Improving the diffraction quality of heat-shock protein 47 crystals

A two-pronged strategy was embarked upon to obtain well diffracting crystals of heat-shock protein 47 by (i) replacing a noncleavable C-terminal His-tag with a cleavable N-terminal His-tag and (ii) developing Adnectin crystallization chaperones. Both approaches yielded better diffracting crystals, and the second approach limited the number of complexes in the asymmetric unit to only one or two, which made for less onerous model building.

Abstract

Heat-shock protein 47 (HSP47) is a potential target for inhibitors that ameliorate fibrosis by reducing collagen assembly. In an effort to develop a structure-based drug-design system, it was not possible to replicate a previous literature result (PDB entry 4au4) for apo dog HSP47; instead, crystal forms were obtained in which pairs of dog HSP47 molecules interacted through a non­cleavable C-terminal His-tag to build up tetramers, all of which had multiple molecules of HSP47 in the asymmetric unit and none of which diffracted as well as the literature precedent. To overcome these difficulties, a two-pronged approach was followed: (i) the His-tag was moved from the C-terminus to the N-terminus and was made cleavable, and (ii) Adnectin (derived from the tenth domain of human fibronectin type III) crystallization chaperones were developed. Both approaches provided well diffracting crystals, but the latter approach yielded crystal forms with only one or two HSP47 complexes per asymmetric unit, which made model building less onerous.

1. Introduction

Heat-shock protein 47 (HSP47) is a serpin (gene name SERPINH1) that functions as an endoplasmic reticulum (ER)-resident, procollagen-specific chaperone. HSP47 is stress-inducible and binds to unstable triple-helical pro­collagen, preventing its unfolding and aggregation in the ER (Ito & Nagata, 2017 ▸). While several molecular chaperones are involved in the biosynthesis of collagen type I (Claeys et al., 2021 ▸), HSP47 is indispensable for triple-helix formation and collagen secretion (Nagai et al., 2000 ▸).

Collagen is the most abundant mammalian protein and is an important component of the extracellular matrix (ECM). Several studies have implicated HSP47 and its role in collagen assembly in the establishment of a proper ECM. Mutations in HSP47 have been linked to several ECM diseases such as osteogenesis imperfecta (Drögemüller et al., 2009 ▸), and knockout mutations in mice have shown that functional HSP47 is required during embryogenesis (Nagai et al., 2000 ▸). The abnormal accumulation of collagen in the ECM is characteristic of fibrosis, which can result in impaired organ function. Inhibition of HSP47 function is anticipated to result in a reduction in collagen secretion; therefore, HSP47 may be a therapeutic target for reducing fibrosis and the treatment of fibrotic diseases. Reducing levels of HSP47 or its activity might also be important in reducing venous thromboembolism in patients who are immobilized for short periods (Thienel et al., 2023 ▸).

The structure of dog HSP47 has been determined in the apo form (PDB entry 4aua) and with several different collagen model peptides (PDB entries 3zha, 4au2, 4au3, 4axy, 7bee, 7bdu and 7bfi; Widmer et al., 2012 ▸; Abraham et al., 2021 ▸). Despite using a published protocol for the growth of apo dog HSP47 crystals, we encountered several challenges in obtaining crystals. These included the precipitation of dog HSP47 with a C-terminal His-tag at the concentrations described and an overall difficulty in getting the protein to crystallize.

This paper describes (i) three new apo crystal forms that we grew using the protein construct described by Widmer et al. (2012 ▸) that all show association through the noncleavable C-terminal His-tag, (ii) our efforts with a new protein construct that produced well diffracting crystals that were useful for structure-based drug design and (iii) a co-crystallization strategy of using HSP47 with Adnectin crystallization chaperones that successfully generated two high-resolution co-crystal structures of HSP47. Adnectins (a trademark of Bristol Myers Squibb) are small, antibody-like proteins based on the tenth domain of human fibronectin type III (¹⁰Fn3) that are selected to bind targets with high specificity and affinity (Lipovšek, 2011 ▸). Adnectins have also successfully been used to generate co-crystal structures for a range of targets such as PXR, EGFR, IL-23 and PCSK9 with minimal disruption of the target compared with protein crystallized without Adnectin (Khan et al., 2015 ▸; Ramamurthy et al., 2012 ▸; Mitchell et al., 2014 ▸).

2. Materials and methods

2.1. Macromolecule production

2.1.1. Expression of dog and human HSP47

All of the HSP47 constructs were gene-synthesized by GenScript USA Inc. (Piscataway, New Jersey, USA) and subcloned into a modified pFastBac1 vector (Invitrogen, Carlsbad, California, USA). Two dog HSP47 constructs (residues 36–418, UniProt C7C419_CANLF) were designed for crystallization. One construct, following Widmer et al. (2012 ▸), had a noncleavable 6×His-tag at the C-terminus; the other construct had a TVMV protease-cleavable His-tag at the N-terminus. We also prepared the analogous human HSP47 constructs; human HSP47 has the same length as dog HSP47 and varies at only five positions (dog→human: L69V, E144D, T235M, W359L and E362D). The vector-encoded N-terminal fusion sequence was MGSSHHHHHHSSGETVRFQGAGA–. Following TVMV cleavage, the residual non-native amino-acid residues GAGA– remained on the N-terminus. Avi-tagged dog and human HSP47(19–418) constructs were designed to provide biotinylated protein for Adnectin selection; both had a TVMV protease-cleavable His-tag at the N-terminus. The vector-encoded N-terminal fusion sequence was MGSSHHHHHHSSGETVRFQ–. Following TVMV cleavage, the Avi tag and residual non-native amino-acid residues GLNDIFEAQKIEWHEDTGHM– remained on the N-terminus.

Baculoviruses expressing all of the dog and human HSP47 proteins were generated using the Bac-to-Bac baculovirus expression system (Invitrogen) according to the manufacturer’s protocol. Baculovirus amplification was achieved using infected Sf9 cells (Expression System, Davis, California, USA) at a 1:1500 virus:cell ratio and cultures were grown for 65 h at 27°C post-transfection.

The expression scale-ups were conducted in Sf9 cell cultures grown to a density of 2.1 × 10⁶ cells ml⁻¹ in ESF921 insect medium (Expression Systems) and infected with virus stock at a 1:200 virus:cell ratio. Cultures were maintained in a volume of 800 ml at 130 rev min⁻¹ and were grown for 65 h at 27°C post-infection. The infected cell cultures were harvested by centrifugation at 2000 rev min⁻¹ for 20 min at 4°C in a Sorvall RC12BP centrifuge. The cell pellets were stored at −70°C until purification. Table 1 ▸ describes the details of expressed proteins that resulted in the deposited crystal structures with PDB codes 9cqe, 9cqf, 9cqg, 9cqh, 9cqi and 9cqj.

