Modelling of microbial polyhydroxyalkanoate surface binding protein PhaP for rational mutagenesis

Summary Phasins are unusual amphiphilic proteins that bind to microbial polyhydroxyalkanoate (PHA) granules in nature and show great potential for various applications in biotechnology and medicine. Despite their remarkable diversity, only the crystal structure of PhaPA h from Aeromonas hydrophila has been solved to date. Based on the structure of PhaPA h, homology models of PhaPA z from Azotobacter sp. FA‐8 and PhaPTD from Halomonas bluephagenesis TD were successfully established, allowing rational mutagenesis to be conducted to enhance the stability and surfactant properties of these proteins. PhaPA z mutants, including PhaPA zQ38L and PhaPA zQ78L, as well as PhaPTD mutants, including PhaPTDQ38M and PhaPTDQ72M, showed better emulsification properties and improved thermostability (6‐10°C higher melting temperatures) compared with their wild‐type homologues under the same conditions. Importantly, the established PhaP homology‐modelling approach, based on the high‐resolution structure of PhaPA h, can be generalized to facilitate the study of other PhaP members.


Introduction
Polyhydroxyalkanoates (PHA) are polymers that are accumulated as intracellular carbon and energy storage compounds in many bacteria (Anderson and Dawes, 1990;Madison and Huisman, 1999;Keshavarz and Roy, 2010). The stored PHA helps the bacteria survive extreme conditions such as temperature shifts, changes of pH and osmotic pressure and/or lack of nutrients (Potter, 2004;Wang et al., 2014). Due to their excellent biodegradability, biocompatibility and thermo-processability, PHAs are not only interesting as environmentally friendly bioplastics (Chen et al., 2001;Potter et al., 2001), but also as biomaterials suitable for implant purposes and as drug delivery matrices (Potter, 2004;Wang et al., 2014).
PHAs are accumulated as intracellular granules surrounded by an organized protein layer composed of PHA synthases, PHA depolymerases and the amphiphilic phasins or PhaPs, which are the object of this study (Potter and Steinbuchel, 2005;Jendrossek and Pfeiffer, 2014). The combination of structural, synthetic and catabolic proteins on the surface of PHA granules forms the complex structure of the carbonosome (Potter and Steinbuchel, 2005;Jendrossek and Pfeiffer, 2014). It is well established that PhaPs bind to the surface of PHA granules Jendrossek, 2011, 2012), forming an interphase between the cytoplasm and the hydrophobic PHA granules to prevent granular coalescence (Wieczorek et al., 1995). A number of studies have demonstrated potential applications of PhaPs in biotechnology and medicine, from protein purification, surface coating and enzyme chaperones to targeted drug delivery (Qi et al., 2000;Handrick et al., 2004;Kuchta et al., 2007;Galan et al., 2011). In view of this, it has become increasingly attractive to understand the structure and function of PhaPs. Apart from their structural role on the surface of PHA granules, other PhaP functions have been reported, including influencing the number and sizes of PHA granules and promoting PHA accumulation (Wieczorek et al., 1995;Kuchta et al., 2007). Some PhaPs were reported to have regulatory activity, such as PhaF from Pseudomonas putida, which regulates PHA synthesis (Prieto et al., 1999;Galan et al., 2011), ApdA from Rhodospirillum rubrum, which induces polyhydroxybutyrate (PHB) depolymerization (Handrick et al., 2004), PhaP 1 of R. eutropha, which enhances the specific activity of PHA polymerase in vivo (Qi et al., 2000), and Pha-P Ah from Aeromonas hydrophila strain 4AK4, which affects its cognate PHA polymerase at different levels (Tian et al., 2005). The effects of PhaPs on PHA accumulation are attributed to several mechanisms, including the activation of genes involved in PHA synthesis (Tian et al., 2005) and the formation of a barrier between the PHA granules and other intracellular components to help avoid negative effects of PHA accumulation on cellular activities (Wieczorek et al., 1995).
