Synthesis of bacteriophage lytic proteins against Streptococcus pneumoniae in the chloroplast of Chlamydomonas reinhardtii

Summary There is a pressing need to develop novel antibacterial agents given the widespread antibiotic resistance among pathogenic bacteria and the low specificity of the drugs available. Endolysins are antibacterial proteins that are produced by bacteriophage‐infected cells to digest the bacterial cell wall for phage progeny release at the end of the lytic cycle. These highly efficient enzymes show a considerable degree of specificity for the target bacterium of the phage. Furthermore, the emergence of resistance against endolysins appears to be rare as the enzymes have evolved to target molecules in the cell wall that are essential for bacterial viability. Taken together, these factors make recombinant endolysins promising novel antibacterial agents. The chloroplast of the green unicellular alga Chlamydomonas reinhardtii represents an attractive platform for production of therapeutic proteins in general, not least due to the availability of established techniques for foreign gene expression, a lack of endotoxins or potentially infectious agents in the algal host, and low cost of cultivation. The chloroplast is particularly well suited to the production of endolysins as it mimics the native bacterial expression environment of these proteins while being devoid of their cell wall target. In this study, the endolysins Cpl‐1 and Pal, specific to the major human pathogen Streptococcus pneumoniae, were produced in the C. reinhardtii chloroplast. The antibacterial activity of cell lysates and the isolated endolysins was demonstrated against different serotypes of S. pneumoniae, including clinical isolates and total recombinant protein yield was quantified at ~1.3 mg/g algal dry weight.


