DNA-Directed Protein Packing within Single Crystals

SUMMARY

Designed DNA-DNA interactions are investigated for their ability to modulate protein packing within single crystals of mutant green fluorescent proteins (mGFPs) functionalized with a single DNA strand (mGFP-DNA). We probe the effects of DNA sequence, length, and protein-attachment position on the formation and protein packing of mGFP-DNA crystals. Notably, when complementary mGFP-DNA conjugates are introduced to one another, crystals form with nearly identical packing parameters, regardless of sequence if the number of bases is equivalent. DNA complementarity is essential, because experiments with non-complementary sequences produce crystals with different protein arrangements. Importantly, the DNA length and its position of attachment on the protein markedly influence the formation of and protein packing within single crystals. This work shows how designed DNA interactions can be used to influence the growth and packing in X-ray diffraction quality protein single crystals and is thus an important step forward in protein crystal engineering.

Graphical Abstract

Protein single crystals are consequential materials for discovering protein structure and function; however, the size and complexity of proteins can make protein crystallization difficult. Furthermore, it is challenging to direct the organization of proteins within crystals. Herein, a DNA modification to the surface of green fluorescent protein influences the organization of this protein within a single crystal. Moreover, protein packing within the protein-DNA crystals is varied substantially with the design of the DNA modification.

INTRODUCTION

Through X-ray crystallography, protein single crystals enable fundamental understanding of protein structure and recognition^1–5 and consequently have been important in the rational design of drugs.^6,7 In addition, they have been used in chiral catalysis⁸ and enantiomeric separations,⁹ and non-crystalline but ordered protein assemblies have been utilized to control cascade reactions.^10–12 However, protein crystallization is challenging because proteins are complex, dynamic molecules comprising thousands of atoms.¹³ Furthermore, the interactions between protein surfaces that drive crystallization are weak, complex, and noncovalent; therefore, researchers interested in such structures have little control over crystallization and the type of crystals that are formed.^14,15

Efforts to control protein crystallization have included modifications that affect charge,^16–18 hydrophobicity,^7,19 protein structure,^20–22 ligand binding,^23–26 and metal binding characteristics,^27–29 and they often involve the introduction of functional groups via site-directed mutagenesis.^30,31 In 2015, we introduced the concept of DNA-modification to control protein crystallization.³² With isotropically and sometimes anisotropically functionalized structures, pseudo-crystalline materials could be realized, but to date, these techniques, in our hands or others, have not yielded structures suitable for single-crystal X-ray diffraction studies.^32–35

Based upon our previous work,^{32–34,36–39} we hypothesized that modifying proteins with single strands of DNA could be used to influence crystallization, and when combined with protein-protein interactions, could yield new crystal forms and atomic resolution structures. Herein, we explore this hypothesis with a model protein, mutant green fluorescent protein (mGFP). The effects of design parameters, including DNA sequence, DNA length, protein amino acid attachment position, and DNA base attachment position were systematically explored with respect to their consequence on protein packing in the crystals (Figure 1). Importantly, for many of the systems studied, X-ray diffraction quality single crystals could be obtained (Table S3), and an elucidation of the resulting structures provided insight into the design parameters that control protein packing within such crystals. Taken together, the data demonstrate that a single DNA modification on the surface of a protein can be used to direct protein packing within a single crystal and, as such, is an important step forward in protein crystal engineering.

Figure 1. Design and Parameter Scope of mGFP-DNA Conjugates Studied.

(A) Schematic of the DNA interaction between mGFP-DNA conjugates with dimensions for the mGFP, the DNA, and the mGFP-DNA linkage.

(B) The design parameters explored include DNA sequence, DNA length, amino acid attachment position, and DNA base attachment position. DNA sequence was varied between self-complementary (scDNA), complementary (cDNA), and non-complementary (ncDNA) (upper left). DNA length was varied between 6 and 18 base pairs (upper right). DNA attachment positions were on the side (residue 148) or edge (residue 176 or 191) of the mGFP β-barrel (lower left). The sites within the DNA for attachment to the proteins were either internal or external (lower right).

