mutations, among which mutant 7 was the most effective (Figure 6b).
However, similar to the P1 lattice, the protein release of the mNb-WT, analogue 1 and 2 crystals were significantly delayed as the solvent content of the crystal decreased.For the P2 1 2 1 2 1 lattice, zinc coordination of the mutant 7 crystal further strengthened the molecular interaction, leading to a slower release rate than mutant 2 crystals.
These results suggested that the protein release rate from crystals could be remarkably extended by a more compact arrangement of protein molecules or stronger internal interaction inside the crystal lattice.Constrained by the very limited amount of protein crystals, the protein release profile reported here was a semiquantitative evaluation of protein release kinetics from different crystals.A more precise and quantitative drug release method will be developed and implemented in the future, when sufficient amount of protein crystals are prepared in the later development stage.However, the rank order of drug release kinetics is expected to be the same as what was reported here in Figure 6b.
Considering the contribution of zinc ion in the lattice packing of mutant 7 protein, we wondered whether the release rate could be further modulated by the existence of zinc ion in the release medium.
The dissolution of protein molecules from crystals should theoretically decelerate with zinc ions.Therefore, we detected the time for mutant 7 crystals to dissolve completely in release medium with a zinc ion concentration gradient.Indeed, increasing zinc ion concentration in the release medium, the dissolution time of mutant 7 crystals increased significantly (Figure 6c).By controlling the zinc ion concentration of the crystal suspension solvent, the release rate of mutant 7 crystals could be further modulated.

| The crystallization propensity of nanobodies was affected by their T m and T agg
The thermodynamic properties of proteins depend on the protein primary sequence, and the protein crystalline propensity may be directly related to the thermodynamic properties of the protein.
Therefore, we explored the possible connections between the crystallinity and the thermodynamic properties of different proteins.We recorded the crystal condition number of the mNb-WT protein and mutants under the same crystal screening batch matrix.Analogue 1 and 2 proteins exhibited a higher number of crystallization conditions; while some mutants completely lost the crystallization capability, such as mutant 1, mutant 3, mutant 4, and mutant 5 (Table 2).The above crystal screening studies were duplicated and the results showed similar crystallization propensity of these nanobodies (data not shown).
We measured the unfolding temperature (T m ) and aggregation temperature (T agg ) of mNb-WT and mutants/analogues to reveal the thermodynamic stability and aggregation propensity of each protein, respectively.The thermodynamic stability and aggregation tendency of proteins changed substantially by one or more mutations (Figure 7a-e).Every tested nanobody exhibited two unfolding events, and the protein structures were destroyed after the second unfolding event, making T m2 more responsive to protein thermodynamic stability (Figure 7b).As the beginning of crystallinity, aggregation plays a critical role in protein behavior.Small aggregates were observed at 266 nm by SLS, and large aggregates were detected by SLS at 473 nm.Small aggregates reflect the aggregation tendency of proteins, such as the protein oligomer assembly.By analyzing the T m2 and T agg (266 nm) of the tested proteins, an interesting fact was demonstrated: proteins with higher T m2 and T agg (266 nm) values around 45-50 C, appeared to crystallize more easily (Figure 7f).With a high T m value (above 61.6 C), the proteins tended to aggregate into precipitants when T agg (266 nm) was below 36.1 C; meanwhile, above 55.6 C, the aggregation tendency was too weak and proteins tended to remain in solution.