Although the degree of conservation among PhaPs is very low, some motifs that are characteristic of most if not all PhaPs have also been described (Mezzina et al., 2014). Based on alignments using the Pfam database, four phasin-related families have been found (Finn et al., 2014). The first family, termed Phasin_2 (PF09361), is the largest and includes 952 sequences found in bacteria belonging to Alpha-, Beta-and Gammaproteobacteria. These include PhaP1 from Ralstonia eutropha, PhaP Ah from Aeromonas hydrophila, PhaP Az from Azotobacter and PhaP TD from Halomonas bluephagenesis TD. The second PhaP family (PF09602) is related to phasins found in Bacillus species, and the third one (PF09650) contains a diverse group of mostly uncharacterized proteins found in different Proteobacteria. The last one (PF05597) consists of proteins from different Proteobacteria, including the most typical phasins belonging to Pseudomonas spp., such as PhaF and PhaI from P. putida (Mezzina and Pettinari, 2016).
A small group of PhaPs has been investigated in greatest detail: PhaP Ah from Aeromonas hydrophila strain 4AK4 has been employed as a protein purification tag that binds to PHA particles for easy recovery, as a drug delivery tool, as a coating agent and as a biosurfactant (Moldes et al., 2004;Banki et al., 2005;Backstrom et al., 2007;Wang et al., 2008). The PhaP-based protein purification system was successfully used to inexpensively purify enhanced green fluorescent protein (eGFP), maltose-binding protein and b-galactosidase . Additionally, PhaP Ah has been shown to be a highly efficient biosurfactant (Wei et al., 2011). In a further potentially groundbreaking application, PhaP Ah fused with targeted cell ligands has been reported to specifically bind to cancer cells . PhaP Az from Azotobacter sp. FA-8 is reported to not only improve growth and PHB accumulation in recombinant Escherichia coli, but also protect non-PHB synthesizing E. coli under both normal and stress conditions, as evidenced by a reduction in heat-shock protein levels, leading to better growth and stronger resistance to both heat shock and superoxide stress (de Almeida et al., 2007(de Almeida et al., , 2011. PhaP Az was also proven to help protein folding by preventing spontaneous thermal aggregation in a chaperone-like manner, both in vivo and in vitro (Mezzina et al., 2014). Due to the many potential applications, research into industrial production of phasins is also ongoing. For example, Halomonas bluephagenesis TD01 was used for high yield expression of its own PhaP TD in an economical way (Lan et al., 2016), and it was found to be a suitable biosurfactant.
Even though numerous phasins have been characterized, only the structure of PhaP Ah has been solved. It was shown to be a tetramer with 8 a-helices adopting a coiled-coil structure, with each monomer having a hydrophobic and a hydrophilic surface, which explains its surfactant properties (Zhao et al., 2016). Although PhaP Az and PhaP TD both have remarkable properties with interesting biotechnological applications, no study has been conducted to clarify the corresponding structures, other than theoretical predictions of their secondary structure (Mezzina et al., 2014).
To address this issue, we established homology models of PhaP Az and PhaP TD based on the crystal structure of PhaP Ah , which allowed us to reveal the biochemical and functional mechanisms of these two PhaPs. Rational mutations were designed based on these homology models, and the enhanced PhaP stability and surfactant properties of the resulting mutant proteins confirmed the validity of the structures. We hope that this publication will provide a generalized method for the homology modelling of PhaP structures to facilitate further studies.

Results
Secondary structure analysis of PhaP Az and PhaP TD The PhaP Az and PhaP TD proteins were successfully expressed in E. coli BL21 (DE3) and Halomonas bluephagenesis TD01 respectively. They were purified using affinity-and size-exclusion chromatography, and their purity confirmed using SDS-PAGE (Fig. S1). Static light scattering (SLS) studies indicated that the average molecular weight of PhaP Az was 90 kDa (Mezzina et al., 2014). As the PhaP Az monomer weight was predicted to be 21 kDa, and based on their overall similarity, we expected both PhaPs to be tetramers in aqueous solution. However, SLS revealed the molecular weight of PhaP TD to be 48 kDa (Fig. S2). As the predicted molecular weight for a PhaP TD monomer is approximately 15 kDa, PhaP TD may be a trimer in solution. How these two PhaPs evolved different oligomerization remains to be elucidated.