Introduction
The increase of antibiotic resistance in the last few decades has severely reduced the reliability of antibiotic treatments (Levy, 2005). In particular, the prevalence and range of multidrug resistant strains has increased alarmingly (Levy and Marshall, 2004;WHO report 2014), and there has been a simultaneous decline in the development of new antibiotics, mainly owing to the lower profitability of antibiotics compared to other drugs (Spellberg et al., 2008;Wright, 2012). Health agencies, including the World Health Organization (WHO) and the Infectious Disease Society of America (IDSA), have warned that antimicrobial resistance is an 'increasingly serious threat to global public health' and that there is a real risk of falling back into a preantibiotic era in the 21st century (Infectious Diseases Society of America (IDSA) 2011; WHO report 2014). Furthermore, many antibiotics are broad spectrum and therefore each treatment affects the human commensal bacterial flora in addition to the target pathogen. Not only does this increase the patient's susceptibility to opportunistic bacterial and fungal pathogens, there is also increasing evidence that disruption to the natural microbiota can play a key role in conditions such as obesity, type 1 diabetes, inflammatory bowel disease, allergies and asthma (Blaser, 2011;Chen and Blaser, 2007). Repeated exposure to broad spectrum antibiotics can also result in the spread of antimicrobial resistance genes within the human microbiome, which in turn can be passed on to pathogenic bacteria (Schmelcher et al., 2012). With the availability of enhanced diagnostic methods it would therefore be highly desirable to develop narrow spectrum antibiotics that specifically target the pathogen, thus reducing the collateral effects on the commensal bacterial flora and in turn the spread of antimicrobial resistance genes (Blaser, 2011).
Endolysins are bacteriophage-encoded enzymes that accumulate in the cytoplasm of infected bacterial cells throughout the lytic cycle (Loessner, 2005). Towards the end of this cycle, the next generation of phage particles is assembled, and the synthesis of a second phage protein known as a holin is initiated. The holin creates pores in the plasma membrane allowing the endolysin to exit the cytoplasm and attack the cell wall of the infected bacterium (Schmelcher et al., 2012;Young et al., 2000). Cleavage of specific bonds within the peptidoglycan results in the lysis of the host cell and the release of the phage progeny (Loessner, 2005). In the case of Gram-positive bacteria, where the peptidoglycan layer is exposed to the bulk solution, such enzymes have been shown to function as 'exolysins', lysing the target bacterium when applied externally (Schmelcher et al., 2012). Gram-positive targeting endolysins are also typically highly specific for the cell wall of the bacterial species from which they are naturally produced (Borysowski et al., 2006), which allows for their use as targeted antibacterial treatments that specifically eliminate the pathogen without harming the commensal human microbiota, in contrast to many conventional antibiotics.
Bacteriophages and their host bacterial strains have co-evolved over billions of years, with the functionality of the phage endolysin representing a key factor in the ensuing arms race: if a host strain develops effective resistance against endolytic attack, then the phage progeny will not be released and that particular phage lineage will not continue. It is thus inferred that any modern-day phage must necessarily have evolved to target components in the cell wall that are essential for bacterial viability, and as such are highly immutable. It was proposed that the development of bacterial resistance against an endolysinbased antibiotic would be less frequent than that seen for conventional antibiotics (Schmelcher et al., 2012), and indeed attempts to induce resistance by treating bacterial populations to sublethal doses of endolysin over several generations (even in combination with a mutagen) have failed to generate any observable effect (Loeffler et al., 2001;Schuch et al., 2002).
Recombinant endolysins have been successfully tested against several different pathogenic bacteria in vitro and in a range of animal models (Entenza et al., 2005;Loeffler et al., 2001;Schuch et al., 2002). The high activity, specificity and low occurrence of resistance make endolysins highly promising candidates for use as novel antibacterial agents in human and veterinary medicine.
Despite the benefits associated with endolysins, they are yet to achieve clinical adoption. The reason for this is largely considered to be economic; the production costs for recombinant therapeutic proteins are significantly higher than small molecule bioactives such as traditional antibiotics (Dove, 2002). It is therefore desirable to develop an inexpensive production system for recombinant endolysins that also fulfils all safety criteria for the production of therapeutic proteins. Microalgae offer several potential advantages for the production of recombinant proteins over more established expression platforms, including a low cost of cultivation using media with simple nutrient requirements and the ability to rapidly develop and scale-up production. A number of key species have also been granted GRAS (Generally Recognised as Safe) status for human consumptionas such, the removal of any endogenous toxins and infectious agents from the product is not a concern (Dove, 2002;Rasala et al., 2010;Specht et al., 2010). Furthermore, the ability of photosynthetic organisms to grow using just light, inorganic substances and CO 2 has the potential to decrease the overall carbon footprint of the production process. Finally, microalgae can be grown in full containment under sterile conditions in simple photobioreactors (PBRs) (Gimpel et al., 2014;Specht et al., 2010), significantly reducing the potential for release of transgenes to the environment and reducing the risk of environmental contamination of the production system.
The eukaryotic microalga Chlamydomonas reinhardtii offers all the aforementioned advantages together with established techniques for the expression of foreign genes from the chloroplast and nuclear genomes (Almaraz-Delgado et al., 2014;Fuhrmann et al., 1999). Expression in the chloroplast is more attractive as the integration of transgenes into the genome occurs via homologous recombination and so the insertion site can be easily defined. In addition, the levels of recombinant proteins that can be achieved in the organelle are markedly higher compared to expression from the nuclear genome (Potvin and Zhang, 2010). Furthermore, the chloroplast compartment offers several specific advantages for the production of endolysins given its evolution from a cyanobacterial endosymbiotic ancestor and hence its similarity to the bacterial cells in which endolysins naturally accumulate. For example, (i) recombinant proteins are retained and accumulate exclusively within the chloroplast (Tran et al., 2013), but the chloroplast lacks a peptidoglycan cell wall to which the endolysin might bind or cleave during cell breakage, thereby hindering cultivation and purification; (ii) endolysins are likely to have a long-half-life in this environment as the endogenous proteases in the chloroplast are homologues of the bacterial proteases to which endolysins have evolved to be resistant (Adam et al., 2006); (iii) proteins synthesized in the chloroplast are not subject to post-translational modifications such as glycosylation, unlike cytoplasmic production platforms. Indeed, the potential exploitation of chloroplasts of plants and microalgae as platforms for the production of therapeutic proteins such as antimicrobials and vaccines has been highlighted in several recent reviews (Bock, 2015;Bock and Warzecha, 2010;Daniell et al., 2016;Rasala and Mayfield, 2015).
The biosynthesis of active endolysins in the chloroplast of tobacco was demonstrated by Oey et al. (2009a) who showed that the PlyGBS endolysin, which targets Streptococcus pathogens, can accumulate in the leaves to levels as high as 70% of total soluble protein. A second study by the same group similarly showed high level production in tobacco of two further Streptococcus endolysins, Pal and Cpl-1 (Oey et al., 2009b). Additionally, this study highlighted the limitations of trying to over-express endolysin genes using an E. coli platform, as both of Pal and Cpl-1 had a lethal effect on the bacterium, potentially caused by unspecific peptidoglycan cleavage at high concentrations or by binding to the cell wall (Oey et al., 2009b). The tobacco chloroplast has also been employed to make other types of antimicrobial proteins and peptides that are technically difficult to produce in established platforms. For example, the clinically relevant antimicrobial peptides retrocyclin and protegrin have complex secondary structures with multiple disulphide bonds, but are able to assemble into a functional form in the chloroplast (Lee et al., 2011). However, there are major challenges associated with the commercial production of recombinant pharmaceuticals in plants grown in fields or greenhouses, as compared to single cells cultured in contained, sterile fermenter-based systems. These challenges include both compliance with good manufacturing practice (GMP) regulations, and the potentially high cost of purifying product from plant biomass under these GMP regulations (Fischer et al., 2012).
In this study, we have investigated the production of endolysins in the chloroplast of C. reinhardtii to examine the suitability of this microalga as a production platform. We have focussed on the Cpl-1 and Pal endolysins which are specific to the human pathogen Streptococcus pneumoniae, as previous studies have demonstrated the antibacterial efficacy of these endolysins both in vitro and in vivo (Garcia et al., 1983;Grandgirard et al., 2008;Jado et al., 2003;Loeffler et al., 2001). We report the generation of transgenic C. reinhardtii lines expressing codon-optimized genes for Pal and Cpl-1, and the stable accumulation of the endolysins to a level of~1% of total soluble protein (TSP). Both crude cell lysates and the isolated endolysins show antibacterial activity against S. pneumoniae, including clinical isolates with resistance against penicillin and co-trimoxazole.

