Novel processes to obtain pneumococcal surface proteins for vaccines

Abstract

Abstract

Current pneumococcal vaccines are based on the protection offered by capsular polysaccharides from only a few from > 100 serotypes; therefore, serotype-independent vaccines composed of pneumococcal surface proteins are being developed. Despite the immense number of publications on the discovery, characterization, and evaluation of new pneumococcal vaccine candidates, there are very few that describe the bioprocess development, which is an essential step to generate material for pre-clinical and clinical tests, to obtain enough protein amount for physical–chemical, biochemical, and biological characterization, and to understand critical product and process attributes. Here, aspects of production and purification processes of pneumococcal surface proteins are reviewed, the most common bioreactor cultivation strategies are discussed, and important features of the purification process are explored to bring new insights about the correlation between protein structure and chromatography. The process development oriented to an industrial scale is an essential step for the success of novel protein-based pneumococcal vaccines and can preclude problems that could be hardly identified at flask scale production. Moreover, the early bioprocess development should favor a smooth scale-up and transfer of the process to GMP facilities for future production of new pneumococcal vaccines.

Key points

• Early bioprocess development is crucial to advancing pneumococcal protein vaccines.

• Bioreactor cultivation can help to identify possible process bottlenecks.

• Structural features of similar proteins can orient purification process development.

Graphical abstract

Introduction

Streptococcus pneumoniae is a Gram-positive bacterium that asymptomatically colonizes the human nasopharynx (Weiser et al. 2018), but it can evade this niche to cause mild diseases, such as otitis media, sinusitis, non-bacteremic pneumonia, among others, or serious life-treating invasive pneumococcal diseases (IPD), such as invasive pneumonia, bacteremia, meningitis, and sepsis (Loughran et al. 2019). It is responsible for high mortality and morbidity worldwide, especially in children younger than 5 years old, elderly, and immunocompromised people, and the leading cause of lower respiratory tract infections (Ikuta et al. 2022). It is also associated with an increasing number of clinical isolates with antimicrobial resistance (Cools et al. 2021).

The current pneumococcal vaccines are based on the protection conferred by capsular polysaccharides: the 23-valent polysaccharide vaccine (PPSV23) composed of 23 purified capsular polysaccharides, and the pneumococcal conjugate vaccines (PCV10, PCV13, PCV15, PCV20) composed of 10 to 20 polysaccharides chemically bound to a carrier protein not related to pneumococcus. Plain polysaccharides do not protect children, because they are T-cell-independent antigens, and the PPSV23 effectiveness against pneumococcal pneumonia is declining in older adults due to epidemiological changes (Chandler et al. 2022; Nakashima and Fukushima 2024). PCVs were developed to induce T-cell-dependent immune response and protect children. PCVs are very effective against the serotypes included in their formulation and mass vaccination promotes herd immunity. However, PCVs contain only a few out of more than 100 known serotypes (Campbell and Hammershaimb 2023), which leads to serotype replacement by non-vaccine serotypes in the vaccinated population, mitigating the benefits of the vaccination over time and introducing the need to constantly improve the vaccine valency (See 2023). Moreover, PCVs present a high cost and variable protection due to the different geographical distribution of serotypes (Lochen et al. 2020).

Novel protein-based vaccines are being proposed to overcome the shortcomings of current vaccines, since proteins are naturally T-cell-dependent antigens, dispensing the expensive conjugation reaction, and these antigens are much more conserved than polysaccharides among clinical isolates, which should avoid the replacement phenomenon. Several protein antigens have been reported to offer serotype-independent protection (Table 1), and some of them went to clinical trials in humans. Nonetheless, it is still unclear which and how many proteins would be necessary to offer full protection to most serotypes. Therefore, alternatives to the subunit vaccines have been proposed as serotype-independent vaccines: whole-cell vaccines and extracellular vesicles (EVs) (Narciso et al. 2024). There are several recent reviews on the characteristics and immune responses offered by promising pneumococcal protein antigens (Nishimoto et al. 2020; Aceil and Avci 2022; Lane et al. 2022; Li et al. 2023). However, only a few works describe the production and purification processes to obtain pneumococcal proteins at the bioreactor scale, an obligatory step to move forward in the development of a protein-based pneumococcal vaccine. This review aims to explore the aspects of bioprocess development to obtain high amounts of pneumococcal proteins and the challenges of the purification process. Early bioprocess development is crucial not only to generate high-quality material for pre-clinical and clinical tests, but also to understand the critical attributes of the product and important aspects to scale up the process and transfer it to GMP facilities.