In silico predictions of phasin secondary structures have found a general feature of phasinsa high percentage of amino acids arranged in an a-helix conformation (Mezzina et al., 2014), together with coiled-coil interacting regions (Maestro et al., 2013;Mezzina and Pettinari, 2016). This common structure suggests that the mode of oligomerization could be determined by the coiled-coil regions. Using MARCOIL (Delorenzi and Speed, 2002;Mezzina et al., 2014), PhaP Az was found to have a maximum probability of 32% of coiled-coil regions for homo-oligomerization, while PhaP TD contains a high coiled-coil probability along its entire sequence, which was similar to PhaP Ah (Fig. 1A).
Thus, the oligomerization of both PhaP Az and PhaP TD may be attributed to coiled-coil regions similar to Pha-P Ah , as evidenced by the structure predictions. As all phasins share the unique ability of binding to hydrophobic granules, analysis of hydrophobic domains is important. Interestingly, hydrophobic cluster analysis (HCA) of PhaP Az and PhaP TD showed that these two proteins do not possess clear hydrophobic domains (Fig. 1B) (Mechin et al., 1995). In fact, hydrophobicity-modulating mutations of PhaP Re were reported to implicate the entire protein in interactions with the PHA polymer granules, without a distinct region responsible for this interaction (Neumann et al., 2008). Nevertheless, judging by the only resolved crystal structure of a phasin (Zhao et al., 2016), a clear distribution of hydrophilic and hydrophobic residues is found on two opposite sides of the PhaP Ah monomer, forming an intrinsically amphiphilic polypeptide without a distinct hydrophobic core (Zhao et al., 2016). These clues suggest that PhaP Az and PhaP TD may possess the same pattern as was observed in PhaP Ah . A sequence alignment has been conducted to find similarities among PhaP Az , PhaP TD and PhaP Ah (Fig. 1C). However, a clear sequence homology could not be found even though their amphiphilic properties had a similar pattern. Nevertheless, blocks of hydrophobic amino acids on the bona fide hydrophobic surface of PhaP Ah (shown in green, Fig. 1C) were reflected in hydrophobic amino acids in the corresponding positions of PhaP Az and PhaP TD . At the same time, hydrophilic amino acids were found in the hydrophobic surface of PhaP Ah (shown in red, Fig. 1C), as was also found in PhaP Az and PhaP TD . A majority of these hydrophilic residues were conserved glutamines.
Rational mutagenesis based on the homology models of PhaP Az and PhaP TD Based on secondary structure analysis of PhaP Az and PhaP TD , a strong structural similarity was found between these two proteins and PhaP Ah . Accordingly, homology models of these proteins were established based on the crystal structure of PhaP Ah using SWISS-MODEL ( Fig. 2A and B) (Biasini et al., 2014). These two modelled structures clearly demonstrated an intrinsic amphiphilic property, even though they did not show clear hydrophobic domains. A very significant phenomenon was observed in both structures via surface electrostatic potential analysis, regarding the distribution of hydrophilic and hydrophobic residues on two opposite sides of individual PhaP Az and PhaP TD monomers ( Fig. 2A and  B). There are 20 hydrophobic amino acids forming the hydrophobic surface of PhaP Az ( Fig. 2A). Most of these residues are leucines and phenylalanines, with hydrophobic residues also found in the corresponding positions of PhaP Ah (Fig. 1C). By contrast, there are 27 hydrophobic amino acids forming the hydrophobic surface of PhaP TD (Fig. 2B), and most of these residues are methionines. The corresponding positions are also covered by hydrophobic residues in PhaP Ah (Fig. 1C). It seems that most of the hydrophobic amino acids are folded on the same layer to form a hydrophobic surface through conformational changes of the a-helices. This may be the reason why phasins have the ability to bind to PHA inclusion bodies, even though PhaP primary structures lack clear hydrophobic domains.