Results
Creation of marker-less transgenic lines of Chlamydomonas reinhardtii expressing pal and cpl-1 in the chloroplast To investigate the synthesis of endolysins in C. reinhardtii, we created transgenic lines in which a synthetic gene encoding either Pal or Cpl-1 was introduced into the chloroplast genome. The coding regions of the genes were designed based on the codon preference for C. reinhardtii chloroplast genes and synthesized de novo. (Nakamura et al., 2000) ( Figure S1). The coding sequence for a C-terminal human influenza haemagglutinin (HA) epitope tag was also added to each gene to facilitate the detection of the proteins. These synthetic genes are referred to as pal and cpl-1. A pal gene without the HA-tag sequence (referred to as pal(HA-)) was also included to investigate whether the HA-tag has an influence on the stability or activity of the endolysin. As shown in Figure 1a, the genes were cloned into the chloroplast transformation vector pSRSapI that uses the chloroplast psaA-1 promoter and 5 0 untranslated region (UTR) to drive transgene expression (Young and Purton, 2014). To compare the efficiency of expression, pal and cpl-1 were also cloned into a second vector, pASapI, that differs from pSRSapI in that it uses the promoter/ 5 0 UTR from the atpA chloroplast gene (Economou et al., 2014). Each vector is designed to target the transgene cassette to a neutral site in the chloroplast genome downstream of the essential photosystem II gene psbH, with restoration of phototrophy in the DpsbH recipient line TN72 used for selection of transformants (Young and Purton, 2014). Such transformants therefore lack any antibiotic-resistance marker and carry the gene of interest (GOI) as the only section of foreign DNA ( Figure 1a).
We analysed colonies of putative C. reinhardtii transformants for the correct insertion of the endolysin genes into the chloroplast genome and homoplasmicity of the polyploid genome using a three-primer PCR assay (Economou et al., 2014). As shown in Figure 1b, homoplasmic transformants were obtained for pSRSapI_cpl-1, _pal and _pal(HA-). Homoplasmic transformants were also obtained for the pASapI_pal and pASapI_cpl-1 constructs, and also for a control transformant line (TN72_control) lacking the GOI in which TN72 was transformed with the empty pSRSapI vector (data not shown). Furthermore, the correct insertion of the expression cassettes carrying pal and cpl-1 into the lines TN72_pal and Tn72_cpl-1 was confirmed by Southern blot analysis using probes binding within the pal and cpl-1 genes, as well as a probe that binds adjacent to the insertion site of the expression cassette ( Figure S2).

Accumulation of Pal and Cpl-1 in the C. reinhardtii chloroplast
We analysed representative TN72_cpl-1 and TN72_pal lines generated using the pSRSapI constructs for the presence of the recombinant protein. This was carried out on crude cell extracts by Western blot analysis using antibodies against the HA-tag. The Cpl-1 protein has an expected size of 40 kDa and Pal an expected size of 36 kDa, and as seen in Figure 2a, both proteins are readily detected in the extracts, confirming the accumulation of Pal and Cpl-1 in the respective lines. For the detection of untagged Pal, antibodies to Pal were raised in rabbits using two synthetic peptides derived from the protein sequence ( Figure S3) and used to compare Pal levels in the TN72_pal and TN72_pal (HA-) lines. As seen in Figure 2b, a 36-kDa band of similar  Figure 1 Schematic diagram of the transformation system used for the introduction of cpl-1 or pal into the chloroplast genome. (a) Homologous recombination (grey crosses) between chloroplast sequences (thick black lines) on the transformation vector pSRSapI and the genome of the recipient strain TN72 allows the introduction of the gene (under the control of the psaA exon 1 promoter/5 0 UTR and rbcL 3 0 UTR from C. reinhardtii) into a neutral locus between psbH and trnE2.
Recombination also replaces the aadA-disrupted copy of psbH with a functional copy of this essential photosystem II gene, allowing transformants to be selected by restoration of phototrophic growth (Wannathong et al., 2016). (b) PCR screening of putative transformants for correct insertion of cpl-1, pal(HA-) and pal and homoplasmicity of the polyploidy chloroplast genome using a set of three primers. Primer F1 (black triangle in (a)) binds to the genome outside of the recombination region. Primer R1 (grey triangle) binds within the aadA cassette of TN72 and results together with primer F1 in a product of 0.88 kb. Primer R3 (white triangle) binds within the gene cassette of the transformants and results together with primer F1 in a product of 1.13 kb. Homoplasmy is scored by a band at 1.13 kb and an absence of a 0.88 kb band arising from untransformed copies of the plastome (Wannathong et al., 2016).
intensity is seen for both the HA-tagged and untagged protein indicating that the HA-tag does not interfere with the stability of this endolysin. Furthermore, we did not observe any reduction in growth rate or biomass production of the Pal and Cpl-1 producing strains in comparison with TN72_control, indicating that the production of the endolysins in the chloroplast is not detrimental to C. reinhardtii growth (shown for TN72_pal in Figure S4). When we compared the levels of Pal and Cpl-1 synthesized using the atpA promoter/5 0 UTR with those from the psaA-1 promoter/5 0 UTR, a higher level was seen for both proteins when using the psaA-1 elements ( Figures S5 and S6), as has been reported previously (Michelet et al., 2011). We therefore used the transgenic lines with the psaA-1 promoter/5 0 UTR for all further studies.
To investigate the stability of the endolysins in the C. reinhardtii chloroplast, we grew cultures of TN72_pal to mid-log phase and specifically inhibited further protein synthesis in the chloroplast with the addition of chloramphenicol. As shown in Figure 3, there is very little decrease in the amount of Pal in the chloramphenicol-treated cultures compared to the untreated controls even after 71 h, and the protein was still detectable 122 h after the start of the inhibition. In contrast, the endogenous D1 protein (a component of photosystem II) decreased to undetectable levels 7 h after the start of the inhibition. In addition, we found that the stability of Pal was further increased during cultivation in the dark (Figure 3). This indicates a high stability of Pal in the C. reinhardtii chloroplast, as has been reported previously for both Pal and Cpl-1 when produced in the tobacco chloroplast (Oey et al., 2009b).