RESULTS

To study how designed DNA interactions can influence the growth and packing in protein single crystals, we designed GFP mutants that could be modified with one DNA strand using cysteine-conjugation methods. A single cysteine residue was positioned at a distinct surface location on all mutants, either on the side (C148 mGFP)³³ or the edge (C176 mGFP or C191 mGFP) of the mGFP β-barrel (Table S1). Crystal structures of C148 mGFP and C176 mGFP were determined (a structure of C191 mGFP is known)⁴⁰ prior to their functionalization with DNA as comparisons to structures obtained when DNA is present. Although crystal structures of native GFP are well known,⁴¹ the position of solvent-accessible cysteine residues on mGFP influences protein packing through the formation of disulfide bonds.⁴⁰ The C148 mGFP was crystallized, and a 1.5 Å structure, in which C148 remains as a thiol was determined in the space group P2₁2₁2₁ (Figures 2A and S7 and Table 2; PDB: 6UHJ). The structure is nearly identical to the majority of GFP structures in the Protein Data Bank (PDB),⁴² with nearly equivalent unit cell parameters and a root-mean-square deviation (rmsd) of 0.2 Å for all atoms from the GFP structure PDB: 4EUL.⁴¹ Crystals of C176 mGFP were characterized where C176 form disulfide bonds (product of oxidation) as a different structure in the space group I222 at 1.9 Å resolution (Figures S9 and S10 and Table 2; PDB: 6UHK).

Figure 2. mGFP-DNA Single-Crystal Structures.

(A) A model of C148 mGFP (top). Four asymmetric units of the C148 mGFP crystal structure (PDB: 6UHJ) in the space group P2₁2₁2₁ (bottom), which is equivalent to previously reported GFP crystal structures (C148 residues represented in blue).⁴¹ Proteins pack densely in this structure, and C148 is involved in an inter-protein interaction.

(B) A model of the C148 mGFP-scDNA-1 design (top). Two asymmetric units of the C148 mGFP-scDNA-1 crystal structure (PDB: 6UHL) in the space group P12₁1 (bottom). ^aC148 mGFP-cDNA-1 and C148 mGFP-cDNA-2 crystallize into nearly identical structures (PDB: 6UHN and 6UHO; Figure S20). In these structures, the DNA does not order past the disulfide mGFP-DNA attachment (inset). Pairs of C148 (blue) orient toward distinct regions of solvent space with a C148-C148 distance of 37 ± 4 Å that is within the theoretical distance for DNA hybridization (27–64 Å).

(C) A model of the C148 mGFP-ncDNA-1 design (top). Two asymmetric units from the C148 mGFP-ncDNA-1 crystal structure (PDB: 6UHP) in the space group P12₁1 (bottom), where each C148 (orange) orients toward distinct regions of solvent space with no free path between C148 residues that would permit DNA hybridization (Figure S23).

Table 2.

Summary of mGFP and mGFP-DNA Crystal Properties

Sample	PDB Code	Space Group	Proteins in Asymmetric Unit	Cell Parameters (Å)	Cell Parameters (°)	Resolution (Å)
C148 mGFP	6UHJ	P212121	1	51.51, 62.90, 69.40	90.00, 90.00, 90.00	1.50
C176 mGFP	6UHK	I222	2	88.93, 91.76, 151.71	90.00, 90.00, 90.00	1.90
C148 mGFP-scDNA-1	6UHL	P1211	2	64.87, 52.29, 86.80	90.00, 94.13, 90.00	1.91
C148 mGFP + scDNA-1	6UHM	P212121	2	58.28, 61.76, 135.32	90.00, 90.00, 90.00	2.10
C148 mGFP-cDNA-1	6UHN	P1211	2	64.92, 52.18, 86.47	90.00, 94.24, 90.00	1.92
C148 mGFP-cDNA-2	6UHO	P1211	2	64.71, 52.22, 86.44	90.00, 94.23, 90.00	1.95
C148 mGFP-ncDNA-1	6UHP	P1211	2	59.05, 51.60, 100.41	90.00, 106.97, 90.00	2.90
C148 mGFP-cDNA-3	6UHQ	C121	1	106.63, 50.58, 56.69	90.00, 110.33, 90.00	2.85
C148 mGFP-scDNA-2	6UHR	P212121	2	50.58, 50.89, 209.19	90.00, 90.00, 90.00	3.00