| DISCUSSION
Although the concept of using protein crystals for long acting therapeutics was proposed earlier, 32,33 and pharmaceutical superiority was observed when high concentration of protein drugs were administrated as crystalline form, 20,34,35 crystalline insulin remains the only marketed crystalline protein drug product. 36One of the major bottlenecks is that protein crystallization process is difficult to occur kinetically, and the protein crystallinity thus pharmaceutical properties were difficult to modulate.
With one or more mutations of surface residues, we successfully changed the thermodynamic properties of a model protein, leading to a shift in the crystallization propensity of pharmaceutical analogues with identical biological activities.With increasing T m and a moderate T agg value, the number of crystallization conditions of protein analogues gradually increased (Figure 7f), indicating the possibility that more crystallizable protein candidates with same efficacy could be acquired by rational mutations, and thermodynamic values such as T m and T agg could be used as readily assessable indicators for crystallization propensity.Certainly, this exciting prospect is hypothesized based on a limited sample size in this study.A concrete correlation between the crystallization properties of a protein and its primary sequence could only be established based on a much larger data library, which is currently undergoing with the assistance of machine learning.
Crystalline proteins could possess different pharmaceutical application perspectives, such as controlled release, improved stability, and so forth. 20,34,35In this study, pseudo "polymorphic" protein crystals could be obtained through rational surface mutations, subsequently, the protein release rate could be modulated by the different packing energy of the polymorphic crystal lattices.Besides, it was suggested that after protein molecules crystallize, some solvent-exposed protein surfaces could be embedded into the crystal lattice to form water inaccessible protein contact interfaces, thus could further reduce the protein release rate.Furthermore, we showed that the dissolution rate of the zinc-protein co-crystals could be further modulated by the zinc concentration in the release medium (Figure 6).On the other hand, protein crystallization could also optimize the solvent-exposed protein surface, reduce the molecular mobility, and prevent the unfolding of protein by fixing the side chains of surface residues, 18 thus collectively enhance protein stability.
It is worth noting that, various approaches could be used to further improve the chance of crystallization and crystal quality of proteins, such as assisting crystallization through other protein stabilizers or ligands, 37,38 truncating the flexible region of target proteins, 39 eliminating the post-translational protein modification, 40 introducing disulfide bonds to enhance protein symmetry into dimers; 41,42 or introducing potential metal coordination (nickel, copper, and zinc); 43 and so forth.Through such approaches, the density and lattice energy of protein crystals could be further improved.
In summary, our findings demonstrated that the protein crystalline form can be modulated by rational surface mutation without impairing pharmaceutical efficacy.Besides, the expected superior stability, 18 high protein concentration, 19 and low viscosity, 20 collectively make crystalline protein a very promising novel intravitreal drug delivery strategy for wet AMD therapy.In fact, this study also suggested that the pharmacological aspects (target binding affinity, biological activity, etc.) and the pharmaceutical aspect (crystallization propensity, formulation design, etc.) of a novel protein drug could be simultaneously considered once a protein design project is started.
With such "product-oriented drug discovery" mentality, an optimized final protein drug product could be obtained much more efficiently, compared with the traditional, step-by-step discovery and development process where various expected or unexpected challenges are to be resolved sequentially with little global and proactive designs.The tested protein sequences are displayed in Figure 2a.The model anti-VEGF nanobody was termed "mNb-WT," which binds to endogenous VEGF protein. 44Mutants 1-7 were rationally designed based on the mNb-WT nanobody.The two nanobody analogues were proprietary anti-VEGF nanobodies; thus, the complete sequences were not disclosed.

| Protein expression and purification
All of the tested proteins were expressed in Escherichia coli BL21

1
The crystal lattice and morphology of protein crystals formed by model nanobody mNb-WT.(a) The surface and cartoon model of protein arrangement in the crystal lattice.The lattice axes are shown as black arrows.The two molecules in the asymmetric unit are gray and pink, and the other asymmetric unit is shown in orange.The CDR region is shown in green, red, and blue, and the sphere model shows the sulfate ions in the structure.(b) The crystal morphology of mNb-WT.The scale bar represents 20 μm.(c) The interaction between two protein molecules inside the asymmetric unit.(d) and (e) The sulfate ion binding sites of mNb-WT.The salt bridge is colored red, and the hydrogen bonds are yellow.(f) The interaction between two asymmetric units.3(K44A/E45A mutation) were designed for the following experiments (Figure2a).To weaken the interactions at the contact interface of the mNb-WT nanobody in the P1 crystal lattice, residues at interface a and interface b (Figure1a), which were representative of all the contact interfaces, were mutated to induce different crystal lattice arrangements.Mutant 4 (Y95F mutation), mutant 5 (Q118N mutation), mutant 6 (S8A mutation), and mutant 7 (Q14N mutation) were designed for crystal lattice comparison.With similar bioactivity to the mNb-WT nanobody, two humanized anti-VEGF nanobodies with mutation series, analogues 1 and 2, were tested together (Figure2a).To remove the influence of protein posttranslational modifications, such as glycosylation and phosphorylation, all proteins were expressed in Escherichia coli without any purifying tag so that any of the discrepancies were induced by surface mutations of the mNb-WT nanobody.