Isolation and quantification of Pal and Cpl-1
For the development of the C. reinhardtii chloroplast as a production platform for therapeutic endolysins, it is important to show that the proteins can be extracted and purified in a biologically active state. Furthermore, it was necessary to produce pure standards for the quantification of Pal and Cpl-1 levels within the algal cell. We isolated the endolysins from crude cell extracts as described in the methods, using ultracentrifugation followed by chromatography with the weak anion exchanger diethylaminoethyl (DEAE) cellulose and choline as specific eluent. Pal and Cpl-1 bind to choline in the S. pneumoniae cell wall, and DEAE is a choline analogue which is commonly used for the affinity chromatography of choline binding proteins (Jado et al., 2003;Sanz et al., 1988). As the final step, we concentrated Pal and Cpl-1 by either ammonium sulphate precipitation or with centrifugal concentrators. As seen in Figure 4, the Pal and Cpl-1 isolated from the algal biomass using this procedure are in a near purified state with only trace amounts of other proteins, including a 55-kDa protein which is most likely the large subunit of the highly abundant Rubisco enzyme.
To optimize the yield of recombinant endolysin, we first analysed the level of Pal in the algal cell at different growth stages and conditions. This indicated that under mixotrophic growth conditions, the concentration of Pal on a per cell basis is highest during the logarithmic phase and then declines once the cells enter stationary phase ( Figures S5, S7 and S8). However, different experiments suggested that a shift from light to dark (i.e. heterotrophic conditions) can limit this decline ( Figures S7 and 3), possibly due to the inactivity of chloroplast proteases that are dependent on ATP generated through photo-phosphorylation as reported by Preiss et al. (2001). However, heterotrophic cultivation of the transformant line from the start of inoculation resulted in an overall reduction in Pal yield per culture volume because of a reduction in cell density under these conditions when compared to mixotrophic growth ( Figure S8). Hence, the maximal yield per culture volume was achieved towards the end of the logarithmic phase under conditions of mixotrophic growth. To quantitate the amount of recombinant protein, we performed Western blot analyses with a dilution series of the isolated endolysins alongside crude cell lysates of TN72_pal and TN72_cpl-1 taken from mid-and late-logarithmic phase cultures (Figure 5a). Using the Odyssey â imaging system (LI-COR) to quantify the amount of antigen in the samples (Figure 5b), we found that both Pal and Cpl-1 represent~1% of total soluble protein (TSP) in C. reinhardtii. This equates to~1 mg of recombinant protein per litre of culture volume at the end of the logarithmic phase during mixotrophic cultivation, or~1.3 mg per gram of algal dry weight (Table 1).