Next, we investigated the effect of introducing a designed DNA interaction between proteins on crystallization and protein packing within a single crystal. We varied design parameters including DNA sequence, DNA length, amino acid attachment position, and DNA base attachment position. In a typical experiment, the surface cysteine on mGFP (Figure S1) was functionalized with pyridyl-disulfide-modified DNA (mGFP-DNA, Table S2) through a thiol-disulfide exchange reaction according to previously published procedures (Supplemental Information and Experimental Procedures).³³ Unreacted DNA and protein were removed using nickel affinity and anion exchange chromatography, respectively. Mono-functionalization of mGFP with DNA and purification of mGFP-DNA conjugates were confirmed using ultraviolet-visible (UV-vis) spectroscopy, SDS-PAGE, and matrix-assisted laser desorption ionization mass spectrometry (MALDI-MS) (Figures S6, S8, S11, S12, S16, S18, S21, S24, and S26–S31). These data conclusively demonstrate the attachment of DNA to mGFP and the purification of the mGFP-DNA conjugates. The mGFP-DNA conjugates were crystallized using vapor diffusion techniques, and hundreds of crystallization conditions (varying salt, precipitant, buffer, and temperature) were screened robotically in a high-throughput manner. The protein packing within each single crystal was characterized by X-ray crystallography for structure determination (see Tables 1,2, and S4-S9).

Table 1.

Sample Designs for mGFP-DNA Conjugates Used for Studying the Effect of DNA Sequence, DNA Length, Amino Acid Attachment Position, and DNA Base Attachment Position on Protein Packing within Single Crystals

Line Number	Sample	PDB Code	mGFP Mutant	DNA Design	Study
1	mGFP	6UHJ	C148	n/a	control
2	mGFP	6UHK	C176	n/a	control
3	mGFP	n/a41	C191	n/a	control
4	mGFP-scDNA-1a	6UHL	C148	H2N-CGCGCG	DNA sequence DNA length amino acid attachment position DNA base attachment position
5	mGFP + scDNA-1	6UHM	C148	H2N-CGCGCG	control
6	mGFP-cDNA-1b	6UHN	C148	H2N-GGCCGG, H2N-CCGGCC	DNA sequence
7	mGFP-cDNA-2	6UHO	C148	H2N-AGAGAG, H2N-CTCTCT	DNA sequence
8	mGFP-ncDNA-1c	6UHP	C148	H2N-TTTTTT	DNA sequence
9	mGFP-cDNA-3	6UHQ	C148	H2N-AAGGAAGGA, H2N-TCCTTCCTT	DNA length
10	mGFP-cDNA-4	did not crystallize	C148	H2N-AAGGAAGGAAGG, H2N-CCTTCCTTCCTT	DNA length
11	mGFP-cDNA-5	did not crystallize	C148	H2N-AGTTAGGACTTACGCTAC, H2N-GTAGCGTAAGTCCTAACT	DNA length
12	mGFP-ncDNA-2	did not crystallize	C148	H2N-TTTTTTTTT	DNA length
13	mGFP-scDNA-1	did not crystallize	C176	H2N-CGCGCG	amino acid attachment position
14	mGFP-scDNA-1	did not crystallize	C191	H2N-CGCGCG	amino acid attachment position
15	mGFP-scDNA-2	6UHR	C148	GCGCT(NH2)AGC	DNA base attachment position

^aSelf-complementary DNA (scDNA).

^bComplementary DNA (cDNA)

^cNon-complementary DNA (ncDNA).