Based on the structural information of PhaP Az and PhaP TD , two significant hydrophilic amino acids embedded in the hydrophobic surfaces were identified as targets for rational mutagenesis of each monomer, which may improve their stability and amphiphilic properties ( Fig. 2A and B). The hydrophilic glutamines Q38 and Q78 of Pha-P Az were mutated to hydrophobic leucines ( Fig. 2A) the most abundant hydrophobic residues in this structure. Subsequently, the hydrophilic glutamines Q38 and Q72 of PhaP TD were mutated to hydrophobic methionines ( Fig. 2B), which again were the most abundant hydrophobic residues in the second structure. The two mutant proteins were successfully expressed and purified from E. coli. These mutations might be able to enhance the stability of their respective tetramers by increasing hydrophobic interactions, as was observed in PhaP Ah (Zhao et al., 2016). Furthermore, our aim was also to obtain enhanced emulsification properties by introducing additional hydrophobic amino acids into the hydrophobic surface.

Rational mutations of PhaP Az and PhaP TD resulted in improved stability
Protein melting temperatures (T m ) were used to study the thermostability of the rational mutants and their respective wild-type parents. The PhaP proteins were gradually heated to 100°C and thermal denaturation was monitored using circular dichroism (CD) spectroscopy to obtain the melting temperatures (T m ) . Two PhaP Az mutants showed higher T m values than the wild type, with approximately 8 degree increase for PhaP Az Q38L and 6 degree increase for PhaP Az Q78L (Fig. 3A). At the same time, significant thermostability increases were observed for PhaP TD mutants, with approximately 10 degree higher T m values for both PhaP TD Q38M and PhaP TD Q72M (Fig. 3A). Importantly, an analysis of the mutants' 1-anilinonaphthalene-8-sulfonate (ANS) binding affinity also showed similar results (Fig. 3B). The circular dichroism results have shown that these proteins still kept some secondary structures rather than complete melting at 100°C (Fig. S3).Wild-type PhaP Az exposed its hydrophobic sites at 50°C, indicating melting at this temperature. On the other hand, its derived mutants Pha-P Az Q38L and PhaP Az Q78L did not show any significant changes at 50°C, and rapid melting was evidenced only at 60°C (Fig. 3B). Similarly, hydrophobic sites of wild- type PhaP TD were exposed at 50°C, while the mutants PhaP TD Q72M and PhaP TD Q38M showed significant changes only at 60°C and 70°C respectively (Fig. 3B). Therefore, significantly enhanced thermal stability was achieved through targeted mutations of the two PhaPs. The observed T m differences are most likely due to closer hydrophobic chain interactions.
As phasins are commonly located on the hydrophobic surface of PHA granules, PhaP structures may be affected by hydrophobic environments, and PhaP Az has already been proved to change its secondary structure in the presence of sodium oleate (Mezzina et al., 2014). It was estimated that approximately 40-45% of the residues in PhaP Az were disordered when it was not bound to a target, and this value decreased to between 23 and 30% upon interaction with sodium oleate (Mezzina et al., 2014). This decrease of disordered secondary structures may be caused by the change of PhaP oligomers. As PhaP Ah does not have obvious hydrophobic regions exposed to the solvent in an aqueous solution (Zhao et al., 2016), PhaP Ah and PhaPs in general need to change their structure to expose their hydrophobic surfaces to contact PHA. The change of oligomers appears as a plausible way to achieve this. When sodium oleate was used as a triggering substance to study the oligomer stability of PhaPs, the secondary  structure of wild-type PhaP Az changed in the presence of 8.5 mM sodium oleate, whereas the mutants Pha-P Az Q38L and PhaP Az Q78L maintained their secondary structures at concentrations of up to 10 mM sodium oleate, and their conformations changed significantly only at 12.5 mM sodium oleate (Fig. 3C). On the other hand, the secondary structure of wild-type PhaP TD changed at 0.75 mM sodium oleate, whereas its mutants PhaP TD Q38M and PhaP TD Q72M tolerated 2.50 mM sodium oleate (Fig. 3C). These results further demonstrate that the mutants were able to withstand higher sodium oleate concentrations than the wild type, clearly illustrating that the targeted mutations improved the proteins' conformational stability in hydrophobic environments by enhancing the hydrophobic interactions between monomer chains, as predicted.

