Engineering

Introduction

Endolysosomal sequestration is a major rate-limiting step in intracellular delivery of biological or chemical therapeutics. In order to elicit responses, therapeutic agents must access the cytoplasm to interact with their corresponding intracellular targets. Most small molecules or macromolecules do not readily diffuse across the cell membrane but are taken up into cells by endocytosis. An exceedingly minor proportion escapes membrane-bound endocytic vesicles prior to acidic and hydrolytic degradation. Many treatment paradigms are therefore marred by poor bioavailability, impaired pharmacokinetics and off-target toxicity.

The ideal drug delivery system would overcome endosomal entrapment, feature multi-drug compatibility, high bioavailability and stimulus responsiveness, and be cost-effective. To resolve this unmet need, our team designed a modular drug delivery platform to mechanically disrupt endosomal compartments as they acidify. We engineered Reb A and Reb B monomer peptides from polymeric Caedobacter taeniospiralis refractile bodies (R bodies) with bioconjugation handles. We demonstrate that modified R bodies retain their mechanical piston-like function, and conjugate them with fluorescent markers to demonstrate their unusually efficient endocytic uptake and cytotolerance. Herein our team also documents the initial iteration of refractile bodies conjugated with acid-labile cargo, aimed at verifying drug release from ruptured endosomes. Our findings highlight the promise of R bodies as efficient, cheap platforms for orthogonal drug and/or excipient conjugation and delivery.

Our project was divided into five phases: expressing and purifying wildtype R bodies, creating new R body variants engineered to display conjugation tags, verifying our bioconjugation strategies, and testing the in vitro application of our modified variants. For more detail on results and protocols, refer to our respective wiki pages.

Check out our Engineering Philosophy for the team philosophy that enabled our engineering success!

---

1. Expression & Purification

Our team found immediate success with R body expression, but purification proved substantially more challenging. R bodies are not amenable to purification by chromatographic techniques due to their poor solubility in aqueous media and unusual size (Cai, 2023; Polka et al. 2016). Polka et al. (2016) describe a simple but relatively harsh purification method that separates R bodies from cellular components through lysis, enzymatic cleanup, centrifugation and repeated washing. However, our team was unable to successfully replicate this protocol. It took us over 6 months of DBTL optimisation cycles to reach sufficient purity for downstream applications. Our team initially used an mNeonGreen-R body fluorescent fusion protein constructed by Polka et al. (2016) to visualise whether R bodies retained their pH-induced extension after our relatively harsh treatment. Once we gained access to more powerful imaging modalities we learnt that the fluorescent R bodies had been giving us repeated false negatives for extension, and finally used wildtype R bodies to verify successful extension post-purification. Our team quickly took to heart that treating every failure as a chance to refine our approach was the cornerstone of progress!

---

Cycle 1.1

Design and Build

Cycle 1.1 optimised conditions for R body expression. Our team was fortunate to inherit two stocks of DH5α strains containing R body plasmids (Reb constructs) created by Polka et al. 2016, namely Reb1 and its fluorescent version Reb206 (Figure 1).

As DH5α E. coli are a standard strain for plasmid propagation (Trivedi et al., 2014), we relied on standard protocols for transferring each construct into BL21 E. coli for recombinant protein expression. See our Protocols page for more details. For protein expression, we faced the following challenge: as complete R body dissociation has not yet been achieved no quantitative methods exist for reliably measuring R body concentrations or yields. As we were limited in imaging modalities, we could only approximately quantify Reb206 expression via fluorescence assays. We therefore decided to optimise expression conditions using Reb206 as a proxy for wildtype expression (the latter was required for conjugation cycles). We decided to first attempt expression at high temperature (37°C) for each construct as it was reported to be optimal for R bodies (Polka et al. 2016; Schrallhammer et al. 2012). We also attempted to optimise Reb206 expression through culturing in LB, enriched TPM medium or culturing a codon-optimised version of Reb206 produced in-house by Team PI Yu Heng Lau in 2018 (YHReb206). We tested enriched media as R body assembly is an energy-intensive process: cells enter a vegetative state once induced (Pond et al., 1989).

Test and Learn

Standard protocols sufficed for propagating our plasmid constructs and transforming them into BL21. We wanted to verify whether any of our experimental conditions (enriched medium or codon optimisation) improved R body yield. This was an important consideration for our future work in Entrepreneurship. We used mNeonGreen fluorescence per cell as a proxy measure of R body expression (Figure 2). Though we applied this data in the Learn part of our DBTL cycle, it should be noted that without data on relative rates of fusion versus wildtype RebB monomer expression, rates of monomer incorporation into assembled R bodies, or rates of successful R body assembly given monomer expression, drawing correlation between monomer expression (fluorescence data) and R body yield is only approximate.