Antibacterial activity of Pal and Cpl-1
To analyse whether Pal and Cpl-1 produced in the C. reinhardtii chloroplast are biologically active, we performed turbidity reduction assays (TRA) with the target bacterium Streptococcus pneumoniae. In a TRA, the lytic activity of an enzyme is measured as a decrease in optical density (OD 595 ) of a bacterial suspension.
Deoxycholate, which induces rapid autolysis of S. pneumoniae, was used as a positive control (Mellroth et al., 2012). As seen in Figure 6, after the addition of a crude extract from TN72_pal or TN72_cpl-1, the OD 595 of a S. pneumoniae suspension decreased in comparison with assays using crude extract from TN72_control or untreated controls. In control assays, the OD 595 showed a gradual decline, which is most likely caused by autolysis of S. pneumoniae. The lytic activity of Pal and Cpl-1 was confirmed in more than 10 independent TRAs. Crude extracts containing either HA-tagged or untagged Pal caused comparable rates of lysis and decreased the OD 595 to a similar extent in five independent TRAs. This suggests that the HA-tag does not interfere with the enzymatic activity of the endolysin. Furthermore, we found that the extracts containing Pal and Cpl-1 were active against different clinical isolates of S. pneumoniae, with all four serotypes (6A, 6B, 6C and 19F) showing lysis including strain ST4157 that is intermediate resistant to penicillin (MIC (minimal inhibitory concentration): 0.125 lg/mL) and resistant to co-trimoxazole (MIC:   (Figure 7a and S9). The specificity of Pal was demonstrated by performing TRAs using Escherichia coli, Staphylococcus aureus and Streptococcus pyogenes. As shown in Figure 7b and S10, the algal extract containing Pal, which was active against S. pneumoniae cells, did not have a measurable effect on the tested bacteria. This suggests that the algal-produced Pal has specificity for S. pneumoniae as described for bacteria-produced Pal and other endolysins specific to Gram-positive bacteria (Loeffler et al., 2001;Schmelcher et al., 2012).
To demonstrate directly that Pal and Cpl-1 kill S. pneumoniae, we determined the number of colony-forming units per mL (cfu/ mL) remaining in a cell suspension after endolysin treatment ( Figure 8). One complication we encountered in initial experiments using the crude algal extracts was the presence of an endogenous bactericidal activity in C. reinhardtii, as has been previously reported (Ghasemi et al., 2007;Jørgensen, 1962). While this activity did not cause cell lysis (see the TN72_control in Figure 6), it did have a marked effect on the viability of S. pneumoniae and therefore masked the effect of the endolysins in the titre assay (data not shown). To demonstrate the killing of S. pneumoniae by Pal and Cpl-1 without the interference of this activity, we repeated the assays with the isolated protein. The addition of 25 lg/mL of Pal or 20 lg/mL Cpl-1 reduced the bacterial titre of strain 16 NP3 by nearly 4 log 10 units (Pal: 3.8 AE 0.5 and Cpl-1: 3.9 AE 0.2 Dlog 10 cfu/mL) in comparison with an untreated control. Treatment of S. pneumoniae D39 with 22.5 lg/mL of Pal reduced the cfu/mL by 3.6 AE 0.3 log 10 units. A treatment with 45 lg/mL of Pal was sufficient to kill all S. pneumoniae cells in suspensions of D39 and 16 NP3, which was a reduction by more than 7 log 10 units for both strains. This supports the TRA results that the recombinant Pal and Cpl-1 are able to effectively kill S. pneumoniae through cell lysis. Furthermore, it shows that the two endolysins retain antibacterial activity during the purification procedure. (a) Cell lysates of TN72_pal and TN72_cpl-1 were compared to a dilution series of known amounts of isolated Pal or Cpl-1 in Western blot analysis using anti-HA antibodies and IRDye â secondary antibodies. The amount of total soluble protein (TSP) in the cell lysates is stated above each lane. For each strain, 'Culture 1' was grown to an OD 750 of 2.5 and 'Culture 2' to an OD 750 of 3.0. The cultures were concentrated five times (lane 1) or 2.5 times (lane 2) in comparison with the initial culture volume before preparation of the cell lysates. (b) The IR fluorescence signals of the serial dilutions were plotted against the protein concentration of the isolated endolysin preparations (taking impurities in the isolated proteins into account) and the equations of the resulting trend lines used to calculate the concentration of Pal/Cpl-1 in the cell lysates.