Rational mutations of PhaP Az and PhaP TD resulted in improved emulsification properties
Emulsification activity can be measured by analysing a protein's ability to reduce water-oil interfacial tension. The mutants PhaP Az Q38L and PhaP Az Q78L showed better properties in reducing water-oil interfacial tension than the wild type at the same protein concentration, demonstrating that an improved emulsification capacity can be obtained by rational mutagenesis (Fig. 4A). The same was also true for PhaP TDits two mutants showed stronger emulsification activity than the parent. Another way to measure emulsification ability is to study the contact angle of drops comprising aqueous solutions of these proteins placed on hydrophobic surfaces. Equal drops containing the same protein concentrations were carefully applied to a hydrophobic poly(3-hydroxybutyrate-co-3-hydroxyhexanoate) and a PHBHHx surface, after which the contact angles reflecting the PhaP's amphiphilic properties were recorded. On hydrophobic PHBHHx films, droplets containing the Pha-P Az mutants Q38L and Q78L had smaller contact angles than droplets containing the wild type (Fig. 4B). Similarly, PhaP TD Q38M had a smaller contact angle than its wildtype parent, and PhaP TD Q72M had the smallest contact angle of all (Fig. 4B). These results indicate improved emulsification ability of the rational mutants compared with their respective wild-type parents.

Discussion
Homology models of PhaP Az and PhaP TD have been constructed successfully for the first time, based on the known crystal structure of PhaP Ah . The resulting modelled structures enable a rational understanding of the proteins' surfactant properties and the related molecular mechanisms of amphiphilicity. Based on these structures, both PhaP Az and PhaP TD are predicted to have a hydrophobic and a hydrophilic surface in each monomer ( Fig. 2A and B). At the same time, two specific point mutations, for PhaP Az and PhaP TD each, were rationally designed based on the homology-modelled structures. In detail, two hydrophilic glutamines and methionines were mutated to hydrophobic leucines in the hydrophobic surfaces of PhaP Az and PhaP TD respectively ( Fig. 2A and  B). These designed mutations were expected to improve the stability of PhaP Az and PhaP TD due to improved hydrophobic interactions between individual chains. As expected, all mutated PhaPs showed increased melting temperatures, indicating enhanced thermal stability ( Fig. 3A and B). At the same time, the mutants demonstrated significantly improved conformational stability and emulsification ability under conditions identical to those of their respective wild-type parents (Figs 3C and 4). Taken together, these results fully support the practical usefulness of the homology-modelling approach used to generate structural models for PhaP Az and PhaP TD , using the known crystal structure of PhaP Ah .
Although PhaPs do not have a high degree of sequence conservation, they do contain a high percentage of a-helices (Zhao et al., 2006;Maestro et al., 2013;Mezzina et al., 2014), together with coiled-coil interacting regions (Mezzina and Pettinari, 2016). The mode of oligomerization is also dictated by the presence of coiledcoil regions. Even though the absence of distinct hydrophobic domains appears to be a common phenomenon of most phasins (Mezzina and Pettinari, 2016), the distribution of hydrophilic and hydrophobic residues on two opposite sides of the PhaP Az and PhaP TD monomers was clearly visible based on the crystal structure of PhaP Ah ( Fig. 2A and B). The common mechanism of PhaP's amphiphilic properties can thus be attributed to the folding of most of the hydrophobic amino acids surrounding the a-helical domain in a way conductive to the formation of a hydrophobic surface. Furthermore, the PhaP's ability to bind to hydrophobic PHA or inclusion bodies is also most likely due to this hydrophobic surface.
The proportion of a-helices in PhaP Az was observed to change depending on the environment (Mezzina and Pettinari, 2016). This change was attributed to a change in the proportion of disordered regions. There may be two reasons leading to the change of a-helix proportion, the first one being the disassembly of PhaP oligomers, as PhaP in an aqueous solution should self-assemble into oligomers without exposure of its hydrophobic surface, while the presence of hydrophobic molecules would induce PhaPs to disassemble so that their hydrophobic surfaces could bind to these molecules (Zhao et al., 2016). The second reason can be attributed to the existence of two different chain conformations, so that some chains may be distorted during the process of oligomer formation (Zhao et al., 2016).