mutant 7 31 F I G U R E 3
recruited zinc ions rather than sulfate ions to help align protein molecules in the crystal lattice.Every mutant 7 molecule interacts with another asymmetric unit mediated by two zinc ions.At the first zinc binding site, the side chain of the D60 residue in asymmetric unit A, the side chain of the D111 residue in asymmetric unit B, and two water molecules formed a strong coordination unit with the zinc ion; meanwhile, the two water molecules also created typical hydrogen bonds with the D60 and R108 residues of asymmetric unit A and the D111 residue of asymmetric unit B to strengthen the periphery of the coordination unit (Figure 5c, top diagram).The other zinc coordination unit consisted of the side chains of residues D73 and K76 in asymmetric unit A, the side chain of residue E45 in asymmetric unit C, and one water molecule.Similar to F I G U R E 2 The sequence and pharmacological properties of rationally engineered mutants (mutants 1-7)/analogues (analogues 1-2) of mNb-WT.(a) Rational mutation design based on the mNb-WT crystal structure.(b) The vascular endothelial growth factor (VEGF) binding ability of mNb-WT and mutants/analogues.(c) The normalized SEC curves of mNb-WT and mutants/analogues.(d) The circular dichroism (CD) curves of mNb-WT and mutants/analogues.(e) The percentage of the secondary structure of mNb-WT and mutants/analogues in (d).No significance among the tested groups.(f) The alignment model of mNb-WT and rational mutations predicted by AlphaFold.F I G U R E 3 Legend on next page.the previous coordination unit, the hydrogen bonds formed among water molecule and the three key residues enhanced the stability of the coordination unit (Figure 5c, bottom diagram).Interestingly, these two coordination units were identical not to the conventional zinc binding motif, the classical zinc finger motif 29 or histidinemediated zinc coordination among protein molecules, 30 but rather to the central catalytic core of metal-binding proteins, such as retroviral integrases.The crystal lattice and morphology of analogues 1 and 2. (a) Comparison of the protein arrangement in the crystal lattice of mNb-WT and analogue 1.The lattice axes are shown as black arrows.(b) The crystal morphology of analogue 1.The scale bar represents 20 μm.(c) The conformational change of residues at the surface of mNb-WT and analogue 1.The two protein molecules A and B of asymmetric units of mNb-WT are shown in gray and pink, respectively; analogue 1 molecules are shown in yellow and cyan; CDR1, CDR2, and CDR3 are shown in green, red, and blue, respectively.(d, e) Comparison of the sulfate ion binding sites of mNb-WT and analogue 1. (f) The protein arrangement in the crystal lattice of mNb-WT and analogue 2. The lattice axes are shown as black arrows.(g) The crystal morphology of analogue 2. The scale bar represents 20 μm.(h) The conformational change of surface residues of mNb-WT and analogue 2. The two protein molecules A and B of asymmetric units of mNb-WT are shown in gray and pink, respectively; analogue 2 molecules are shown in cyan and purple; CDR1, CDR2, and CDR3 are shown in green, red, and blue, respectively.(i, j) Comparison of the sulfate ion binding sites of mNb-WT and analogue 2.