Table 1. Macromolecule-production information for dog HSP47 and Adnectins used in crystallization.

Construct name	Dog HSP47(36–418)-His6	Dog GAGA-HSP47(36–418)	Adnectin-44	Adnectin-53
PDB codes	9cqe, 9cqf, 9cqg	9cqh, 9cqi, 9cqj	9cqi	9cqj
Source organism	Dog (Canis lupus familiaris)	Dog (Canis lupus familiaris)	Human (Homo sapiens)	Human (Homo sapiens)
Expression vector	Baculovirus	Baculovirus	pET-9d	pET-9d
Plasmid-construction method	Gene synthesis	Gene synthesis	Gene synthesis	Gene synthesis
Expression host	Sf9	Sf9	E. coli BL21(DE3)pLysS	E. coli BL21(DE3)pLysS
Expression details	ESF921 medium	ESF921 medium	Autoinduction medium	Autoinduction medium
Complete amino-acid sequence of the construct produced	MLSPKAATLAERSAGLAFSLYQAMAKDQAVENILLSPVVVASSLGLVSLGGKATTASQAKAVLSAEQLRDEEVHAGLGELLRSLSNSTARNVTWKLGSRLYGPSSVSFAEDFVRSSKQHYNCEHSKINFRDKRSALQSINEWAAQTTDGKLPEVTKDVERTDGALLVNAMFFKPHWDEKFHHKMVDNRGFMVTRSYTVGVTMMHRTGLYNYYDDEKEKLQIVEMPLAHKLSSLIILMPHHVEPLERLEKLLTKEQLKIWMGKMQKKAVAISLPKGVVEVTHDLQKHLAGLGLTEAIDKNKADLSRMSGKKDLYLASVFHATAFEWDTEGNPFDQDIYGREELRSPKLFYADHPFIFLVRDTQSGSLLFIGRLVRPKGDKMRDELLEHHHHHH	GAGALSPKAATLAERSAGLAFSLYQAMAKDQAVENILLSPVVVASSLGLVSLGGKATTASQAKAVLSAEQLRDEEVHAGLGELLRSLSNSTARNVTWKLGSRLYGPSSVSFAEDFVRSSKQHYNCEHSKINFRDKRSALQSINEWAAQTTDGKLPEVTKDVERTDGALLVNAMFFKPHWDEKFHHKMVDNRGFMVTRSYTVGVTMMHRTGLYNYYDDEKEKLQIVEMPLAHKLSSLIILMPHHVEPLERLEKLLTKEQLKIWMGKMQKKAVAISLPKGVVEVTHDLQKHLAGLGLTEAIDKNKADLSRMSGKKDLYLASVFHATAFEWDTEGNPFDQDIYGREELRSPKLFYADHPFIFLVRDTQSGSLLFIGRLVRPKGDKMRDEL	MGVSDVPRDLEVVAATPTSLLISWDAPPYYVRYYRITYGETGGNSPVQEFTVPGSKSTATISGLKPGVDYTITVYAEHPYYDPSSSYYFSSKPISINYRTPHHHHHH	MGVSDVPRDLEVVAATPTSLLISWYHPEQYTEYYRITYGETGGNSPVQEFTVPGERETATISGLKPGVDYTITVYAVGAEQYGGGPDAPISINYRTPHHHHHH

2.1.2. Purification of dog HSP47(36–418)-His₆ and human HSP47(36–418)-His₆

All steps were performed at 4°C unless noted otherwise. Frozen Sf9 cells harvested from a 5 l culture were suspended in 500 ml buffer A [50 mM HEPES pH 8.0, 300 mM NaCl, 5%(v/v) glycerol, 1 mM TCEP] plus ten tablets of cOmplete EDTA-free protease-inhibitor cocktail (Roche Applied Science), 2 mM CHAPS and 20 units per millilitre of Benzonase nuclease (EMD Millipore). The cells were lysed by nitrogen cavitation using a cell-disruption vessel (Parr) pressurized at 375 psi (∼2.58 MPa) for 30 min at 4°C. The lysate was clarified by sedimentation at 143 000g for 60 min (Thermo F10S-6X500Y rotor) at 10°C and the supernatant was loaded onto a HisTrap FF Crude 5 ml column (Cytiva) equilibrated with buffer A with 10 mM imidazole. The column was washed sequentially with nine column volumes each of buffer A with 10, 20 and 30 mM imidazole. The protein was eluted with three column volumes of 500 mM imidazole in buffer A. The nickel-column eluate was loaded onto a HiLoad 26/600 Superdex 200 column (Cytiva) equilibrated with buffer B [25 mM HEPES pH 7.5, 300 mM NaCl, 5%(v/v) glycerol, 0.1 mM EDTA, 1 mM DTT]. The purified protein was pooled, concentrated to 1 mg ml⁻¹ at 10°C and stored at 10°C prior to crystallization setup. The final yield of purified dog HSP47(36–418)-His₆ protein was 6.4 mg per litre of culture, while the human HSP47(36–418)-His₆ protein had a lower yield. The protein purity was ≥98% as determined by SDS–PAGE analysis with Coomassie Brilliant Blue staining (data not shown). Electrospray ionization TOF mass spectrometry was used to confirm the identity of each purified protein. The single, major observed mass species for dog HSP47(36–418)-His₆ was 44 254.3 Da, which was consistent with an N-acetylated initiating methionine preceding HSP47 residue Leu36 and no other post-translational modifications (calculated mass of 44 254.7 Da). The single, major observed mass species for purified human HSP47(36–418)-His₆ was 44 170 Da, which was consistent with an N-acetylated initiating methionine preceding HSP47 residue Leu36 (calculated mass of 44 169.6 Da) as the only post-translational modification.

Both the dog and human HSP47(36–418)-His₆ proteins were found to precipitate when stored on ice at 10 mg ml⁻¹. After filtration of precipitate, the concentration was reduced to 4 mg ml⁻¹, after which precipitation did not continue. The precipitated protein could not be re-solubilized in storage buffer. Concentration and dispensing were therefore performed at 10°C. The protein was flash-frozen at ∼10 mg ml⁻¹ and stored at −80°C.