Table 1.

Surface protein antigens for next-generation pneumococcal vaccines

Antigen	Characteristics	Clinical trial	References
PspA	3 families and 6 clades based on sequence similarity, binds to lactoferrin, protecting the bacterium from lactoferrin-killing; inhibits C3 deposition, reducing phagocytosis; can bind to lactate dehydrogenase, GAPDH, and other host cell proteins, enhancing pneumococcal virulence	Yes	Lane et al. 2022
PspC (CbpA, SpsA, or Hic)	11 variant groups, binds to secretory IgA, inhibits factor H and C3, can bind to lactate dehydrogenase and polymeric immunoglobulin receptor (pIgR)	No	Zhang et al. 2000, Haleem et al. 2019
PcpA (CbpN)	Highly conserved, adhesion and biofilm formation	Yes	Visan et al. 2018
CbpD, CbpE, CbpF, CbpG, CbpM	Choline-binding proteins with sequence variability in their N-termini, cell-wall association, and diverse functions related to cell-wall degradation, adhesion, binding to plasminogen (CbpE) and fibronectin (CbpM)	No	Aceil and Avci 2022
PhtA, PhtB, PhtD, PhtE	Generally conserved pneumococcal histidine triad proteins, inhibit complement, bind zinc and maybe nickel	yes (PhtD,E)	Miller et al. 2018
LytA	Highly conserved N-acetylmuramoyl-l-alanine amidase, inhibits complement, releases Ply	No	Ramos-Sevillano et al. 2015
LytB, LytC	Well-conserved cell wall hydrolases, nasopharynx colonization and biofilm formation	No	Ramos-Sevillano et al. 2011, Bai et al. 2014
PsaA	Highly conserved, binds manganese and zinc, protects against oxidative stress, promotes bacterial adhesion	Yes	Li et al. 2023
PiuA, PiaA	Generally conserved iron transport lipoproteins responsible to iron uptake, related to bacterial survival	No	Kohler et al. 2016, Li et al. 2023
PcsB	Highly conserved, secreted 45-KDa protein, peptidoglycan hydrolase, belongs to the CHAP (for cysteine, histidine-dependent amidohydrolase/peptidase) family of peptidoglycan endopeptidases, involved in cell division and septum formation	Yes	Vollmer et al. 2019, Giefing et al. 2008
StkP	Highly conserved, serine/threonine protein kinase, involved in regulation of cell growth and cell division, virulence, competence, and stress resistance	Yes	Vollmer et al. 2019, Giefing et al. 2008
PsrP	Pneumococcal serine-rich repeat protein present in approx. 50% isolates, binds to keratin 10 on lung cells, causing biofim formation and adherence	No	Middleton et al. 2021
Pneumolysin* (Ply)	Highly conserved, cholesterol-dependent cytolysin, 53 kDa, binds to host cell membrane cholesterol, oligomerizes and forms pores, leading to cell death; binds to MRC-1, which enhances survival in the airways	Yes	Nishimoto et al. 2020; Subramanian et al. 2019

*Ply is not a surface protein, but it is released during infection

Pneumococcal protein-based vaccines

The antigens described in Table 1 are produced in Escherichia coli as recombinant proteins for obvious reasons: the amount that could be recovered directly from S. pneumoniae cells is lower than can be obtained in recombinant processes (Garcia et al. 1987). Nonetheless, some pneumococcal proteins have been purified directly from S. pneumoniae. Pneumolysin (Ply) was purified from pneumococci by precipitation methods, ion exchange, and hydrophobic interaction chromatography until 1987 (Shumway and Klebanoff 1971; Johnson et al. 1982; Paton et al. 1983; Kanclerski and Möllby 1987); after that, the production became entirely recombinant. Neuraminidase is another intracellular protein that was purified from pneumococcus cultures (Lock et al. 1988). Also, native PspA and LytA were obtained from pneumococcus culture supernatant (Briles et al. 1996, Holtje and Tomasz. 1976) and changed to recombinant production of N-terminal fragments, since the choline binding region was shown to be irrelevant for the immune response (Briles et al. 2000; Nabors et al. 2000; García et al. 1987).