An mGFP-DNA Single-Crystal Structure

As a proof of concept that DNA interactions can modify the growth and packing of protein single crystals, we first studied the crystallization of mGFP modified with a 6 base pair (bp) self-complementary DNA strand (scDNA-1) at the C148 position (mGFP-scDNA-1, Table 1: line 4). DNA conjugation did not inhibit the protein’s ability to crystallize, as the mGFP-scDNA-1 conjugate crystallized into thin plates (~100 × 200 × 10 mm). Significantly, a 1.9 Å resolution crystal structure in the space group P12₁1 was determined (Figures 2B and S13 and Table 2; PDB:6UHL). Furthermore, the structure has different unit cell parameters and protein packing with respect to the C148 mGFP crystal structure, indicating that the DNA modification plays a role in how the proteins are organized. In fact, the unit cell parameters and protein packing in the mGFP-scDNA-1 crystal are different relative to all previously reported GFP crystal structures. The crystal structure shows electron density for mGFP and the disulfide mGFP-scDNA-1 attachment, but not DNA. The flexibility of the linker used for protein conjugation (see Figure S2 for linkage structure) likely prevents DNA from ordering in the crystal. However, the mGFP-scDNA-1 protein packing is consistent with the presence of hybridized DNA. Pairs of C148 residues orient toward distinct regions of solvent space and are separated by 37 ± 4 Å, a distance that corresponds well with the length of the duplexed DNA within the protein single crystals (theoretical distance for a 6 bp duplex DNA, including the two alkyl linker molecules in the contracted or extended form, is 27–64 Å).⁴³ As an additional control experiment to confirm that covalent attachment of scDNA-1 to mGFP directs the mGFP-scDNA-1 crystal structure, a physical mixture of C148 mGFP and scDNA-1 was subjected to identical crystallization conditions as the conjugate (Table 1: line 5). The crystals resulting from the physical mixture show a structure with a disulfide bond between surface cysteines (Figures S14 and S15 and Table 2; PDB: 6UHM), where mGFP packing is exclusively directed by inter-protein interactions. Taken together, these results show that the covalent attachment of a 6 bp scDNA strand to mGFP leads to a change in protein-protein contacts during crystallization and, ultimately, different protein packing.

DNA Hybridization Directs mGFP-DNA Packing

To explore whether DNA-directed protein packing using complementary strands is independent of specific sequence, two sets of 6 bp complementary DNA strands were designed (cDNA-1 and cDNA-2, Table 1: lines 6 and 7). The C148 mGFP was functionalized with the cDNA sequences separately, and then corresponding mGFP-DNA conjugates were mixed immediately prior to subjecting the mixture to crystallization experiments. Both mGFP-cDNA-1 and mGFP-cDNA-2 crystallized into thin plates, showing the same crystal morphology as mGFP-scDNA-1 crystals. Furthermore, 1.9 Å crystal structures for mGFP-cDNA-1 and mGFP-cDNA-2 have the same space group P12₁1 and nearly equivalent unit cell parameters as the mGFP-scDNA-1 structure (Figures 2B, S17, and S19 and Table 2; PDB: 6UHN and 6UHO, respectively). The rmsd between mGFP-scDNA-1, mGFP-cDNA-1, and mGFP-cDNA-2 structures are less than 0.2 Å for all atoms, confirming that the protein packing of these structures is essentially equivalent (Figure S20). Therefore, (self-)complementary mGFP-DNA conjugates with a DNA length of 6 bp crystallize into practically identical single-crystal forms, regardless of DNA sequence.