2.1.3. Purification of dog GAGA-HSP47(36–418) and human GAGA-HSP47(36–418)

Frozen Sf9 cells of dog His-TVMV-GAGA-HSP47(36–418) and human His-TVMV-GAGA-HSP47(36–418) harvested from a 5 l culture were lysed and purified by nickel-affinity purification as described for dog and human HSP47(36–418)-His₆. The nickel-column eluate was loaded onto a HiLoad 26/60 Superdex 200 column (Cytiva) equilibrated with buffer A [50 mM HEPES pH 8.0, 300 mM NaCl, 5%(v/v) glycerol, 1 mM TCEP]. The His-tag was cleaved with His-TVMV protease (1:10 ratio of TVMV:HSP47) overnight. TVMV protease was removed by applying the reaction mixture onto a HisTrap FF 5 ml column (Cytiva) equilibrated with buffer A with 20 mM imidazole. The column-flowthrough fraction was concentrated to 10 ml and buffer-exchanged by dialysis using a 15 ml dialysis cassette in buffer B [25 mM HEPES pH 7.5, 300 mM NaCl, 5%(v/v) glycerol, 0.5 mM EDTA, 1 mM DTT]. The final yield of purified dog GAGA-HSP47(36–418) was 8.5 mg per litre of culture and that for GAGA-human HSP47(36–418) was 6.0 mg per litre of culture. The protein purity was ≥98% as determined by SDS–PAGE analysis with Coomassie Brilliant Blue staining (data not shown). Electrospray ionization TOF mass spectrometry was used to confirm the identity of each purified protein. The single, major observed mass species for dog GAGA-HSP47(36–418) was 43 273 Da, which was consistent with the expected mass for the protein lacking any post-translational modifications (calculated mass of 43 272.6 Da). Likewise, the single, major observed mass species for human GAGA-HSP47(36–418) was 43 188 Da, which was consistent with the expected mass for the unmodified protein (calculated mass of 43 187.5 Da).

2.1.4. Purification of biotinylated dog HSP47(19–418) and human HSP47(19–418)

Frozen Sf9 cells of dog His-TVMV-Avi-HSP47(19–418) and human His-TVMV-Avi-HSP47(19–418) harvested from a 5 l culture were lysed and purified by nickel-affinity purification as described for dog and human HSP47(36–418)-His₆. The nickel-column eluate was loaded onto a HiLoad 26/60 Superdex 200 column (Cytiva) equilibrated with buffer A [50 mM HEPES pH 8.0, 300 mM NaCl, 5%(v/v) glycerol, 1 mM TCEP]. The His-tag was cleaved with His-TVMV protease (1:10 ratio of TVMV:HSP47) overnight. The His-tag-cleaved HSP47 proteins were biotinylated in vitro with His-BirA and biotin as described by Cull & Schatz (2000 ▸). TVMV protease and His-BirA were removed by applying the reaction mixture onto a HisTrap FF 5 ml column (Cytiva) equilibrated with buffer A with 20 mM imidazole. The column-flowthrough fraction was concentrated to 10 ml and buffer-exchanged by dialysis using a 15 ml dialysis cassette in buffer B [50 mM HEPES pH 8.0, 300 mM NaCl, 5%(v/v) glycerol, 2 mM DTT]. The final yield of purified biotinylated dog HSP47(19–418) protein was 4.2 mg per litre of culture. The final yield of purified biotinylated human HSP47(19–418) was 1.67 mg per litre of culture. The protein purity was ≥98% as determined by SDS–PAGE analysis with Coomassie Brilliant Blue staining (data not shown). Electrospray ionization TOF mass spectrometry was used to confirm the protein identity and biotin modification.

2.2. Adnectin selection

HSP47-binding Adnectins were generated using three rounds of mRNA display followed by three rounds of yeast surface display (Table 2 ▸ and Supplementary Fig. S1). The mRNA display selection scheme followed a previously described method (Xu et al., 2002 ▸). Briefly, three naïve libraries based on ¹⁰Fn3 with diversified BC, DE and FG loops (libraries 1 and 2) or diversified loops and selected positions along the exposed β-strands (library 3) were selected against either human or dog HSP47 which had previously been biotinylated. Positive selections were performed at room temperature with the target in solution (Table 3 ▸) and the mRNA-displayed library before being immobilized on magnetic streptavidin beads (DynaBeads M-280; ThermoFisher M-280). For rounds 2 and 3, the library first underwent three sequential negative-selection steps with streptavidin beads only prior to positive selection with target to reduce the prevalence of bead binders.

Table 2. Adnectin selection strategy.

Selection conditions are described in Table 3 ▸. Outputs from S1 and S2 in yeast round 1 were merged to create S1/2 for yeast rounds 2 onwards. Outputs from S4 and S5 in yeast round 1 were merged to create S4/5 for yeast rounds 2 onwards

Selection arm	Naïve library	Target species	mRNA round 1	mRNA round 2	mRNA round 3	Yeast round 1	Yeast round 2	Yeast round 3
S1	Library 1	Human	Condition A	Condition B	Condition B	Condition C	Condition D	Conditions D, E
S2	Library 2	Human	Condition A	Condition B	Condition B	Condition C
S3	Library 3	Human	Condition A	Condition B	Condition B	Condition C	Condition C	Condition D
S4	Library 1	Dog	Condition A	Condition B	Condition B	Condition C	Condition D	Conditions D, E
S5	Library 2	Dog	Condition A	Condition B	Condition B	Condition C
S6	Library 3	Dog	Condition A	Condition B	Condition B	Condition C	Condition D	Condition D

Table 3. Adnectin selection conditions.

Condition	Target concentration (nM)	Incubation time (min)	Selection buffer†
A	100	60	Reducing selection buffer
B	100	30	Reducing selection buffer
C	100	30	PBSA
D	10	30	PBSA
E	1	30	PBSA

^†Reducing selection buffer is composed of 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄, 137 mM NaCl, 2.7 mM KCl, 2 mM DTT, 0.125% Tween-20, 1 mg ml⁻¹ BSA, 0.1 mg ml⁻¹ sheared salmon sperm DNA. PBSA is composed of 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄, 137 mM NaCl, 2.7 mM KCl, 0.1%(w/v) BSA.