In general, the recombinant antigens were produced under T7/lac promoter with a His-tag and purified using immobilized metal affinity chromatography (IMAC) as the sole purification step, using additional steps only when the desired purity was not reached, for example, in studies for structural characterization (Luo et al. 2018). This one-IMAC-step approach is adequate to obtain proteins to test in animal models and may be acceptable for early stages of clinical development; but to meet the requirements for human use, at least two different separation principles might be applied to the purification process (De Luca et al. 2020), and the His-tag should most probably be removed (Prasanna et al. 2019; Köppl et al. 2022).

The alternative to produce and purify several recombinant proteins is to cultivate S. pneumoniae to produce whole-cell vaccines (Morais et al. 2019; David et al. 2022), extracts (Entwisle et al. 2017), or EVs (Codemo et al. 2018; Yerneni et al. 2021). Pneumococcal cultivation process is feasible and has long been applied for capsular polysaccharide production. However, pneumococcus is an aerotolerant anaerobe microorganism unable to perform the complete oxidation of carbon sources, and the main end-product of fermentation is the acid lactic that inhibits the cell growth (Carvalho et al. 2013). As a result, the cell density is generally low, which poses challenges to the process scale-up, especially for EVs production because the amount of these vesicles that can be obtain per liter of culture is very low (Parveen and Subramanian 2022). Another issue is the complex composition of the final products, which makes their characterization difficult in comparison to isolated purified antigens (Parveen and Subramanian 2022). The main advantage of these alternatives is the production of all pneumococcal surface proteins at once in their native conformation, while the production and purification of each individual protein would be more costly and complex, and to achieve the correct conformation could be also challenging.

An animal-origin free medium (Liberman et al. 2008), a GMP batch process (Gonçalves et al. 2014), and an intensified perfusion process (Campos et al. 2020) were developed to produce a pneumococcal inactivated whole cell vaccine, using the non-encapsulated mutated strain RM200 (Lu et al. 2010). The perfusion process and culture medium could be applied to pneumococcal extracts and EVs, because perfusion allows threefold higher cell biomass than batch process (Campos et al. 2020); consequently, the amount of protein in the cell extract would be increased, as well as EVs titers, which could be continuously recovered from the microfiltrate during the perfusion cultivation.

Production of recombinant pneumococcal recombinant proteins

The bioprocess development oriented to industrial scale is an essential step for the success of novel protein-based pneumococcal vaccines. Although Escherichia coli cultivation for recombinant protein production is considered a well-established platform (McElwain et al. 2022), each heterologous gene and protein has particularities that will demand process adjustments. High-cell density culture is normally needed to achieve high protein titers and economic viability (Cardoso et al. 2020). Articles reporting the production of recombinant pneumococcal proteins in bioreactors are very rare in the literature, and all proteins in Table 2 are variants of pneumococcal surface protein A (PspA), except for a fusion protein PspA-PdT and PdT.

Table 2.

Gene and protein information of recombinant pneumococcal proteins produced in bioreactor

Protein	Origin strain	GenBank	Protein ID	References
PspA1	St435/96	AY082387.1	AAL92492.1	Rodrigues et al. 2024
PspA245	St245/00	Na	Na	Barazzone et al. 2011, Horta et al. 2011
PspA3	St259/98	AY082389.1	AAL92494.1	Horta et al. 2011, Horta et al. 2012, Carvalho et al. 2012, Horta et al. 2014, Horta et al. 2015
PspA4Pro	St255/00	EF649969.1	ABR53734.1	Campani et al. 2016, Figueiredo et al. 2017, Cardoso et al. 2020, Cardoso et al. 2022
PspA4	EF5668	U89711.1.1	AAC62252.1	Xi et al. 2018
PdT (3-point mutations) and His-TEV-PdT	NCTC7466 (Ply)	X52474.1(Ply)	CAA36714.1(Ply)	Fusco et al. 2024
hybrid PspA-PdT	St94/01(PspA2), NCTC7466 (Ply)	Na	Na	Zane et al. 2023