We learnt that codon optimisation did not improve R body yield, but enriched media more than doubled R body yield. However, it also doubled the time taken for the expression culture to reach OD600 = 0.4. As our future work required small yields we continued expressing with LB media. Expression in TPM remained an important consideration for future entrepreneurship work, particularly with respect to scaling-up operations.

Next Cycle

Although we collected evidence that suggested R bodies were successfully expressed, we could not be certain of correct assembly at high temperatures. We could not know how robust R body assembly is at high temperatures, and were concerned that accelerated protein aggregation and misfolding at high temperatures might impede assembly (Rosa et al., 2017). In cycle 1.2, We needed to purify the protein and confirm their contraction-extension function to confirm our success.

---

Cycle 1.2

Design and Build

We now needed to design an unconventional protein purification strategy. R bodies are not amenable to purification by chromatographic techniques such as FPLC due to their poor solubility in aqueous media and unusual size (Cai, 2023; Polka et al. 2016). However, R bodies are reported to be stable under extreme treatments such as 5% mercaptoethanol, 8M urea, boiling, 6M guanidine thiocyanate and even shearing into fragments does not abrogate their pH-sensitive mechanical transitions (Cai, 2023; Pond et al., 1989). This enables, indeed necessitates, the use of harsh purification conditions which all reported purification protocols exploit (Lalucat, 1988). Various cell lysis methods are typically followed by DNase treatment for clean up, and rounds of harsh detergent washes and centrifugation to pellet clean R bodies. We therefore first chose to replicate Polka et al.’s purification strategy with fluorescent Reb206, as we were limited in our imaging modalities. While wildtype Reb1 are not visible with standard brightfield microscopy, individual Reb206 proteins can be easily visualised with a fluorescence filter due to their size. Thus we could visually confirm whether R bodies were purified and whether they survive (i.e. retain pH-sensitive extension) post-purification.

Test and Learn

Replicating Polka et al.’s protocol yielded very poor results. We did not obtain a fine, white sediment as reported in relevant literature but rather a stringy, viscous mass indicative of DNA contamination (Figure 3). We tried to resuspend the pellet to measure A260:A280 to confirm the presence of DNA, but resuspension was incomplete and the opaque, viscous solution did not yield intelligible data. Microscopy at pH 7.5 (when the sample should feature homogeneous masses of small contracted coils) revealed amorphous heterogeneous clumps with both fluorescent and non-fluorescent components, and extensive bacterial contamination. When resuspended in pH 5 buffer to induce R body extension, only a transient inflation of these clumps was observed (Video 1). This implied some mechanical force was being generated (which was further evidence that expression in Cycle 1.1 was successful), but perhaps complete conformational change was impeded by impurity.

As we were replicating Polka et al.’s protocol, our team also observed cloudy white supernatants matching descriptions of suspended R body solutions, and observed fluorescence in supernatant fractions. This suggested R bodies were not precipitating, either due to low centrifugation speeds or DNA contamination impeding settling (or a combination thereof).

Next Cycle

Our team confirmed that fluorescent structures, most likely Reb206 R bodies, were trapped within amorphous clumps. Observation and microscopy indicated the greatest degree of contamination originated from DNA and unlysed bacteria. In the next DBTL cycle we focused on eliminating this contamination, and avoiding R body loss in supernatants.

---

Cycle 1.3

Design and Build

After a literature review our team concluded that Polka et al.’s approach represented the archetypical R body purification method, and no protocol existed that departed from Polka et al.’s main principles. We were on our own. Over the next half-year our team trialed a variety of methods to improve lysis and eliminate DNA contamination. Table 1 compiles the approaches sampled by our team. We collate them here into a single DBTL cycle as our fundamental goal - visualising Reb206 extension by achieving sufficient purity - remained unchanged. The loss of R bodies into supernatant observed in Cycle 1.2 was easily resolved by increasing centrifugation speeds.

Test and Learn

In order to best understand the changes made we recommend referring to our Protocols page for comparison to the original purification method.