The outputs from mRNA display round 3 were taken into yeast surface display following a previously published method (Koide et al., 2012 ▸). For round 1, Adnectin display was monitored in FACS (Aria III) by a C-terminal cMyc tag with an anti-cMyc-chicken primary antibody (ThermoFisher) and a goat-anti-chicken-A488 secondary antibody (ThermoFisher). Target engagement was measured with an anti-biotin rabbit DyLight549 antibody (Rockland) to reduce the prevalence of any streptavidin binders from the earlier selection rounds. Outputs from yeast selection arms 1 and 2 and arms 4 and 5 from round 1 were pooled prior to yeast surface display round 2 as the starting libraries all shared the same outermost PCR primers. Selection pressure was increased from yeast display round 2 onwards by decreasing the target concentration for arms showing a significant binding signal in the FACS sort gate (>20× cells versus the no same gate for no target control) and by adjusting the FACS sort gates to isolate only small cells (low forward scatter) with high Adnectin expression (cMyc) to bias the library toward Adnectins with good expression. Since all arms of selection showed a binding signal in round 3, this round was repeated under a set of standard conditions to aid in hit picking: 100 nM human HSP47, 100 nM dog HSP47 and 100 nM HSP90 as an irrelevant control. All selection outputs were barcoded (Nextera XT v2) with PCR primers specific to the Adnectin flanking regions and sequenced by Next Generation Sequencing (NGS; MiSeq 600cycle v3 kit), except for the 100 nM HSP90 samples as too few cells were in the sort gate to provide a meaningful comparison. NGS data were processed using custom software developed to provide sequence and abundance information for all Adnectins found in the selected populations. Hits were picked from the NGS data based on a combination of abundance, equal binding to both human and dog HSP47 and enrichment at lower target concentrations. See the Supplementary Data File with NGS results for 77 hits.

2.3. Expression and purification of HSP47-Adnectins

After selection, 77 Adnectins, each with a C-terminal His₆ tag, were cloned into pET-9d vectors and transformed into Escherichia coli BL21(DE3)pLysS cells. The Adnectins were produced on a test scale in a plate-based high-throughput expression and purification process (see Table 1 ▸ for the sequences of the two used in the structures deposited in the wwPDB and Supplementary Table S1 for the loop sequences of all 16 Adnectins chosen for test-scale expression). Glycerol stocks of the chosen Adnectins were stamped into 1 ml LB (Lennox formulation, Gibco) containing 50 µg ml⁻¹ kanamycin in 96 deep-well plate starter cultures (Whatman) and grown at in a Glascol Vertiga shaker at 37°C overnight at 700 rev min⁻¹. Starter cultures were inoculated into 4 ml cultures in a 24 deep-well plate (Seahorse) containing MagicMedia E. coli Expression Medium (ThermoFisher) and grown at 37°C for 4 h followed by 20°C for 20 h at 400 rev min⁻¹. After centrifugal harvesting, Adnectins were purified from the soluble fraction using metal-chelation chromatography (600 µl Repligen RoboColumn 5 mm internal diameter × 30 mm Nickel Sepharose 6 FF) and tested for monomericity (anSEC Superdex 75 Increase 5/150 in 100 mM sodium phosphate, 150 mM sodium chloride pH 7.2) and mass confirmation via Agilent ZORBAX RRHD 300SB-C8 2.1 × 50 mm, 1.8 µm RP in 0.1% formic acid/acetonitrile at 0.5 ml min⁻¹ and 75°C on a Thermo QE-MS.

At scale-up, glycerol stocks for several HSP47-Adnectins were plated on LB-Kan agar plates (TekNova) and cultured overnight at 37°C. Single colonies were chosen to inoculate starter cultures grown overnight at 37°C in LB-Kan. Production cultures were grown from the starter cultures by a 1:40 dilution into MagicMedia-Kan and were grown at 37°C and 250 rev min⁻¹ for 18 h. The cultured cells were harvested by centrifugation and stored at −80°C until purification. The pellets were mechanically lysed in 20 mM sodium phosphate, 0.5 M sodium chloride pH 7.4 with cOmplete EDTA-free protease-inhibitor tablets (Roche Applied Sciences) and centrifuged at 14 000g to clear insertion bodies and debris. The cleared lysates were purified using a 5 ml Cytiva HisTrap HP column equilibrated in 20 mM sodium phosphate, 0.5 M sodium chloride pH 7.4, washed with the addition of 40 mM imidazole and eluted in three column volumes with the same buffer with 400 mM imidazole. The eluate was further purified via HiLoad 16/600 Superdex 200 in 20 mM HEPES pH 7.5, 300 mM NaCl, 4 mM DTT pH 7.5. Adnectins were subsequently complexed with HSP47 for crystallization.

2.4. Initial screening of HSP47-binding Adnectins by ELISA

Adnectins from high-throughput protein expression and purification were assayed for binding to HSP47 via a direct sandwich ELISA. Three washes of 100 µl phosphate-buffered saline with 0.1%(w/v) Tween-20 were used between each of the subsequent steps. Nunc 96-well Maxisorp Plates (ThermoFisher) were coated overnight with an anti-His antibody and blocked with StartingBlock T20 buffer (ThermoFisher) for 1 h. His-tagged, high-throughput-expressed Adnectins in blocking buffer were diluted 1:20, and 50 µl was applied to the plate and incubated for 1 h at room temperature. The target was applied as a 100 µl solution of 100 nM biotinylated human or dog HSP47 or an irrelevant biotinylated target followed by a 1 h room-temperature incubation. Plates were developed by first adding 100 µl of 1:10 000 diluted Pierce High-Sensitivity Streptavidin-HRP (ThermoFisher) in blocking buffer for 30 min, followed by 100 µl Stabilized Chromogen TMB (Life Tech). After developing for 5–15 min, the reaction was quenched by adding 100 µl 2 N hydrogen peroxide (VWR) and the absorbance was measured at 450 nm.

2.5. Confirmation of Adnectin–dog HSP47 complex formation by size-exclusion chromatography with multi-angle static light scattering (SEC–MALS)

Each purified Adnectin was complexed with dog HSP47 in a 1:1 (Adnectin:dog HSP47) molar ratio. The mixtures were then incubated at room temperature for 1 h to allow formation of the complexes. To assess the complexes, SEC–MALS analysis was performed using an analytical HPLC–SEC system (Agilent 1260 HPLC) with a YMC-Pack Diol-120 column (4.6 mm internal diameter × 30 cm), and the molar-mass shift resulting from complex formation was measured using a DAWN Wyatt MALS detector. The mobile phase used for the analysis consisted of 100 mM sodium phosphate, 150 mM sodium chloride pH 7.2; the flow rate was set to 0.3 ml min⁻¹. Dog HSP47 and Adnectin were injected separately as controls at 10 µg. For each complex, 20 µg was then injected. The SEC–MALS analysis revealed a mass shift, indicating the formation of complexes between Adnectin and dog HSP47. However, some free Adnectin and dog HSP47 were still present in the samples. In total, 24 Adnectin–dog HSP47 complexes were evaluated using this method and complex formation was confirmed (see Supplementary Fig. S2 for a representative chromatogram).