Discussion
In this study, we investigated the synthesis of two bacteriophage endolysins specific to the human pathogen S. pneumoniae in the chloroplast of C. reinhardtii. We found that Pal and Cpl-1 accumulate as predominately full-length proteins in the transgenic lines. Furthermore, for Pal we demonstrated that this endolysin is highly stable despite the presence of a suite of proteases within the algal chloroplast (Adam et al., 2006). The antibacterial activity of Pal and Cpl-1 against different serotypes of S. pneumoniae, including clinical isolates, was demonstrated using both cell lysates and the isolated proteins. Incubation with 25 lg/mL Pal or 20 lg/mL Cpl-1 resulted in a decrease in colonyforming units (cfu) for S. pneumoniae of nearly four log 10 units. These activities are similar to the antibacterial activities and minimal inhibitory concentrations (MIC) described for Pal and Cpl-1 synthesized in E. coli or tobacco (Loeffler and Fischetti, 2003;Loeffler et al., 2001;Oey et al., 2009b;Rodr ıguez-Cerrato et al., 2007), which suggests that Pal and Cpl-1 are produced in a predominately active form in the algal chloroplast. The production of the endolysins did not have any negative impact on the viability of the alga suggesting that the production in C. reinhardtii could be performed using continuous cultivation systems. Both proteins account for approximately 1% of total soluble protein (TSP) during mixotrophic cultivation, which equates to~1.3 mg of recombinant endolysin per gram of cell dry weight. These yields are similar to previous reports on the production of recombinant proteins in the C. reinhardtii chloroplast, which range from 0.1% to 5% of TSP (Dreesen et al., 2010;He et al., 2007;Manuell et al., 2007;Rasala et al., 2010;Sun et al., 2003;Tran et al., 2013), although significantly lower than yields reported for tobacco chloroplasts which are typically 10%-30% of the TSP in leaves, and in extreme cases are as much as 70% (Bock, 2015). The reason for this difference is not clear, but might reflect the much tighter anterograde regulation of chloroplast gene expression by the numerous nuclear-encoded factors in the single chloroplast of the algal cell (Douchi et al., 2016;Lefebvre-Legendre et al., 2015). Many of these factors act on the 5 0 UTR of specific gene transcripts and are required for RNA stability or translation initiation; they therefore also regulate transgenes driven from the same 5 0 UTR. Competition between endogenous and foreign transcripts with the same 5 0 UTR for low abundance factors might therefore limit the level of recombinant protein. Indeed, higher yields were achieved using the psbA promoter/5 0 UTR to drive transgene expression in a psbA deletion mutant (Manuell et al., 2007;Rasala et al., 2010). However, the generation of nonphotosynthetic transgenic lines negates the key value of the alga as a light-driven cell factory; we have therefore focused on maximizing endolysin production in a photosynthetic line. Further studies of the mechanisms of anterograde regulation in the C. reinhardtii chloroplast combined with the redesign of UTR elements to eliminate regulatory elements (Specht and Mayfield, 2013) should result in significant improvements in protein production levels (Rasala and Mayfield, 2015).
Microalgae are gaining increasing attention as alternative expression platforms due to the simple and cheap nutrient requirements, as well as easy and cost-effective scalability, which will potentially result in low production costs for recombinant proteins (Barrera and Mayfield, 2013;Gimpel et al., 2014). Furthermore, the GRAS status of several green algal species including C. reinhardtii, which lack endotoxins and infectious agents, might decrease downstream processing costs and enables the use of dried algal biomass for the administration of therapeutics and vaccines (Dreesen et al., 2010;Gregory et al., 2012). However, C. reinhardtii cultures reach lower final cell densities and have lower recombinant protein yields than conventional expression systems such as mammalian cells, yeast or bacterial systems (Andersen and Krummen, 2002;Jørgensen et al., 2014). Therefore, to determine whether the C. reinhardtii chloroplast has the potential to be cost-competitive with conventional expression platforms, an accurate costing for the production of recombinant proteins at scale in C. reinhardtii needs to be generated.
This study represents a first report and 'proof of concept' for the production of endolysins in the C. reinhardtii chloroplast. The use of endolysins as antibacterial agents is a relatively new, but potentially immense field. It is estimated that, in total, 10 31 bacteriophages exist globally (Abedon, 2008), so there is a huge untapped resource of endolysins that could be investigated for their potential use as antimicrobials in a wide range of applications (Borysowski et al., 2006). A recent study even describes an endolysin that can traverse epithelial membranes and eliminate intracellular Streptococcus pyogenes cells (Shen et al., 2016). Additionally, several research groups have started to create designer endolysins by removing cell wall binding domains, adding additional catalytic domains or combining domains from different endolysins (Schmelcher et al., 2012). Furthermore, endolysins combined with polycationic peptides have been created that are effective against Gram-negative bacteria. These enzymes are referred to as artilysins (Briers et al., 2014). The C. reinhardtii chloroplast might offer advantages especially for the production of endolysins and artilysins specific to Gram-negative bacteria. These enzymes are less specific and can have adverse effects on bacterial production systems containing the peptidoglycan substrate (Briers et al., 2014;Oey et al., 2009b). The successful production in the C. reinhardtii chloroplast of a Gram-negative endolysin that cannot be produced in bacterial systems would demonstrate a clear advantage of this microalga as an expression platform for endolysins.
Overall it is acknowledged that the field of therapeutic protein production in the green microalga C. reinhardtii is in its infancy, and at present is not a viable alternative to more mature technologies. That being said, the use of an algal platform does hold a number of intrinsic advantages over other systems including cost of cultivation and sustainability. The path to greater adoption of this and similar platforms will likely start with niche applications specifically tailored to the unique characteristics exhibited by the algal host; the particular suitability of the C. reinhardtii chloroplast for recombinant endolysin expression can be seen as a first step on this road.

C. reinhardtii strains and growth conditions
The recipient strain used to create the transformant lines was TN72, a psbH deletion mutant of the cell wall-deficient C. reinhardtii strain cw15.mt+ (Wannathong et al., 2016). TN72 was maintained on Tris-acetate-phosphate (TAP) medium at 20°C and 5-10 lmol/m 2 /s light, as were transformant lines following selection on high salt minimal (HSM) medium (Harris, 2009). Liquid cultures were grown in Erlenmeyer flasks at 100-200 lmol/m 2 /s, 120 rpm shaking and 25°C in either illuminated shaking incubators or Algem photobioreactors (Algenuity, Stewartby, UK) for growth studies. Growth was measured by recording the optical density at 750 nm with a Unicam UV/Vis Spectrometer (Thermo Electron Corporation, USA) or 740 nm in the Algem photobioreactor.

Bacterial strains and growth conditions
Activity assays were performed with the S. pneumoniae strain 16 NP3 (serotype 19F) or D39. Additional assays were with clinical isolates: ST65 = serotype 6A, H08212 0259; ST176 = serotype 6B, H08052 0052; ST1390 = serotype 6C, H05252 0075; ST4157 = serotype 6B, 35 NP1. The S. pneumoniae strains and S. pyogenes (ATCC 19615, Streptococcus group A) were grown on Columbia blood agar plates overnight at 35°C under anaerobic or aerobic conditions, respectively. Liquid cultures were grown in trypticase soy broth (30 g/L trypticase soy broth, 3 g/L yeast extract) overnight at 35°C without shaking. Anaerobic conditions were generated using Oxoid AGS CO 2 Gen Compact gas packs and an anaerobic jar from Oxoid. S. aureus (ATCC 28213) liquid cultures were grown in ISO-Sensitest broth from Oxoid overnight at 35°C with shaking. All target bacteria were obtained from the Royal Free Hospital (London, UK).