Next, we wanted to confirm the importance of DNA complementarity on the resulting crystal structure. The C148 mGFP was functionalized with a T6 non-complementary DNA strand (mGFP-ncDNA-1, Table 1: line 8) and crystallized. The mGFP-ncDNA-1 conjugates formed needle-like crystals, a distinct crystal morphology from mGFP and the three 6 bp (self-)complementary mGFP-DNA conjugates. Moreover, a 2.9 Å resolution crystal structure in the space group P12₁1 was determined for mGFP-ncDNA-1 with unit cell parameters and protein packing that are different from both those of mGFP and (self-)complementary mGFP-DNA conjugates (Figures 2C and S22 and Table 2; PDB: 6UHP). Clearly, the presence of non-complementary single-stranded DNA still influences packing outcomes of mGFP, likely by filling space and altering the crystal contacts that may form between neighboring mGFP proteins in the crystal. However, the protein packing in the mGFP-ncDNA-1 structure is not consistent with DNA duplexing, as each C148 residue orients toward a different region of solvent space with no free path between C148 residues that would permit DNA hybridization (Figure S23). This result indicates the importance of DNA complementarity on protein packing outcomes in protein-DNA crystals and illustrates that protein packing within single crystals (mGFP-scDNA-1, mGFP-cDNA-1, and mGFP-cDNA-2) can be directed using programmable DNA interactions.

Because no direct evidence of electron density for DNA was observed in the electron density maps for the mGFP-DNA crystals structures, to confirm the presence of the DNA, crystals were incubated with the DNA-intercalating dye TOTO-3 and imaged using confocal microscopy. TOTO-3 is a cationic, DNA duplex-sensitive dye that shows a several thousand-fold increase in fluorescence upon DNA intercalation due to decreased rotational freedom, which enforces a planar conformation.^44,45 Before dye addition, crystals of C148 mGFP, C148 mGFP-ncDNA-1, and C148 mGFP-cDNA-1 showed mGFP fluorescence (485 nm excitation and 500–550 nm emission filter), but no TOTO-3 fluorescence (640 nm excitation and 663–738 nm emission filter) (Figures S3–S5). When TOTO-3 was added to crystals of C148 mGFP, as expected, no TOTO-3 fluorescence was observed because the mGFP crystals do not contain DNA (Figure 3A). In contrast, a strong TOTO-3 fluorescence was observed for mGFP-ncDNA-1 (Figure 3B) and mGFP-cDNA-1 crystals (Figure 3C), providing evidence for the presence of DNA within the crystals of mGFP-ncDNA-1 and mGFP-cDNA-1. Surprisingly, no significant difference in the ratio of mGFP to TOTO-3 fluorescence was observed between mGFP-ncDNA-1 and mGFP-cDNA-1 crystals (Figure 3D). Although TOTO-3 is duplex-sensitive in solution, the behavior of TOTO-3 in protein crystals is less understood.^44,45 In our case, it is possible that TOTO-3 dye could interact with confined single-stranded DNA in protein crystals in a way that enforces planarity and induces fluorescence. Overall, the evidence for the presence of DNA in mGFP-ncDNA and mGFP-(s)cDNA crystals from the microscopy experiment, when combined with crystallographic evidence that DNA complementarity determines crystallization outcomes, shows that protein packing in single crystals can be modulated by DNA hybridization interactions.

Figure 3. Confocal Microscopy Evidence for DNA in C148 mGFP-(s) cDNA and C148 mGFP-ncDNA Crystals.

(A–C) Confocal microscopy images of (A) C148 mGFP, (B) C148 mGFP-ncDNA-1, and (C) C148 mGFP-cDNA-1 crystals after soaking the crystals for 30 min in the intercalating dye, TOTO-3. The images are in bright field (left), a green channel (middle, 485 nm excitation and 500–550 nm emission filter), and a far-red channel (right, 640 nm excitation and 663–738 nm emission filter). All scale bars represent 50 μm.

(D) The ratio of green to far-red fluorescence signals was compared across multiple crystals. The lower signal ratio in C148 mGFP-ncDNA-1 and C148 mGFP-cDNA-1 crystals compared with C148 mGFP crystals indicates the presence of DNA in C148 mGFP-ncDNA-1 and C148 mGFP-cDNA-1 crystals.