PspA is present in all clinical isolates, and the mature native protein (without the signal peptide) is divided into four domains: N-terminal α-helix domain exposed outside the capsule that includes the clade-defining region (CDR), proline-rich domain sometimes interrupted by a non-Pro block, choline-binding domain, and hydrophobic C-terminal domain (Yother and Briles 1992, Yother and White 1994). PspA was classified into three families and six clades according to the variation of the amino acid sequences of the CDR, as follows: PspAs from the same clade have > 90% similarity in the CDR, they are clustered in the same family when the CDR similarity is about 72%, and if the similarity is < 50%, PspAs are separated in different families (Hollingshead et al. 2000). Epidemiological data shows that families 1 (clades 1–2) and 2 (clades 3–5) are the most frequent in clinical isolates, while family 3 (clade 6) is rare (Chang et al. 2021; Hollingshead et al. 2000, 2006). Usually, only the N-terminal α-helix and proline-rich domains are cloned because they are the region accessible to the immune system interaction (Shaper et al. 2004; Park et al. 2021).

In addition to the proteins listed in Table 2, there are works that evaluated the influence of induction conditions and other cultivation parameters on recombinant PsaA (Larentis et al. 2011, Larentis et al. 2012), ClpP (Einsfeldt et al. 2011), and Ply (Marini et al. 2014) in shaken flasks using design of experiments, which can be considered the first step of the bioprocess development, as the conditions defined in flasks could be applied to the bioreactor production (Kanno et al. 2019). Although one could infer that proteins evaluated in clinical trials should have been produced in the bioreactor, this information is not available or difficult to find, as is the case of Intercell AG and Sanofi proteins (Posfay-Barbe et al. 2011; Schmid et al. 2011; Ljutic et al. 2012; Iyer et al. 2012; Andrade et al. 2018).

Several strategies can be used to reach high-cell density of E. coli, and the most common is the carbon-limited fed-batch. The feeding profile can vary, and the exponential flow rate is among the best strategies, because it is effective in preventing the substrate overflow and, at the same time, maintaining the cell growth rate at the desired value (Shiloach and Fass 2005). A pre-defined exponential feeding based on mass balance equations was applied to produce recombinant PspA245 and PspA3 N-terminal fragments (Table 2), using either glucose or glycerol as carbon source, reaching around 60 g/L of dry cell weight (DCW) and 3 g/L of recombinant protein (Barazzone et al. 2011; Carvalho et al. 2012). The automatic start of feeding was implemented using a neural network committee (Horta et al. 2011), and the feeding was optimized using a model-based adaptative control that allowed to achieve 9.2 g_DCW/L.h and > 20 g/L of PspA3 (Horta et al. 2012, 2014; Nicoletti et al. 2013), using a capacitance sensor to online monitoring the biomass formation (Horta et al. 2015).

To address the impacts of pure oxygen supply on process economics, an airlift bioreactor was investigated to produce PspA4Pro using an auto-induction medium free of animal-origin compounds (Campani et al. 2016). This medium showed higher specific production of PspA4Pro (mg/g_DCW) and higher productivity in simple batch bioreactor operation mode than using the chemically defined medium for fed-batch, while fed-batch showed higher PspA4Pro volumetric production (g/L) and higher plasmid stability (Figueiredo et al. 2017); thus, it is an alternative for recombinant pneumococcal protein production in laboratory settings with no facilities to perform fed-batch cultures. A different composition of an auto-induction medium was also employed to produce another PspA4 in bioreactor (Xi et al. 2018), but the cell concentration and PspA4 production were lower than reported by Figueiredo et al. (2017), most probably due to the lower amounts of carbon sources used: 5 g/L glycerol, 2 g/L glucose, and 2 g/L lactose (Xi et al. 2018).

An auto-induction medium was also applied to PdT and His-TEV-PdT production in bioreactors (Fusco et al. 2024). This work compared 3 gene versions to obtain a genetically detoxified pneumolysin (PdT) with 3-point mutations (C42G, W433F, and D385N), which presents only 0.0001% of the wild-type toxin cytotoxicity (Berry et al. 1995). Two gene versions have the same protein product, but one of them was codon-optimized for E. coli synthesis, while the third version has the coding sequences for an N-terminal His-tag and a TEV protease cleavage site (His-TEV-PdT). The paper showed the influence of the codon initiation region of mRNA on the synthesis of recombinant proteins and the correlation between 5′ mRNA opening energy and protein titers, revealing that mRNA analyses can help reach high-recombinant protein titers in a shorter time than using only traditional bioprocess optimization strategies (Fusco et al. 2024). Therefore, this work illustrates how an early up-scaling to the bioreactor can preclude further issues for industrial production.