Plasmid construction
Endolysin coding sequences (including a C-terminal human influenza haemagglutinin (HA) tag sequence) were codonoptimized for the C. reinhardtii chloroplast to a Codon Adaption Index (CAI) of 0.8 using the Kazusa CAI table (Nakamura et al., 2000) and synthesized de novo by GENEART (Regensburg, Germany). A 5 0 SapI and a 3 0 SphI restriction site were included in each synthetic gene to facilitate cloning into the chloroplast expression vectors pASapI (atpA promoter/5 0 UTR used to drive GOI expression) and pSRSapI (psaA exon 1 promoter/5 0 UTR) as described in Wannathong et al. (2016). Plasmids were prepared from E. coli DH5a by alkaline lysis (Sambrook and Russell, 2001) and checked by SphI digestion and by Sanger DNA sequencing (Wolfson Institute for Biomedical Research, University College London). For algal transformation, larger plasmid preparations were made using a QIAfilter Plasmid Midi kit (Qiagen).

Transformation of C. reinhardtii
The endolysin genes were introduced into the psbH-trnE2 intergenic region of the chloroplast genome of TN72 by homologous recombination, restoring a functional psbH gene and the ability to grow photosynthetically. The transformation procedure was based on that described by Kindle (1990). Cells were grown to mid-log phase (1-2 9 10 6 cells/mL) in TAP medium, harvested by centrifugation (8000 g, 8 min) and resuspended in TAP medium to 2 9 10 8 cells/mL. The cell suspension (300 lL per transformation) was vortexed for 15 s together with 0.3 g of glass beads (Sigma-Aldrich, diameter 0.4-0.6 mm) and 2-20 lg of circular plasmid DNA. Molten top agar (3.5 mL HSM + 0.5% agar) at 42°C was mixed with the cell suspensions and poured onto HSM plates. The plates were incubated for 1 h in the dark, followed by 3-4 weeks at 20°C under 50 lmol/m 2 /s light. To confirm gene insertion and homoplasmicity of transformants, DNA was extracted using the Chelex 100 method as described by Werner and Mergenhagen (1998), and 2 lL of the DNA extract used in PCR reactions with the primers F1, R1 and R3 (Young and Purton, 2014).

Southern blot analysis
Genomic DNA (4 lg) of the recipient strain TN72, TN72_pal and TN72_cpl-1 was cut with the restriction enzymes SphI and EcoRI and separated on an agarose gel (1% w/v). The DNA was transferred to a Hybond N (Amersham Biosciences) membrane as described in Sambrook and Russell (2001). Subsequently, the blot was hybridized with DNA probes (produced either by PCR or the digest of a plasmid) that bind within the pal and cpl-1 genes, as well as a probe that binds just before the insertion site of the expression cassette in all strains. DNA labelling and detection was performed with the DIG High Prime DNA Labelling and Detection Starter Kit II (Roche Diagnostics GmbH) according to the manufacturer's instructions.

SDS-PAGE and Western blot analysis
SDS-PAGE was carried out using the Mini-PROTEAN Tetra cell system (Bio-Rad) and gels based on the recipe by Laemmli (1970) containing 15% acrylamide. Samples were supplemented with sample loading buffer, boiled at 99°C for 3 min and centrifuged at 21 000 g for 2 min. The gels were run at 120-150 V for 90-120 min. To visualize all proteins, the gels were stained with Coomassie Brilliant Blue R solution for 1 h and destained for at least 2 h. The stained gels were scanned using the Odyssey â imaging system from LI-COR.
For Western blot analyses, the proteins were transferred to Hybond-ECL nitrocellulose membranes (GE Healthcare), at 19 V for 1 h using a Trans-Blot SD semi-dry electrophoretic transfer cell (Bio-Rad). The membranes were blocked overnight in TBS-T (TBS + 0.1% Tween) with 0.5% milk and incubated with the primary antibody (a-HA antibody from rabbit (Sigma-Aldrich product H6908) diluted 1:2000 or anti-Pal antibodies) for 1-3 h followed by incubation with the secondary antibody for 1 h. Both antibodies were resuspended in TBS-T + 0.5% milk, and the membranes were washed after each incubation 39 in TBS-T for 5-10 min. When ECL IgG horseradish peroxidase-linked secondary antibodies (GE Healthcare, 1:10 000) were used, the membranes were incubated with SuperSignal â West Pico Chemiluminescence Substrate (Thermo Scientific) for 5 min and exposed to Hyperfilm ECL (GE Healthcare). For quantitative Western blot analyses, IRDye â secondary antibodies (Dylight TM 800, Thermo Scientific, 1:20 000) and the Odyssey â Infrared Imaging system (Li-COR Biosciences) were used. Custom made anti-Pal antibodies were raised by Eurogentec (Belgium) using two peptides of 17 amino acids (see Figure S2) attached to keyhole limpet haemocyanin carrier protein.