2.6. Dog GAGA-HSP47(36–418)–Adnectin complexation

Alexa Fluor 488 and 555 NHS ester dyes (ThermoFisher) were used in conjunction with a RockImager (Formulatrix) fitted with multi-fluorescence imaging (MFI) to confirm protein–protein complexes in crystals prior to X-ray data collection. 1 µM Alexa Fluor 488 NHS ester dye was added to each Adnectin and 1 µM Alexa Fluor 555 NHS ester dye was added to dog GAGA-HSP47(36–418). The samples were incubated for 1 h at 4°C. The complex was formed by mixing dog GAGA-HSP47(36–418) with a 1.5-fold molar excess of Adnectin. The complex was incubated overnight at 4°C prior to concentrating the sample using a 10K molecular-weight cutoff spin concentrator (Millipore).

The complex sample was separated from excess Adnectin using a HiLoad 26/60 Superdex 75 column (Cytiva) equilibrated with buffer consisting of 40 mM HEPES pH 7.5, 200 mM NaCl, 5%(v/v) glycerol. Fractions of the complex of dog GAGA-HSP47(36–418) and Adnectin were pooled and concentrated to 25 mg ml⁻¹ for crystallization.

2.7. Crystallization

Crystallization conditions were determined using a Mosquito drop setter (SPT Labtech) by screening against the following commercial screens: PACT, MCSG-1, MCSG-2, MCSG-3 and MCSG-4, NR-LBD, TOP96, Wizard 1+2 and JCSG from Molecular Dimensions, Index, PEGRx, Crystal Screen HT, Crystal Screen 1+2 and SaltRx from Hampton Research and Ammonium Sulfate Suite from Rigaku.

Crystallization conditions for dog HSP47(36–481)-His₆, dog GAGA-HSP47(36–418), dog GAGA-HSP47(36–418)–Adnectin-44 and dog GAGA-HSP47(36–418)–Adnectin-53 are described in Table 4 ▸.

Table 4. Crystallization.

Construct name	Dog HSP47(36–418)-His6	Dog HSP47(36–418)-His6	Dog HSP47(36–418)-His6
PDB code	9cqe	9cqf	9cqg
Method	Sitting-drop vapor diffusion	Sitting-drop vapor diffusion	Sitting-drop vapor diffusion
Plate type	SD-2	SD-2	SD-2
Temperature (°C)	4	4	4
Protein concentration (mg ml−1)	10.0	15.01	13.8
Buffer composition of protein solution	20 mM HEPES pH 7.5, 300 mM NaCl, 4 mM DTT	25 mM HEPES pH 7.5, 300 mM NaCl, 5%(v/v) glycerol, 0.1 mM EDTA, 1 mM DTT	25 mM HEPES pH 7.5, 300 mM NaCl, 5%(v/v) glycerol, 0.1 mM EDTA, 1 mM DTT
Composition of reservoir solution	0.18 M ammonium sulfate, 0.09 M Tris pH 8.5, 22.5%(w/v) PEG 3350, 0.2 M nondetergent sulfobetaine (NDSB) 221	20%(w/v) PEG 3350, 200 mM lithium citrate tribasic	21.7%(w/v) PEG 3350, 50 mM 1lithium citrate tribasic
Volume and ratio of drop	500 nl, 1:1	300 nl, 1:1	300 nl, 1:1
Volume of reservoir (µl)	80	80	80
Drop setting	Mosquito	Mosquito	Mosquito
Seeding	No	No	No
Composition of the cryoprotectant	75%(v/v) well solution, 10%(v/v) PEG 400, 10%(v/v) glycerol, 5%(v/v) water	75%(v/v) well solution, 10%(v/v) PEG 400, 10%(v/v) glycerol, 5%(v/v) water	75%(v/v) well solution, 10%(v/v) PEG 400, 10%(v/v) glycerol, 5%(v/v) water

Construct or complex name	Dog GAGA-HSP47(36–418)	Dog GAGA-HSP47(36–418)–Adnectin-44	Dog GAGA-HSP47(36–418)–Adnectin-53
PDB code	9cqh	9cqi	9cqj
Method	Sitting-drop vapor diffusion	Sitting-drop vapor diffusion	Sitting-drop vapor diffusion
Plate type	SD-2	SD-2	SD-2
Temperature (°C)	19	19	19
Protein concentration (mg ml−1)	16	25	25.7
Buffer composition of protein solution	25 mM HEPES pH 7.5, 300 mM NaCl, 5%(v/v) glycerol, 0.1 mM EDTA, 1 mM DTT	40 mM HEPES pH 7.5, 200 mM NaCl, 5%(v/v) glycerol	40 mM HEPES pH 7.5, 200 mM NaCl, 5%(v/v) glycerol
Composition of reservoir solution	100 mM MMT† pH 8.5, 25%(w/v) PEG 3350, 500 mM lithium sulfate	2.4 M sodium malonate pH 7.0	100 mM HEPES pH 7.5, 25%(w/v) PEG 3350, 200 mM lithium sulfate monohydrate
Volume and ratio of drop		400 nl, 1:1	500 nl, 1:1
Volume of reservoir (µl)	80	80	80
Drop setting	Mosquito	Mosquito	Mosquito
Seeding	No	No	No
Composition of the cryoprotectant	75%(v/v) well solution, 10%(v/v) PEG 400, 10%(v/v) glycerol, 5%(v/v) water	3.4 M sodium malonate pH 7.0	75%(v/v) well solution, 10%(v/v) PEG 400, 10%(v/v) glycerol, 5%(v/v) water

^†MMT is a mixture of DL-malic acid, MES monohydrate and Tris.

2.8. Data collection and processing

Data were processed using autoPROC (Vonrhein et al., 2011 ▸), with integration using XDS (Kabsch, 2010 ▸), isotropic scaling using either XSCALE (Kabsch, 2010 ▸; Table 5 ▸, PDB entries 9cqe and 9cqf) or AIMLESS (Evans & Murshudov, 2013 ▸; Agirre et al., 2023 ▸; Table 5 ▸, PDB entries 9cqg, 9cqh, 9cqi and 9cqj) and anisotropic scaling using STARANISO (Tickle et al., 2018 ▸).

Table 5. Data collection and processing.

Values in parentheses are for the outer shell.