Another example of the importance of the early bioprocess up-scaling can be found in the process development to produce a fusion protein composed of PspA and PdT (Zane et al. 2023). Although the fusion of both proteins in tandem was successfully obtained in flasks, purified by IMAC, and used to immunize mice, showing protection against the lethal challenge (Goulart et al. 2013), this chimeric protein was shown to be unstable when the process was scaled up to a 10-L bioreactor, and new gene constructs were prepared to include molecular linkers between the antigens (Zane el al. 2023). The rigid linker that forms an alpha-helix increased the protein stability from 2 weeks to a year at − 20 °C, while the fusion protein remained stable up to a year at 4 °C with the flexible linker (Zane et al. 2023).

The preference for production of recombinant pneumococcal proteins in E. coli is due to its well-known genome, rapid cell growth, and high yields (McElwain et al. 2022). However, one of the main disadvantages of this host system is the presence of lipopolysaccharide (LPS) in the outer membrane of E. coli. LPS can elicit an endotoxin response in humans even at minimal concentrations (Taguchi et al. 2015), and its hydrophobic lipid component, lipid A, is recognized as responsible for its toxic effects (Schneier et al. 2020). Stringent regulatory standards are in place globally to control the LPS content in biopharmaceutical products, and diverse strategies can be employed to reduce LPS levels during downstream processing, but there is no universal protocol for LPS removal from cell extracts and the methods must be tailored to the specific characteristics of each case (Schneier et al. 2020). To address this issue, an endotoxin-free E. coli strain named ClearColi BL21(DE3), which produces a modified LPS called lipid IV_A that does not cause an endotoxin response in humans (Watkins et al. 2017), was employed to obtain PspA4Pro (Cardoso et al. 2022). This work was the first reporting the cultivation of ClearColi in a bioreactor and showed that ClearColi presented similar PspA4Pro yields, reaching 146 mg PspA4Pro/g_DCW, with lower productivities than conventional E. coli, due to the low specific growth rate of ClearColi (Cardoso et al. 2022); thus, this strain is a potential host for industrial production of pneumococcal recombinant proteins.

Purification of recombinant pneumococcal surface proteins

Detailed information about industrial-oriented purification processes of recombinant pneumococcal proteins is also hardly found in the literature. As mentioned above, most works reported standard procedures used for His-tagged recombinant proteins at a small scale, and they are not being discussed here. Fusion tag technology was largely applied for the production of recombinant pneumococcal proteins, since they can facilitate purification, and some tags can increase protein solubility, but these tags present also disadvantages, for example, they can interfere with the target protein activity, and it could be necessary to remove them, increasing the complexity of the purification process (Köppl et al. 2022). Among the few exceptions that do not employ tags, there are works reporting that recombinant PhtD, PhtE, PcpA, and LytB were expressed in Escherichia coli as soluble proteins and purified with combinations of ion-exchange chromatography (Posfay-Barbe et al. 2011), reaching purity values higher than 95% (Ljutic et al. 2012), without any additional information about columns and purification conditions. Also, untagged PspA4Pro (Figueiredo et al. 2017), PspA4 (Xi et al. 2018), and PspA1 (Rodrigues et al. 2024) were purified using ion exchange (Table 3 and Fig. 1).

Table 3.

Characteristics of PspA computed using ProtParam (https://web.expasy.org/protparam/)

	His-tag	Amino acids	Molecular mass (Da)	Absorbance 280 nm (0.1%)	Isoelectric point	Acid/basic amino acids	Aliphatic index	GRAVY
PspA1 (St435)	N	325	36,508.54	0.367	4.79	87/61	59.69	− 1.317
PspA1 (St245)	Y	378	42,902.85	0.569	5.31	90/74	59.07	− 1.394
PspA2 (St94)	Y	380	42,747.87	0.571	5.43	84/71	62.66	− 1.311
PspA3 (St259)	Y	355	39,368.76	0.227	4.98	86/62	69.77	− 1.164
PspA4Pro (St255)	N	382	42,719.54	0.314	4.82	98/72	70.50	− 1.186
PspA4 (EF5668)	N	419	46,877.49	0.318	4.57	110/68	65.44	− 1.286
PspA5 (St122)	Y	402	45,192.92	0.264	4.84	100/70	71.52	− 1.219

Fig. 1.