F I G U R E 4
The crystal lattice alignment of mNb-WT and mutant 2. (a) The protein arrangement in the crystal lattice of mNb-WT and mutant 2. The crystal axes are shown as black arrows.(b) The crystal morphology of the mutant 2 crystal.(c) The conformational change of residues at the surface of mNb-WT and mutant 2. The two protein molecules A and B of the asymmetric unit of mNb-WT are shown in gray and pink, respectively; mutant 2 is shown in purple; CDR1, CDR2, and CDR3 are shown in green, red, and blue, respectively; and the mutation site is shown in green.(d, e) The conformational change of the surface residues at sulfate ion binding sites.F I G U R E 5 The crystal lattice arrangement of mNb-WT and mutant 7. (a) The protein arrangement in the crystal lattice of mNb-WT and mutant 7.The lattice axes are shown as black arrows.The two protein molecules A and B of mNb-WT are shown in gray and pink, respectively; mutant 7 is shown in lime-green, and CDR1, CDR2, and CDR3 are shown in green, red, and blue, respectively.(b) The crystal morphology of mutant 7.The scale bar represents 20 μm.(c) The zinc coordination units in the mutant 7 lattice.The zinc-binding residues are shown in green, yellow and cyan for different molecules, yellow dashed lines indicate dipolar bonds and hydrogen bonds at the zinc-binding unit, and water is shown as a red sphere.(d) The conformational change of the whole surface residues of mNb-WT and mutant 7 molecules.The mutation sites are colored cyan.The details of residues at zinc binding sites of mNb-WT and mutant 7 are shown in (e) and (f).The residues of mNb-WT are shown in gray, and the green, cyan and orange residues belonged to mutant 7, of which orange residues were key residues with zinc binding in mutant 7 crystal.The red spheres are the waters in the mutant 7 lattice, and the blue spheres are in the mNb-WT lattice.(g, h) Comparison of mNb-WT and mutant 7 of the sulfate ion binding site.

F I G U R E 6
Protein release kinetics from different crystals.(a) Images of the dissolution process of different crystals.The scale bar represents 20 μm.(b) The percentage size change of the protein crystals relative to the initial size.The error bar is the standard deviation (SD).(c) The dissolution time of mutant 7 crystals in a linear gradient of zinc ions.The error bar is the SD.The p values are labeled in the figure, and the n numbers of the 20, 60, 100, 150, 200, 250, and 300 mM zinc ion groups were 5, 5, 6, 5, 7, 7, and 5, respectively.

T A B L E 2
The crystal propensity of mNb-WT and mutations S3: Crystal Screen ½ kit, Hampton Research.S6: Index kit, Hampton Research.S10: JCSG plus kit, Molecular Dimensions.S20: JBScreen Basic HTS kit, Jena.S21: JBScreen Kinase HTS kit, Jena.S22: JBScreen Classic HTS I kit, Jena.S23: JBScreen Classic HTS II kit, Jena.The red numbers are crystallization condition numbers after combining the same condition components.

F I G U R E 7
Relationship between thermodynamic properties and crystallization propensity.(a) The T m curves of mNb-WT and mutants 1-7 and analogues 1-2.(b) The T m values of a. (c) The aggregation curves of tested proteins at 266 nm.(d) Aggregation curves of the tested proteins at 473 nm.(e) The T agg values of (c) and (d).(f) The relationship between crystal numbers, T m2 and T agg (266 nm).The tested sample was labeled as red circles.

(
DE3) competent cells to avoid the potential influence of posttranslational modification.The expression and purification procedure of all of the tested proteins were similar.The bacteria with constructed plasmids were cultured in lysogeny broth medium with 100 μg/ml ampicillin at 37 C until the OD 600 reached 0.6, and isopropyl β-D-1-thiogalactopyranoside (InalcoPharm., 1758-1400) at a final concentration of 1 mM was employed to induce the expression of target nanobodies.At 6 h postincubation, the expressed target protein was enriched as inclusions in the cytoplasm of E. coli.The bacteria were centrifuged at 4000 rpm/min for 10 min and resuspended in Buffer A (20 mM Tris-HCl, pH = 8.0, 150 mM NaCl).

1
Crystal lattice comparison of mNb-WT and mutations/analogues