PDB code	9cqe	9cqf	9cqg
Protein	Dog HSP47(36–418)-His6	Dog HSP47(36–418)-His6	Dog HSP47(36–418)-His6
Diffraction source	17-ID, APS	17-ID, APS	17-ID, APS
Wavelength (Å)	1.0	1.0	1.0
Temperature (K)	100	100	100
Detector	Dectris EIGER2 X 9M	Dectris EIGER2 X 9M	Dectris EIGER2 X 9M
Crystal-to-detector distance (mm)	350	450	250
Rotation range per image (°)	0.25	0.25	0.25
Total rotation range (°)	180	180	360
Exposure time per image (s)	0.040	0.078	0.040
Space group	P21	C2221	P1
a, b, c (Å)	140.2, 75.3, 163.9	117.8, 129.9, 501.9	87.3, 92.0, 123.6
α, β, γ (°)	90, 110.8, 90	90, 90, 90	85.2, 70.7, 83.2
Mosaicity (°)	0.44–0.57	0.10–0.11	0.17–0.23
Isotropic processing
Resolution range (Å)	153.19–3.80 (3.87–3.80)	250.97–3.38 (3.44–3.38)	116.50–2.95 (3.00–2.95)
2σ operational diffraction limit† (Å)	4.28	3.71	3.21
Total No. of reflections	106116 (5318)	374323 (18909)	182449 (7521)
No. of unique reflections	31672 (1561)	54172 (2622)	73067 (3692)
Completeness (%)	99.0 (99.4)	99.8 (99.6)	96.6 (97.0)
Multiplicity	3.4 (3.4)	6.9 (7.2)	2.5 (2.0)
〈I/σ(I)〉	6.2 (2.2)	7.8 (2.2)	9.6 (2.2)
CC1/2	0.994 (0.822)	0.993 (0.834)	0.995 (0.751)
Rr.i.m.	0.167 (0.680)	0.216 (0.954)	0.090 (0.580)
Anisotropic processing
Resolution range (Å)	153.19–3.18 (3.45–3.18)	250.97–2.93 (3.21–2.93)	116.50–2.47 (2.71–2.47)
Ellipsoidal diffraction-limit cutoffs	0.877a* − 0.481c*: 3.13 Å	a*: 3.24 Å	0.668a* − 0.271b* + 0.693c*: 3.04 Å
	b*: 3.31 Å	b*: 3.29 Å	0.448a* + 0.886b* + 0.121c*: 2.65 Å
	0.116a* + 0.993c*: 3.93 Å	c*: 2.92 Å	−0.234a* + 0.216b* + 0.948c*: 2.47 Å
2σ operational diffraction limit† (Å)	4.06	3.58	3.10
Total No. of reflections	134820 (6991)	437083 (20106)	233970 (10795)
No. of unique reflections	39821 (1991)	63276 (3165)	89165 (4158)
Completeness (%)	92.6 (57.2)	93.9 (65.6)	90.9 (59.7)
Multiplicity	3.4 (3.5)	6.9 (6.4)	2.6 (2.6)
〈I/σ(I)〉	5.0 (1.4)	7.1 (1.6)	7.9 (1.5)
CC1/2	0.992 (0.499)	0.992 (0.658)	0.995 (0.455)
Rr.i.m.	0.210 (1.063)	0.237 (1.201)	0.106 (0.987)
Overall B factor from Wilson plot (Å2)	83	96	76

PDB code	9cqh	9cqi	9cqj
Protein or complex	Dog GAGA-HSP47(36–418)	Dog GAGA-HSP47(36–418)–Adnectin-44	Dog GAGA-HSP47(36–418)–Adnectin-53
Diffraction source	17-ID, APS	17-ID, APS	17-ID, APS
Wavelength (Å)	1.0	1.0	1.0
Temperature (K)	100	100	100
Detector	Dectris EIGER2 X 9M	Dectris EIGER2 X 9M	Dectris EIGER2 X 9M
Crystal-to-detector distance (mm)	150	200	200
Rotation range per image (°)	0.25	0.25	0.25
Total rotation range (°)	180	180	180
Exposure time per image (s)	0.020	0.020	0.020
Space group	C2	C2	P21212
a, b, c (Å)	180.0, 115.5, 187.5	94.3, 82.5, 98.3	86.8, 129.1, 78.2
α, β, γ (°)	90.0, 106.8, 90.0	90.0, 116.1, 90.0	90.0, 90.0, 90.0
Mosaicity (°)	0.12	0.20–0.23	0.41–0.94
Isotropic processing
Resolution range (Å)	95.94–2.51 (2.56–2.51)	88.27–2.34 (2.38–2.34)	78.22–2.29 (2.33–2.29)
2σ operational diffraction limit† (Å)	2.75	2.59	2.48
Total No. of reflections	397471 (20615)	98974 (4813)	253899 (12552)
No. of unique reflections	123193 (6150)	28471 (1410)	39717 (1920)
Completeness (%)	98.2 (98.0)	99.6 (99.9)	98.8 (98.6)
Multiplicity	3.2 (3.4)	3.5 (3.4)	6.4 (6.5)
〈I/σ(I)〉	9.6 (2.1)	11.8 (2.1)	12.9 (2.1)
CC1/2	0.996 (0.746)	0.997 (0.764)	0.998 (0.731)
Rr.i.m.	0.096 (0.711)	0.091 (0.676)	0.079 (0.874)
Anisotropic processing
Resolution range (Å)	179.53–2.01 (2.28–2.01)	88.27–1.94 (2.10–1.94)	78.22–2.08 (2.19–2.08)
Ellipsoidal diffraction-limit cutoffs	0.984a* + 0.177c*: 2.01 Å	0.837a* − 0.547c*: 1.88 Å	a*: 2.08 Å
	b*: 2.39 Å	b*: 2.08 Å	b*: 2.08 Å
	−0.389a* + 0.921c*: 2.55 Å	0.240a* + 0.971c*: 2.41 Å	c*: 2.22 Å
2σ operational diffraction limit† (Å)	2.65	2.46	2.42
Total No. of reflections	486247 (24531)	127998 (6670)	302494 (15776)
No. of unique reflections	150858 (7544)	36513 (1827)	46990 (2350)
Completeness (%)	92.1 (60.7)	93.6 (66.9)	93.4 (53.2)
Multiplicity	3.2 (3.3)	3.5 (3.7)	6.4 (6.7)
〈I/σ(I)〉	8.2 (1.8)	9.7 (1.5)	11.1 (1.4)
CC1/2	0.996 (0.662)	0.997 (0.547)	0.998 (0.620)
Rr.i.m.	0.105 (0.785)	0.111 (0.994)	0.086 (1.325)
Overall B factor from Wilson plot (Å2)	41	31	51

^†The operational diffraction limit is such that an isotropic data set 100% complete to that limit, according to the cutoff criterion used, would contain the same number of reflections.

2.9. Structure determination and refinement

The structures were determined by molecular replacement with Phaser (McCoy et al., 2007 ▸) using chain A from PDB entry 4aua (Widmer et al., 2012 ▸). In addition, the data that gave rise to PDB entry 9cqh, given the similarity of the unit-cell parameters to those of PDB entry 4aua, was attempted using the entire model of PDB entry 4aua. The structures were refined with BUSTER (Bricogne et al., 2019 ▸; Smart et al., 2012 ▸) and the models were fitted into the electron density with Coot (Emsley & Cowtan, 2004 ▸; Emsley et al., 2010 ▸). Refinement details are provided in Table 6 ▸. ALIGN (Cohen, 1997 ▸) was used for superpositions.