Modification	Rationale
Lengthening enzymatic and detergent incubations	Lengthened incubation with lysozyme and DNase to maximise their effect yielded no improvement in purity, although it improved bacterial lysis. Lengthened incubation with 1% SDS solution before washing, inspired by Bedingfield et al. (1984), likely enabled more contaminants to be solubilised into the supernatant and improved purity appreciably. The SDS incubation also facilitated lysis.
Increasing lysozyme and DNase concentrations in the initial incubation step	Increasing concentrations only marginally improved purity. Stringiness and amorphous clumps remained.
Introducing a second lysozyme and DNase incubation	Given that longer incubations and higher concentrations in the first incubation step only marginally improved purity, we hypothesized the enzymes might be inactivated during the first incubation step and replaced them with fresh enzyme in a second incubation step. This improved purity significantly.
Altering buffer composition	Although Polka et al. utilised EDTA in buffers, likely to inactivate intracellular proteases, reports of R body resistance to proteolytic digestion (Pond et al., 1989) and concerns that EDTA might impair DNase and lysozyme activity led us to remove it. No appreciable difference was observed with or without EDTA.
Cell lysis by sonication	Sonication was trialed to improve lysis. Harsh sonication produced foaming and deteriorated purity. Gentle sonication yielded approximately equivalent results to high lysozyme concentrations, so for fear of mechanically damaging R bodies sonication was abandoned.
Increasing ionic strength	High ionic strength has been reported to separate DNA from histone proteins by screening electrostatic attraction (Zhong et al., 2022). We tested increasing salt concentrations but observed on improvement.
Washing with nonionic Triton X-100 detergent instead of SDS	There appears little consensus in the literature as to which detergent is preferable (Polka et al., 2016; Schralhammer et al., 2012). We found that switching from SDS to Triton had the greatest effect on improving purity.

Combining these strategies ultimately yielded a pure sample of Reb206 (see Figure 4 in Cycle 1.4), but still no extension at low pH could be visualized with brightfield microscopy. We had ruled out contamination as the issue behind extension. It was nonetheless a worthwhile effort as speaking with experts such as Professor John Rasko and Dr Nicholas Hunt made us acutely aware of the rigorous purity standards necessary for administration in patients, meaning significant progress had been achieved towards our goal of developing a human therapeutic.

Next Cycle

We sought access to improved imaging modalities so that fluorescence would be unnecessary to detect extension. This would allow us to test our purification protocol on wildtype R bodies (Reb1) as a positive control.

---

Cycle 1.4

Design and Build

In this cycle we gained access to phase contrast and differential interference contrast (DIC) functionality. We could now examine the effects of our purification strategy on Reb1 to determine whether issues with extension were Reb206-specific or a result of our purification protocol.

Test and Learn

We observed Reb1 extension at low pH, which confirmed that Reb206 was non-functional. Surprisingly, even in pH 3 a significant proportion of our R bodies remained contracted. Overall this results enabled us to move onto the next phase of our project: designing and expressing R body variants for bioconjugation.

Next Cycle

Our results opened several avenues of investigation: why were our Reb206 constructs non-functional? Why did some Reb1 R bodies extend at low pH but not others? In future cycles team would likely trial different expression conditions to assess whether temperature-dependent misassembly may have contributed to these observations. We would also investigate whether gentler purification might improve functional R body yield.

---

Genetic Assembly & Conjugation

We wanted to engineer R bodies with a modular cargo-carrying function using orthogonal conjugation strategies. Physical models by Cai (2023) resolve the ribbon-like structure of R bodies composed of RebA and RebB monomers aligned along their vertical axis (that is, horizontal dimensions aligned in the same plane). The N termini and C termini are spatially separated across the outer and inner ribbon surface. This structure allows us to engineer up to four different conjugation handles at once onto the N termini and C termini of RebA and RebB. Orthogonality is crucial as it increases the combinations of cargo or excipient molecules that may be conjugated at once without interference. Polka et al.’s Reb206 construct also demonstrates that multiple versions of a single Reb monomer integrate may integrate into one R body. In future investigations, characterizing monomer assembly mechanisms and ratios will further expand the variety of conjugation handles that can be simultaneously displayed.

Our team considered a range of conjugation strategies. We ultimately selected:

• Sortase A enzymatic ligation due to its reliability and FDA-approval in antibody-drug conjugates (Chen et al. 2016; Taylor et al. 2022). See Cycle 2.1.
• Non-canonical amino acid incorporation by amber codon suppression due to easy testing (Lang & Chin, 2014; Nodling et al., 2019; Waals & Ovaa, 2014). See Cycle 2.2.
• Thiol-maleimide conjugation due to its in vitro stability and precedent in FDA-approved trastuzumab (Fontaine et al., 2014). See Cycle 2.3.
• 2-pyridinecarboxyaldehyde selective N terminus modification as it can be applied post hoc to any engineered R body variant (Macdonald et al., 2015). See Cycle 2.4.