PhaPs have also been applied as fusion partners for drug targeting and protein purification Grage et al., 2009), as biosurfactants (Wei et al., 2011), and as chaperones for better protein folding (Mezzina et al., 2015), positioning PhaPs as candidate proteins for high-value applications. However, in spite of their impressive functional diversity and biotechnological applications, only the structure PhaP Ah has been resolved to date (Zhao et al., 2016).
As PhaP structures are important for mechanistic studies and rational mutagenesis, a generalized method was developed in this study to construct PhaP structure models based on the known high-resolution structure of Pha-P Ah , greatly facilitating the study of other PhaP members, including secondary structure analysis of PhaPs to find amphiphilic features similarity to those of PhaP Ah . A general PhaP homology-modelling approach using SWISS-MODEL can be applied if the PhaP in question has a highly amphiphilic nature similar to Pha-P Ah , which can reasonably be expected in most cases. The results obtained for PhaP Az and PhaP TD demonstrate the accuracy of this method, paving the way for the analysis of many more PhaPs, hopefully leading to the discovery of new properties and applications.

Experimental procedures
Bacterial strains and plasmids. All bacterial strains and plasmids used in this study are summarized in Table 1. E. coli S17-1 was used as donor strain for the Halomonas TD conjugation process (Fu et al., 2014).
Vector construction and site-directed mutagenesis. For protein expression, the gene encoding PhaP Az (UniProtKB -Q8KRE9) was cloned into vector pGEX-6P-1 and expressed in E. coli BL21(DE3) as a fusion protein with an N-terminal GST tag. The gene encoding PhaP TD (UniProtKB -F7SLK4) was inserted into the vector pSEVA331 with a C-terminal His tag. E. coli S17-1 was selected as the donor for conjugation with Halomonas bluephagenesis TD01.
Site-specific mutations were introduced using the Fast Mutagenesis System kit (Transgen Biotech Company).
Protein expression and purification. The PhaP Az protein was expressed in E. coli BL21 (DE3) grown in LB medium (10 g/L NaCI, 10 g/L tryptone and 5 g/L yeast extract). A final concentration of 0.5 mM isopropyl-b-D-1thiogalactopyranoside (IPTG) was added to the cultures when the OD 600 reached 0.6-0.8, followed by overnight expression at 37°C. The PhaP TD protein was expressed in Halomonas bluephagenesis TD01 grown in 60-LB medium (60 g L À1 NaCI, 10 g L À1 tryptone and 5 g L À1 yeast extract) overnight at 37°C without induction (constitutive expression). Subsequently, cells were harvested by centrifugation at 4000 9 g and 4°C for 20 minutes and resuspended in lysis buffer containing 50 mM Tris (pH 8.0), 500 mM NaCl for PhaP Az , and HBS buffer (1.5 mM Na 2 HPO 4 •2H 2 O, 50 mM HEPS, 150 mM NaCl) for PhaP TD . After cell disruption, the lysates were centrifuged at 13,000 rpm for 1 h. The cleared lysates containing GST-PhaP Az were transferred to a glutathione sepharose 4B column, and the cleared lysates containing PhaP TD -His were transferred to a Ni-NTA agarose column. After washing the column with 5 column volumes (CV) of lysis buffer, the GST-tagged fusion protein was digested in situ using 1 mg/mL 3C PreScission protease, whereas the His-tagged fusion protein was eluted using elution buffer comprising 500 mM imidazole. The released PhaP Az without the GST tag and the His-tagged PhaP TD were further purified by gel-filtration chromatography on a Superdex 200 high-performance column using 50 mM Tris-Cl (pH 8.0), 500 mM NaCl and HBS buffer (1.5 mM Na 2 HPO 4 •2H 2 O, 50 mM HEPS, 150 mM NaCl) respectively. The bed volume was 24 mL and the bed dimension was 109300-310 mm. The injection volume was 1 mL and the flow rate was 0.5 mL min À1 . The same protocols were applied to the respective mutants.