Preparation of C. reinhardtii extract
Chlamydomonas reinhardtii cultures were grown to an OD 750 of 2-3, harvested by centrifugation at 8000 g for 10 min and resuspended in 20 mM sodium phosphate (Na-Pi) buffer (pH 6.9) + protease inhibitor (Roche cOmplete, EDTA-free). Cells were either broken by three cycles of freezing and thawing (liquid nitrogen, 30°C) or sonication with the Cup Horn (Qsonica, LLC), followed by centrifugation at 21 000 g for 5 min (=crude extract).

Endolysin isolation and quantification
After cell breakage, cell suspensions were centrifuged at 5000 g for 20 min, followed by ultracentrifugation at 100 000 g for 1 h. Ultracentrifugation extracts were applied to a diethylethanolamine (DEAE) cellulose column. The column was washed with six column volumes (CV) of Na-Pi, eight CV of Na-Pi + 1.5 M NaCl and four CV of Na-Pi + 0.1 M NaCl. Pal and Cpl-1 were eluted with 2 CV of Na-Pi + 0.1 M NaCl + 6.5% (w/v) choline. The elution fractions with the highest amount of Pal and Cpl-1 were determined in protein microarrays (2 lL spots on nitrocellulose membranes, followed by Western blot analysis protocol), combined and dialysed using Slide-A-Lyzer G2 dialysis cassettes (20 000 MWCO) (Thermo Scientific). After dialysis, the elution was concentrated 10-fold by ammonium sulphate (AS) precipitation: Pal was precipitated with 35% AS, and Cpl-1 with 50% AS. Alternatively, the chromatography as described above was performed with an € AKTA pure system (GE Healthcare Life Sciences) and the fractions with the highest amount of endolysin were concentrated with centrifugal concentrators (5000 MWCO). Protein concentrations were determined using Bradford reagent (Sigma-Aldrich) and bovine serum albumin as standard (taking impurities into account).
Endolysin concentration within C. reinhardtii lines TN72_pal and TN72_cpl-1 was quantified using serial dilutions of the isolated Pal or Cpl-1 in Western blot analyses with anti-HA antibodies and IRDye â secondary antibodies. The IR fluorescence signals of the serial dilutions were plotted against the protein concentration (determined by Bradford assay, taking impurities into account), and the equations of the resulting trend lines were used to calculate the amount of Pal/Cpl-1 in cell lysates. The amount of total soluble protein (TSP) in the cell suspensions was determined by Bradford assay after cell breakage and centrifugation at 13 000 g for 15 min (Manuell et al., 2007). The cell dry weight was determined after lyophilization of the cell suspensions for 16 h.

Assays of endolysin activity
S. pneumoniae cultures were grown to mid-log phase, (S. pyogenes, E. coli and S. aureus overnight) harvested and resuspended in 20 mM Na-Pi. Most TRAs were performed at 37°C in 96-well microtitre plates with 20 lL of crude extract in a final volume of 200 lL using a FLUOstar OPTIMA Microplate Reader (BMG Labtech Ltd) recording the OD 595 over times courses of 60-180 min. The assays in Figure 7b and S10 were performed in cuvettes with 50 lL of crude extract in a final volume of 1 mL at room temperature. For the preparation of the crude extracts used in the TRAs, cultures of TN72_pal, TN72_cpl-1 and TN72_control were grown to the same OD 750 , harvested by centrifugation and resuspended in 1/100 of the culture volume. After cell breakage, the suspension was centrifuged at 21 000 g for 5 min and the supernatant was used as crude extract in the TRAs.
To determine the colony-forming units (cfu) that survived the endolysin treatment, samples from TRAs with the isolated endolysins were diluted in serial 10-fold dilutions and plated in triplicate onto Columbia blood agar plates. After incubation overnight, the cfu were counted and the decrease in cfu/mL calculated relative to an untreated control. For the cfu assays with S. pneumoniae D39, frozen aliquots (-80°C) from an overnight culture were thawed and incubated with isolated endolysin at 37°C for up to 200 min. The decrease in cfu was determined as described above.

Supporting information
Additional Supporting Information may be found online in the supporting information tab for this article: Figure S1 The pal and cpl-1 sequences codon-optimised for the C. reinhardtii chloroplast. Figure S2 Southern blot analysis confirming the predicted insertion of pal and cpl-1 into the plastome of TN72. Figure S3 Protein sequence of Pal showing the peptide sequences used to raise antibodies. Figure S4 Growth of C. reinhardtii TN72_pal compared to the TN72_control. Figure S5 Comparison of the amount of Pal produced using the atpA or psaA exon 1 promoter/5 0 UTR elements. Figure S6 Comparison of the amount of Cpl-1 produced using the atpA or psaA exon 1 promoter/5 0 UTR elements. Figure S7 Yield of Pal at different growth stages and the influence of darkness on the protein stability during stationary phase. Figure S8 Yield of Pal per culture volume at different growth stages during mixotrophic and heterotrophic growth. Figure S9 Turbidity reduction assay showing the lytic activity of crude cell extracts containing Pal and Cpl-1 against clinical isolates of S. pneumoniae. Figure S10 Turbidity reduction assays showing the specific lytic activity of Pal produced in C. reinhardtii against Streptococcus pneumoniae compared to Escherichia coli, Streptococcus pyogenes and Staphylococcus aureus.