DNA Interaction Length Influences mGFP-DNA Packing

Because complementary DNA interactions can direct protein crystallization, we sought to determine if DNA length provides another parameter for affecting crystal packing arrangements. To investigate the effect of DNA interaction length on crystallization outcome, DNA interactions at various lengths (6, 9, 12, and 18 bp) were designed, and mGFP-DNA conjugates incorporating these interactions were synthesized. Although a single-crystal form was observed for the three 6 bp DNA duplexes discussed above, an increase in DNA duplex length to 9 bp (mGFP-cDNA-3, Table 1: line 9) led to a 2.9 Å structure in the space group C121 (Figures 4A and S25 and Table 2; PDB: 6UHQ). The protein packing within this structure is distinct from other mGFP-DNA structures and, importantly, pairs of C148 residues again orient toward distinct regions of solvent space, separated by 41 ± 6 Å, a distance that agrees with the length of the duplexed DNA (theoretical distance for a 9 bp duplex, including the two alkyl linker molecules in the contracted or extended form, is 37–75 Å). However, when longer DNA ligands (12 bp, mGFP-cDNA-4, Table 1: line 10; and 18 bp, mGFP-cDNA-5, Table 1: line 11) were investigated, no crystallization was observed. This suggests that above an upper threshold for DNA duplex length, DNA prevents the formation of single crystals. Similarly, increasing the length of ncDNA from 6 to 9 bases (mGFP-ncDNA-2, Table 1: line 12) precluded crystallization. Taken together, mGFP-DNA crystallization and structural outcomes depend strongly on the length of the designed DNA.

Figure 4. DNA Design Influences mGFP-DNA Packing.

(A) A model of the C148 mGFP-cDNA-3 design (top). Four asymmetric units of the C148 mGFP-cDNA-3 crystal structure (PDB: 6UHQ) in the space group C121 (bottom). Increasing DNA duplex length by 3 bp led to this structure. Pairs of C148 (red and purple) orient toward distinct regions of solvent space with a C148-C148 distance of 41 ± 6 Å that is within the theoretical distance for DNA hybridization (37–75 Å).

(B) A model of the C148 mGFP-scDNA-2 design (top). Two asymmetric units of the C148 mGFP-scDNA-2 crystal structure (PDB: 6UHR) in the space group P2₁2₁2₁ (bottom). Changing the location of mGFP-DNA attachment position led to this structure. Pairs of C148 (blue) orient toward distinct regions of solvent space with a C148-C148 distance of 30 ± 6 Å that is within the theoretical distance for DNA hybridization (8–45 Å).

Protein-DNA Attachment Position Influences mGFP-DNA Packing

In addition to exploring how DNA design can influence crystal structures, protein-DNA attachment position represents another powerful design parameter, where changing attachment location can guide protein-protein interactions and therefore protein packing. The amino acid attachment position was varied by changing the location of the cysteine more than 15 Å from the middle of the side of the mGFP β-barrel (C148 mGFP) to the edge of the mGFP β-barrel (C176 mGFP and C191 mGFP). The C176 mGFP and C191 mGFP were functionalized with scDNA-1 (C176 mGFP-scDNA-1, Table 1: line 13; and C191 mGFP-scDNA-1, Table 1: line 14), the same DNA which directed the crystallization and structure of C148 mGFP-scDNA-1. In contrast, C176 mGFP-scDNA-1 and C191 mGFP-scDNA-1 conjugates did not crystallize, perhaps due to the high flexibility of loops at the edge of the mGFP β-barrel. Moreover, residues at position 176 and 191 are located at interfaces in the C148 mGFP crystal structure, potentially indicating their involvement in crystal packing contacts. These results exhibit the importance of amino acid attachment position on crystallization outcomes.

Next, DNA base attachment position was changed from an external to an internal DNA base, which allows shorter inter-protein distances. Additionally, DNA strands with an internal base attachment position may be designed with short sticky end overhangs, which can lead to DNA ordering in single crystals.^46,47 The C148 mGFP was functionalized with a 6 bp scDNA strand with a 2 base sticky end (C148 mGFP-scDNA-2, Table 1: line 15), and this conjugate crystallized into a crystal form in the space group P2₁2₁2₁ (Figures 4B and S32 and Table 2; PDB: 6UHR). Similar to other mGFP-DNA crystal structures, DNA did not order, obscuring the effect of the sticky ends on crystal formation. That said, pairs of cysteines orient toward distinct regions of solvent space at a distance (30 ± 6 Å) that agrees with the length of the duplexed DNA (theoretical distance for an 8 bp duplex with internal attachment position is 8–45 Å), further confirming that DNA interactions can be extensively designed to influence the crystallization and packing of proteins. This structure suggests an additional layer of control provided by the DNA ligand and will be the subject of future investigations involving linker flexibility and sticky end design.