Multiple sequence alignment of PspA variants using MUSCLE (Madeira et al. 2024) at https://www.ebi.ac.uk/jdispatcher/msa/muscle?stype=protein. The image was generated with Jalview software (https://www.jalview.org), and the general characteristics of PspA variants are displayed in Table 3. Genbank accession number: PspA1 (AY082387.1), PspA2 (OP871266.1, amino acids 1–302), PspA3 (AY082389.1), PspA4Pro (EF649969.1), PspA4 (U89711.1.1), and PspA5 (EF649970.1). The PspA1 (St245) is described in Goulart et al. (2013), and the gene sequence was not deposited

Besides His-tag, other tags were occasionally used. The glutathione-S-transferase tag (GST-tag) was employed to purify Ply and the maltose-binding protein (MBP-tag) to purify the domain 4 of Ply for structural characterization. GST-Ply was separated in a glutathione-Sepharose affinity column, and the GST fusion was cleaved by a protease (Lee et al. 2018), while MBP fusion with Ply domain 4 was purified by affinity chromatography on an amylose-Sepharose column and the domain 4 released by TEV protease digestion (Marshall et al. 2015).

It was shown that the choline-binding domain of surface proteins can adsorb into DEAE and Q ligands of anion exchange resins, which then operate as affinity chromatography and the adsorbed proteins are eluted using a choline solution (Sanchez-Puelles et al. 1992, Garcia et al. 1999, Vollmer and Tomasz 2001, Paterson et al. 2006). In addition, new purification tags were designed based on the choline-binding domain (Caubín et al. 2001; Stamsås et al. 2013). Although this strategy was important to isolate and characterize choline-binding pneumococcal proteins, it seems to be abandoned, probably because the immunologically relevant domains are exposed on the cell surface and the choline-binding domain is buried in the cell wall, as it serves to attach the proteins to phosphorylcholine residues of teichoic and lipoteichoic acids of the cell wall (Aceil and Avci 2022). Therefore, recombinant pneumococcal proteins are normally cloned without the choline-binding domain for application in vaccines, as is the case of PspA.

PspA fragments from clades 1 and 3, without the choline-binding domain, were produced with N-terminal His-tag (Table 3 and Fig. 1), and during the development of their downstream processing, the best purification results were observed when the anion exchange chromatography (AEX) was performed first, followed by immobilized metal affinity chromatography (IMAC), instead of the most usual procedure for His-tagged proteins: IMAC followed by AEX (Barazzone et al. 2011; Carvalho et al. 2012). The chromatography sequence AEX followed by IMAC showed higher yield and purity of PspA and a longer lifetime of IMAC than the opposite sequence (Barazzone et al. 2011; Carvalho et al. 2012). In addition, these results encouraged the cloning of untagged PspA4Pro and PspA1 (Figueiredo et al. 2017; Rodrigues et al. 2024), because critical process parameters for purification using AEX were already known and only a few adjustments were necessary. PspA3 was loaded into Q-Sepharose FF at pH 6.0 (Carvalho et al. 2012), while the loading pH was increased to 6.5 for PspA245 (Barazzone et al. 2011), PspA1 (Rodrigues et al. 2024), and PspA4Pro (Figueiredo et al. 2017), given their isoelectric points are higher than that of PspA3 (Table 3). For the elution, 300 mM NaCl was used for PspA245 (Barazzone et al. 2011), PspA3 (Carvalho et al. 2012), and PspA4Pro (Figueiredo et al. 2017); 250 mM NaCl yielded better results for PspA1, which was attributed to differences in the amino acid residues exposed on the surface of PspA1 for interaction with the resins (Rodrigues et al. 2024), although the amino acid composition of both proteins is quite similar (Fig. 2).

Fig. 2.