Table 6. Structure refinement.

Values in parentheses are for the outer shell.

PDB code	9cqe	9cqf	9cqg
Protein	Dog HSP47(36–418)-His6	Dog HSP47(36–418)-His6	Dog HSP47(36–418)-His6
Resolution range (Å)	45.85–3.18 (3.36–3.18)	250.97–2.93 (3.10–2.93)	116.50–2.47 (2.63–2.47)
Completeness (%)	73.1 (9.4)	76.0 (9.5)	69.4 (8.2)
σ Cutoff	0	0	0
No. of reflections, working set	37779 (752)	60201 (1208)	84700 (1689)
No. of reflections, test set	2024 (45)	3075 (58)	4465 (95)
Final Rcryst	0.263 (0.330)	0.228 (0.328)	0.243 (0.334)
Final Rfree	0.288 (0.386)	0.252 (0.318)	0.264 (0.355)
Cruickshank DPI	0.74	0.48	0.37
No. of non-H atoms
Protein	23196	24135	24351
Water	0	7	45
Total	23196	24142	24396
R.m.s. deviations†
Bond lengths (Å)	0.007	0.009	0.009
Angles (°)	0.9	1.0	1.0
Average B factors (Å2)
Protein	91	79	73
Water		31	39
Ramachandran plot‡
Favored (%)	94.7	92.9	92.8
Outliers (%)	0.6	0.5	0.3

PDB code	9cqh	9cqi	9cqj
Protein or complex	Dog GAGA-HSP47(36–418)	Dog GAGA-HSP47(36–418)–Adnectin-44	Dog GAGA-HSP47(36-418)–Adnectin-53
Resolution range (Å)	28.87–2.01 (2.17–2.01)	26.71–1.94 (2.04–1.94)	78.22–2.08 (2.15–2.08)
Completeness (%)	61.5 (6.0)	72.9 (10.6)	87.2 (18.5)
σ Cutoff	0	0	0
No. of reflections, working set	143234 (2845)	34664 (692)	44631 (894)
No. of reflections, test set	7538 (171)	1833 (38)	2359 (46)
Final Rcryst	0.234 (0.295)	0.214 (0.275)	0.248 (0.310)
Final Rfree	0.254 (0.298)	0.245 (0.304)	0.273 (0.357)
Cruickshank DPI	0.217	0.170	0.229
No. of non-H atoms
Protein	20627	3795	7220
Ion (SO4)	40
Glycerol	12
Water	623	170	109
Total	21302	3965	7329
R.m.s. deviations†
Bond lengths (Å)	0.008	0.008	0.008
Angles (°)	1.0	1.0	1.0
Average B factors (Å2)
Protein	50	345	55
Ion (SO4)	89
Glycerol	42
Water	39	35	45
Ramachandran plot‡
Favored (%)	96.0	96.5	96.2
Outliers (%)	0.4	0.6	0.1

^†Restraints are from Engh & Huber (2006 ▸) as implemented in BUSTER.

^‡From MolProbity (Chen et al., 2010 ▸).

3. Results

3.1. Crystallization and structure determination of C-terminally His-tagged dog HSP47

With a literature path (Widmer et al., 2012 ▸) to obtaining an apo structure of dog HSP47, we expected this to be a straightforward system, but it turned out to be anything but. We started with E. coli-expressed protein as described in Widmer et al. (2012 ▸). In our hands, the C-terminally His-tagged protein tended to precipitate at 4°C at the protein concentrations described in Widmer et al. (2012 ▸). Even when we switched to the same protein expressed in baculovirus-infected insect cells, the protein exhibited essentially the same behavior, but its yield was higher. It should be noted that the HSP47 constructs that we and Widmer et al. (2012 ▸) used lacked the N-terminal signal sequence and additional N-terminal residues present in native HSP47. Because of this, the recombinant HSP47 constructs were expressed in the cytosol and, in the case of the insect-cell expression system, the proteins were not translocated into the ER lumen or glycosylated despite the presence of two potential N-linked glycosylation sites (i.e. Asn120 and Asn125). Moreover, crystallization was difficult, and we were never able to grow crystals with a C-terminally His-tagged construct that replicated PDB entry 4au4 either in crystal form or diffraction limit (Table 5 ▸; PDB entries 9cqe, 9cqf and 9cqg). We also attempted to crystallize the analogous human HSP47 protein, but it was even more poorly behaved and never exceeded a diffraction limit of ∼8 Å. Instead, with dog HSP47 we kept obtaining crystal forms with eight molecules per asymmetric unit consisting of two tetramers, with the tetramers arranged differently to each other depending on the crystal form (Fig. 1 ▸; PDB entries 9cqe, 9cqf and 9cqg). We noticed that in each of these crystal forms pairs of dog HSP47 molecules interacted through the C-terminal His-tag, and we suspected that this might be the problem with this construct. We took two approaches to obviate this problem, both of which yielded better behaved protein, more straightforward crystallization and better resolution: (i) the use of an N-terminal cleavable His-tag for purification rather than a C-terminal noncleavable His-tag and (ii) the use of crystallization chaperones in the form of Adnectin molecules.

Figure 1.

Superposition of one tetramer from each of the P2₁ (PDB entry 9cqe; red), C222₁ (PDB entry 9cqf; blue) and P1 (PDB entry 9cqg; cyan) crystal forms, showing that the tetramers superimpose well but the second tetramer in each asymmetric unit is arrayed differently in the three crystal forms. To superimpose PDB entries 9cqe (P2₁) and 9cqf (C222₁) on PDB entry 9cqg (P1), the C^α positions of chains A, B, D and E (a tetramer) were superimposed. The resulting C^α r.m.s.d. was 1.2 Å for 1486 (of 1524 possible) pairs for PDB entry 9cqe on PDB entry 9cqg and 0.8 Å for 1509 (of 1545 possible) pairs for PDB entry 9cqf on PDB entry 9cqg.