Effect of DNA on Interfaces between Proteins in Single Crystals

Because different crystal structures were observed depending on the presence and design of DNA, we sought to understand the discrete changes in protein-protein interactions across observed single crystals. Toward this end, interfaces between mGFP in all obtained crystal structures were analyzed using PDBePISA to study the effect of introduced DNA on protein packing (Figures S33–S35 and Tables S10–S21).⁴⁸ Although some new interfaces arose, crystals of C148 mGFP-DNA show a protein-protein interface with 82%−100% amino acid residue similarity to the interface with the largest area observed in C148 mGFP (Table S19). Importantly, this shows that despite dramatic changes in the overall protein packing, the largest protein-protein interface is conserved across all C148 mGFP and C148 mGFP-DNA crystals (Figure S33). The second largest interface in crystals of C148 mGFP is also conserved for crystals of C148 mGFP-scDNA-1, C148 mGFP-cDNA-1, and C148 mGFP-cDNA-2 (89%, 100%, and 73% similar, respectively) and partially conserved for crystals of C148 mGFP-ncDNA-1 (50% similar) (Figure S35 and Table S21). Interestingly, this interface is not observed with increases in DNA length or changes in DNA base attachment position, indicating that the nature of protein packing is dictated by an interplay between protein-protein and DNA hybridization interactions. This suggests we could use existing protein crystal structures to inform the placement of DNA ligands to engineer protein packing within single crystals.

DISCUSSION

Growth of protein single crystals involves complex protein-protein interactions that are challenging to design and predict. This work demonstrates how replacing such interactions with highly programmable DNA interactions could enable structural control over protein packing within single crystals. We report the first protein single-crystal structure where DNA hybridization interactions between the surfaces of proteins direct the packing of proteins within the crystal. Furthermore, we demonstrate that DNA complementarity, DNA length, and protein-DNA attachment position all influence crystallization and protein-packing structural outcomes. Although the resulting crystal structure is shown to be independent of DNA sequence (while maintaining complementarity), crystallization only occurred for mGFP-DNA conjugates when DNA duplexes were less than or equal to 9 bp. Interestingly, changing the DNA length or the attachment of DNA to the protein through an internal base modification afforded more crystal structures that further demonstrate the versatility of this approach and the large design space to be explored. An analysis of protein–protein interfaces within mGFP and mGFP-DNA crystals reveals that interfaces may be conserved, regardless of overall crystal structure. Taken together, the work presented herein is an essential step toward designing and engineering protein packing within single crystals and could lead to future applications in protein structure determination and functional protein crystal materials.

EXPERIMENTAL PROCEDURES

See Supplemental Information for materials and methods.

HIGHLIGHTS.

Single-protein crystallization of protein-DNA conjugates

DNA modification design influences protein packing in protein-DNA crystals

Protein-protein interfaces can be conserved from protein-only to protein-DNA crystals

The Bigger Picture.

Uncovering protein structure and function typically relies on protein crystals. However, proteins are large, complex, and dynamic molecules that can be difficult to crystallize. Furthermore, controlling protein organization within crystals is challenging but could allow engineering of materials that exploit highly specific protein functions. Herein, we modify green fluorescent protein (GFP) packing within single crystals through the addition of DNA on the GFP surface. We demonstrate that DNA interaction, length, and attachment location dramatically affect GFP organization within crystals. Importantly, DNA functions together with protein-protein interactions, providing a potential route to predictably manipulate protein packing with DNA. These findings suggest that judicious DNA design and attachment onto proteins may direct protein crystallization and dictate how proteins are arranged in crystals, ultimately facilitating the discovery and application of protein structure and function.