Amino acid composition of PspA variants listed in Table 3, showing the number (a) and the percentage (b) of residues. The sequences were analyzed using ProtParam (https://web.expasy.org/protparam/). Residues are classified as follows: acidic (D, E), basic (H, K, R), polar (N, Q, S, T), non-polar (A, G, I, L, M, P, V), and aromatic (F, W, Y)

PspA3 was adsorbed into a cation exchange chromatography (CEX) at pH 6.0, the same pH value for loading into AEX, which was unexpected given the theoretical isoelectric point (pI) of this protein (Table 3). This unexpected behavior was attributed to two distinct charge regions present in the PspA coiled-coil structure (Carvalho et al. 2012). The PspA structure has a negatively charged portion that is exposed to the bacterial surface and a positively charged region directed to the cell wall (Jedrzejas et al. 2001). A similar phenomenon was observed during the purification of the pneumococcal penicillin-binding protein PBP2a, which bound to a CEX at pH 8.0 despite its overall predicted pI is 6.0, with a net negative charge of − 8 at pH 8.0 (Helassa et al. 2012). The adsorption of PBP2a on CEX was also attributed to distinct regions with net positive and net negative charges in opposite poles of the molecule, corresponding to cytoplasmic and periplasmic domains of PBP2a (Helassa et al. 2012).

PspA4Pro and PspA1 strongly adsorbed into CEX at pH 4.0, below their predicted pI (Table 3), but at ionic strengths higher than normally applied for most CEX processes (Figueiredo et al. 2017; Rodrigues et al. 2024). Thus, this atypically strong adsorption could also be related to the regions with opposite charges present in the PspA structure. The elution of PspA4Pro with high purity from CEX was achieved using 800 mM NaCl (Figueiredo et al. 2017), while PspA1 was eluted at the same conditions with low purity, because host cell proteins (HCP) eluted together (Rodrigues et al. 2024). Hence, a negative CEX was performed at pH 6.0 and PspA1 was recovered in the flowthrough, leaving HCP adsorbed to the resin (Rodrigues et al. 2024). Although not obtained without a tag yet, it will be interesting to check at which pH values PspA2 and PspA5 would adsorb into CEX considering their theoretical pI (Table 3).

The purification process of another version of untagged PspA4 (strain EF5668) included an adsorption step of PspA4 directly from the cell extract into hydroxylapatite (Xi et al. 2018). The adsorption into hydroxylapatite was also successfully applied to purify a PsaA-PspA fusion (Guo et al. 2021). Compared to the HCP and LPS removal by precipitation with the cationic detergent cetyltrimethylammonium bromide (CTAB), which was applied to PspA1, PspA4Pro, and PspA-PdT purification processes (Figueiredo et al. 2017; Cardoso et al. 2022; Zane et al. 2023; Rodrigues et al. 2024), hydroxylapatite has the advantage of avoiding the use of another detergent besides Triton X-100 used for cell lysis (Xi et al. 2018). Nonetheless, it is important to emphasize that CTAB does not adsorb to AEX or IMAC resins; therefore, it is eliminated in the first chromatography step (Figueiredo et al. 2017).

After hydroxylapatite elution, two chromatography sequences were evaluated for PspA4 (strain EF5668) purification: AEX into the Capto-DEAE column at pH 6.5 followed by CEX into SP-Sepharose HP column at pH 5.0, and the sequence CEX followed by AEX, which showed the best results, reaching 96% of PspA4 purity (Xi et al. 2018). The comparison of the two PspA4 variants revealed that the one from strain EF5668 is more acidic and has higher molecular mass (Table 3), while the PspA4Pro from strain St255 has a higher content of hydrophobic residues (Fig. 2). Similarly, the comparison between PspA4Pro and PspA1 from St435 showed that despite the conserved coiled-coil structure and proline-rich region (Fig. 3), small differences in amino acid composition, especially in the amino acids buried or exposed, determined adjustments that culminate in distinct purification processes (Rodrigues et al. 2024).

Fig. 3.

PspA structure prediction using Alphafold 3 (Abramson et al. 2024) available at https://alphafoldserver.com

It is interesting to note that the PspA1 from strain St245, PspA2 from strain St94, and PspA4 from strain EF5668 were all cloned with the non-Pro block (Figs. 1 and 3) that can be present within the proline-rich region of some PspAs (Hollingshead et al. 2000). The presence of this non-Pro block was associated with PspA instability and removed from the hybrid PspA-PdT molecules redesigned to include a molecular linker (Zane et al. 2023). The non-Pro block may also explain the instability of PspA4 at pH < 6.0 (Xi et al. 2018).