3.2. Crystallization and structure determination of dog GAGA-HSP47(36–418)

The first approach to come to fruition was the N-terminally His-tagged and cleaved protein dog GAGA-HSP47(36–418), which expressed better and was stable at higher concentrations in solution at 4°C than the C-terminally His-tagged protein and readily grew better diffracting crystals. The crystals seemed to be the same crystal form as PDB entry 4aua (C2; a = 178.8, b = 116.2, c = 188.4 Å, β = 107.7°), compared with the data that gave rise to PDB entry 9cqh (C2; a = 180.0, b = 115.5, c = 187.5 Å, β = 106.8°). Molecular replacement with the seven molecules of the aymmetric unit of PDB entry 4au4 yielded decent molecular-replacement statistics (Table 7 ▸) but, retrospectively, a mistake in preparing the Phaser-derived coordinates for refinement led us, at the time, to believe that something was wrong with the molecular-replacement solution. We then tried molecular replacement with chain A of PDB entry 4au4, which found seven molecules per asymmetric unit, yielding a better initial TFZ and better initial and refined LLG scores (Table 7 ▸), and after an initial round of refinement yielded R_free = 0.276 and R_work = 0.258 (PDB entry 9cqh). Thus, despite the different constructs, PDB entries 4au4 and 9cqh are the same crystal form.

Table 7. Phaser molecular-replacement statistics for C2 crystals of dog GAGA-HSP47(36–418), the data set that gave rise to PDB entry 9cqh.

Model	Monomers	RFZ	TFZ	PAK	LLG	Refined TFZ	Refined LLG
PDB entry 4aua, asymmetric unit	7	20.2	14.9	7	513	125.7	20206
PDB entry 4aua, chain A	7	17.6	67.1	1	12850	116.7	30079

We also crystallized human GAGA-HSP47(36–418), but while it diffracted better than the C-terminally His-tagged human HSP47, the 2σ anisotropic operational diffraction limit (the definition is given in the footnote to Table 5 ▸) was 5.08 Å and thus it will not be further discussed.

3.3. Adnectin screening

Human and dog HSP47-binding Adnectins were generated from naïve libraries using mRNA display followed by yeast surface display (Supplementary Fig. S1). For characterization of the binding to HSP47 by ELISA and of the percentage of monomer by size-exclusion chromatography (SEC), 77 Adnectin NGS hits were initially selected for high-throughput protein purification. Of the initial 77 NGS hits, 16 Adnectins from 13 separate sequence families were selected for scale-up as ELISA positive (>40× over an irrelevant target) and >50% monomeric by analytical SEC (Supplementary Table S2). The top eight Adnectins were confirmed to form complexes with HSP47 via an apparent molecular-weight shift in SEC–MALS (Supplementary Fig. S2) and progressed into co-crystallization screens.

3.4. Crystallization and structure determination with Adnectin chaperones

The other approach of using crystallization chaperones also yielded well diffracting crystals for two of the top eight Adnectins, which had the advantage of having only one or two HSP47 molecules per asymmetric unit and consequently yielded smaller unit cells and higher resolution (PDB entries 9cqi and 9cqj). Interestingly, despite an agnostic selection strategy, both Adnectin molecules bound in the collagen model peptide binding site, but were oriented at different angles (Fig. 2 ▸).

Figure 2.

Cartoon diagrams of dog HSP47 bound to Adnectins, showing that they bind in the collagen model peptide site. (a) HSP47 (cyan)–Adnectin-44 (magenta) complex (PDB entry 9cqi) showing that a single complex is present in the asymmetric unit. (b) HSP47 (cyan, red)–Adnectin-53 (magenta, blue) complex (PDB entry 9cqj) showing that two complexes are present in the asymmetric unit. (c) Superposition of the HSP47 of the HSP47–Adnectin-53 complex on the HSP47 of the HSP47–Adnectin-44 complex, but showing only the HSP47 (cyan) from the Adnectin-44 complex with Adnectin-44 (magenta) and Adnectin-53 (blue). Although the epitopes of Adnectin-44 and Adnectin-53 overlap, the orientation of the Adnectin relative to HSP47 differs between the two complexes.

4. Discussion

After considerable unsuccessful attempts to reproduce the literature crystal form of apo dog HSP47 (PDB entry 4au4) and obtaining crystal forms that were undesirable, we took two approaches to developing a working crystallization system that would be suitable for iterative structure-based drug design: (i) switching the noncleavable His-tag from the C-terminus to the N-terminus and making it cleavable, and (ii) using crystallization chaperones in the form of Adnectins; an approach previously used successively by our colleagues with PXR (Khan et al., 2015 ▸). Both yielded better diffracting crystal forms, but the crystallization-chaperone approach also yielded crystal forms with only one or two complexes per asymmetric unit, which made model fitting less onerous.

In all constructs we found human HSP47 to be less well behaved than dog HSP47, which is somewhat surprising given that only five sequence differences exist. We believe that four of these changes (L69V, E144D, W359L and E362D) are unlikely to be the cause of the differences, which is supported by comparing the AlphaFold-predicted human HSP47 structure (AF-P50454-F1; Jumper et al., 2021 ▸; Berman et al., 2003 ▸) with the experimentally determined dog HSP47 structure, which shows minimal differences even in the case of W359L, where one might have anticipated some rearrangements to accommodate the smaller leucine. However, the T235M change occurs on the surface in a β-strand with the sequence SYTVGVTMMHRT and we suspect that the third consecutive methionine, although Met236 and Met237 are buried, contributes to the less well behaved human protein. Fig. 3 ▸ shows the relative positions of the five amino-acid differences between the dog and human proteins.

Figure 3.

Cartoon representation of dog HSP47 showing the five amino-acid differences between the dog and human proteins.

Adnectin selections against HSP47 successfully generated a wide range of binders, with only a few Adnectins having significant sequence similarity within the FG loops, which are analogous to CDR3 of an antibody (Lipovšek, 2011 ▸). Species cross-reactivity between human and dog was also high among the Adnectins tested, which was likely to be due to the 98% sequence identity between the two protein constructs. Both co-crystal structures of HSP47 with Adnectin reported here (PDB entries 9cqi and 9cqj), and a third HSP47–Adnectin structure that was not reported due to suboptimal diffraction, are from Adnectins that bind to the collagen model peptide binding site. However, the other Adnectin binders to HSP47 could have bound elsewhere but did not yield co-crystals.

We often find that challenges exist in replicating literature results for structure. Neither of the two approaches that were successfully used here for HSP47 are novel or will work in all situations, but after troubleshooting is completed and unsuccessful one should take alternative approaches as early as possible.

Supplementary Material

PDB reference: apo dog HSP47(36–418) with C-terminal His-tag, space group P2₁, 9cqe

PDB reference: space group C222₁, 9cqf

PDB reference: space group P1, 9cqg

PDB reference: apo dog HSP47(36–418), N-terminal His-tag cleaved, space group C2, 9cqh

PDB reference: dog GAGA-HSP47(36–418), complex with Adnectin-44, 9cqi

PDB reference: complex with Adnectin-53, 9cqj