In silico prediction and homology modelling. MARCOIL was used to predict coiled-coil regions (Delorenzi and Speed, 2002;Mezzina et al., 2014), hydrophobic cluster analysis was used to predict hydrophobic domains (Mechin et al., 1995), and homology models based on the crystal structure of PhaP Ah were built using SWISS-MODEL (Biasini et al., 2014).
Protein melting temperature (T m ) measurements using circular dichroism spectroscopy. Circular Dichroism spectroscopy was carried out on a qCD Chirascan-auto instrument equipped with the qBiC Biocomparability Suite, using dynamic multimode spectroscopy (DMS). Heatinduced denaturation of PhaP was directed from 30°C-100°C with a heating rate of 1°C per min. The 220 nm wavelength was used to draw the thermal change graphs. Global 3 software (Applied Photophysics, UK) was used to calculate the melting temperatures. An S curve was applied to fit the data to the following equation (Benjwal, 2006): y=Ab+{(AtÀAb)/(1 + exp[(x 0 -x)/w]} whereby y is the experimentally observed CD signal at a given temperature, x is the fraction of the native state at temperature y, Ab, At and w are constant terms of this equation and X 0 is the transition point of the curve which referring to a protein's melting temperature.
CD measurements were conducted in a buffer comprising 50 mM Tris-Cl (pH 8.0) and 500 mM NaCl for PhaP Az , and HBS buffer for PhaP TD . The protein concentration was 0.5 mg/ml for both the wild-type proteins and the mutants.
ANS fluorescence spectroscopy. An F-2500 fluores cence spectrophotometer was used to record fluorescence spectra. The slit width was 5 nm for both emission and excitation. The excitation wavelength was 380 nm for the extrinsic ANS fluorescence. The 1sulfonic-8-anilinonaphthalene acid (ANS) extrinsic probe was added to the protein solutions at a molar ratio of 75:1 (probe:protein) and incubated at 25°C in the dark for 30 min prior to measurements. ANS binding affinity was measured via emission spectra recorded from 400 to 600 nm. The measurements were conducted in a buffer comprising 50 mM Tris (pH 8.0) and 500 mM NaCl for PhaP Az , and HBS buffer for PhaP TD . The protein concentration was 0.5 mg ml À1 for both wild type and the mutants.
Determination of water-oil interfacial tensionpendant drop method. The water-oil tension between water (PhaP protein solution) and soya bean oil was determined using an optical contact angle measurement and contour analysis system (DataPhysics, Germany). The set-up was used to capture an image of a liquid drop that hangs on a dosing needle and to subsequently analyse it via the DataPhysics SCA 22 software module (DataPhysics, Germany) using the pendant drop method (Berry et al., 2015).
Water contact angle measurements. The OCA20 (the model number of contact angle evaluation system) contact angle evaluation system was applied to measure the water contact angles of protein solution drops which were placed on sample films. Comparison of contact angles of drops comprises 25 mg/L protein solutions of wild-type PhaP Az and PhaP TD and the indicated mutants. Droplets with or without protein in 3 lL of deionized water were carefully placed onto poly (3hydroxybutyrate-co-3-hydroxyhexanoate) or PHBHHx surfaces. The average contact angles were obtained by measuring drops placed on at least three positions of the same film.
Static light scattering (SLS). Protein molecular masses were determined via SLS using the DAWN HELEOSTM II eighteen-angle static light scattering system (Wyatt Technology, USA) connected to a gel-filtration chromatography system equipped with a Superdex 200 high-performance column (see above for specifics of gel filtration). The Superdex 200 high-performance column was flowed by HBS buffer (1.5 mM Na 2 HPO 4 •2H 2 O, 50 mM HEPS, 150 mM NaCl). The system was preequilibrated with buffer for more than 8 h and subsequently calibrated with 1 mg/ml BSA. The samples were prepared as above and concentrated to 1 mg mL À1 , after which they were injected into the SLS analyser at a 0.5 mL minutes À1 flow rate at 16°C. The molecular mass was calculated using ASTRA5.3.4.14 software (Wyatt Technology, USA).

Supporting information
Additional Supporting Information may be found online in the supporting information tab for this article: Fig. S1. SDS-PAGE gels of purified PhaP Az and PhaP TD .