Finally, chromatography modeling was applied to optimize PspA4Pro purification in AEX (Benedini et al. 2020, 2023). Steric mass action (SMA) and modified Langmuir (LM) isotherms were used together with the equilibrium dispersive model (EDM) for parameter estimation using real cell extracts containing PspA4Pro produced in different culture conditions and pure PspA4Pro, and the simulation results showed that both, SMA and LM, represented well the chromatograms of complex protein mixtures (Benedini et al. 2020). In addition, the model simulations showed that PspA4Pro was eluted at the beginning of the peak and proposed an alternative elution schedule that provided a 34% increase in the purity (Benedini et al. 2020). The SMA/EDM equations were applied to model different ionic strength elution profiles and improve parameter estimation, and new simulations showed better purification results with 120 and 250 mM NaCl for elution than the original with 150 and 300 mM NaCl (Benedini et al. 2023).

Future of recombinant surface proteins for new pneumococcal vaccines

There is an increasing need to develop the bioprocesses for production and purification of many other pneumococcal surface proteins besides PspA and PdT, as illustrated in Table 1. Among them, PspC presents the highest challenge because 11 variant groups of PspC have been described so far (Park et al. 2021). Other antigens that present a certain degree of sequence variation among clinical isolates are PiuA, PiaA, PhtA, PhtB, PhtD, and PhtE, and the selection of immune cross-reactive variants may be needed. Considering that the high number of different pneumococcal surface proteins would demand a huge experimental effort, future works could focus on proteins that are not yet in clinical trials (Table 1).

Although most researchers have indicated that a single representative of PspA family 1, along with another from family 2, would provide sufficient coverage for all clinical isolates of Streptococcus pneumoniae—given that these two families account for 80 to 99.7% of isolates, depending on the study location (Chang et al. 2021)—the potential risk of a replacement phenomenon, albeit low, cannot be completely excluded. This concern is particularly relevant in regions with elevated numbers of isolates from PspA family 3. At least in those places, epidemiological surveillance would be necessary to indicate if a representative from family 3 might be also included in future formulations of protein-based vaccines. Alternatively, a consensus PspA sequence or a chimeric molecule containing multiple epitopes of PspA proteins from all clades could be designed using bioinformatic tools (Afshari et al. 2023).

Recently, a novel approach called immune interface interference (I3) was proposed to develop bacterial vaccines targeting surface proteins that interact with host immune mediators (Croucher 2024). However, I3 vaccines should face the diversity of immune interface proteins, such as three families of PspA and 11 distinct variant groups of PspC, because they are constantly under the selection pressure of the host immune system. Once again, bioinformatic tools could help select variants and epitopes to design chimeric proteins (Bahadori et al. 2024). The practical application of such chimeric molecules will require the development of bioreactor cultivation and purification processes to allow their actual properties to be assessed.

From the bioprocess point of view, it would be interesting to develop novel bioinformatic tools to use experimental data to predict the behavior of structurally related molecules during chromatography, placing together chromatography and protein modelling. This would accelerate the process development to purify untagged PspAs from all clades and variants of other pneumococcal surface proteins, such as PspC, and could be eventually applied to any protein.

In conclusion, the bioreactor production to generate enough material to develop the purification process, independently of the protein chosen, is a crucial step to prevent issues that could hardly be identified at flask scale production. This is especially important for multiepitope and hybrid proteins, as shown by Zane et al. (2023). Moreover, the early bioprocess development should lead to a smooth scale-up and transfer to the GMP facility for future vaccine production.

Author contribution

VMG conceived, wrote, and revised the manuscript.

Funding

The author would like to thank the Sao Paulo Research Foundation (FAPESP grant number 2017/24832–6), National Council for Scientific and Technological Development (CNPq grant number 310973/2022–8), and Fundação Butantan for financial support.

Data availability

The author declares that all data are available within the paper or within referenced sources. Should any data files be needed, they are available upon reasonable request.

Declarations

Ethics approval

This article does not contain data from any studies with human participants or animals performed by the author.

Competing interests

The author has patents deposited on the production of a vaccine formulation against pneumococcal infection containing PspA1 and PspA4Pro and on the design of multiepitope pneumococcal vaccines.