We created new Part for each strategy. Our team also considered using tyrosine-based PTAD click chemistry (Dorta et al., 2020). We designed primers and began the cloning procedure but did not complete testing due to time constraints. Several other strategies were deemed compatible with R bodies but were not tested: native chemical ligation, serine-threonine peptide ligation, typical and atypical split intein conjugation, SpyTag-SpyCatcher, protein farnesyltransferase and N-myristoyltransferase. See our Description page for more details. Future investigation of these strategies would enhance the R body conjugation toolbox.

---

Cycle 2.1

Design and Build

S. aureus sortase A (SrtA) catalyses a highly specific, directional transpeptidase reaction between a C-terminal LPXTG motif and N-terminal oligoglycine (Antos et al., 2009; Figure 6). The closest existing analogue to our drug delivery system, conventional ADCs, suffer from pharmacologic unpredictability due to heterogeneous modification and variable drug-to-antibody ratios (Lucas et al., 2018). SrtA conjugation has thus emerged as a promising option to improve ADC pharmacologic predictability and has been used to generate enhanced counterparts of several ADCs (Beerli et al., 2015). Applied to our delivery system, SrtA site-specific conjugation would generate homogeneously modified drug carriers with defined drug-to-carrier ratios.

As the SrtA reaction is directional, we were limited to conjugating on the C-terminus. We also wanted to conjugate a fluorescent cargo (we chose mNeonGreen due to its improved stability) for easy downstream validation. Research from Chen et al. (2016) had optimised an enhanced SrtA (eSrtA) enzyme that ligated best with an LPETG motif, and this enzyme was commercially available. Thus three constructs needed to be created and expressed: mNeonGreen with an N-terminus oligoglycine, and Reb1 where RebA or RebB had a C-terminus LPETG motif. Hereon we refer to these constructs as GGG-mNeonGreen, RebA-LPETG and RebB-LPETG respectively.

To clone our constructs, overlap extension PCR was used with Reb1 or mNeonGreen templates to insert the necessary sortase motifs and overlaps. The fragments were combined by Gibson assembly. They were transformed into DH5α E. coli for amplification and BL21 E. coli for expression. Meanwhile eSrtA was expressed from a commercial plasmid (Addgene #75144) and purified by the 6xHis/Ni-NTA system (see Results for details).

Test and Learn

We wanted to confirm that C-terminal (or N-terminal) modifications to R body monomers preserve their extension-contraction behaviour. RebA-LPETG and RebB-LPETG both exhibited this activity (Figure 7A). As relatively large peptide motifs did not obstruct R body mechanical transitions we concluded the very C-terminal of R bodies was unlikely to feature specific interactions needed for conformational change.

Our choice of a fluorescent cargo allowed us to easily and qualitatively verify that conjugation was successful (Figure 7B). These results are a scientific curiosity as they represent successful aqueous-phase conjugation on a resuspended insoluble protein, which is also one of the largest natural protein assemblies on which conjugation has been attempted to the extent of our knowledge.

Next Cycle

As no biochemical tests exist to quantify R body yields or concentrations, we could not quantify the efficiency of conjugation. This will be critical to characterise prior to any therapeutic application. In our protocol we also used an enormous excess of GGG-mNeonGreen and eSrtA to force reaction equilibrium towards the product. Future cycles could reduce and optimise reagent concentrations, which would assist in scaling up production for entrepreneurial purposes.

---

Cycle 2.2

Design and Build

Amber codon suppression is a genetic code expansion technique that introduces non-canonical amino acids (ncAA) with powerful reactive handles into proteins. An ncAA added to the cell growth medium is specifically recognized by an orthogonal aminoacyl-tRNA synthetase and attached to an orthogonal amber suppressor tRNA, which is decoded by the ribosome in response to an amber codon (TAG) introduced into the gene of interest, allowing the synthesis of a protein with a site-specifically introduced ncAA (Lang & Chin, 2014; Figure 8).

Amber codon suppression can be challenging to implement and its efficiency varies unpredictably with sequence context, cell type, specific ncAA or translation rates (Meineke et al., 2023; Bartoschek et al., 2021). Therefore we chose to engineer our TAG amber codon only at the N-terminus of RebA or RebB monomers. Thus, if the ncAA was not successfully incorporated we would obtain no R body product, and the mere presence of purified R bodies signifies successful ncAA incorporation. Validating ncAA incorporation at the C-terminus poses a far greater problem that we did not attempt due to time constraints. The same overlap extension PCR technique with Gibson assembly as in Cycle 2.1 was used to create our constructs, hereon named TAG-RebA and TAG-RebB.

Once the ncAA is incorporated into our protein, our team needed to apply an orthogonal chemical conjugation strategy to a fluorescent marker. We chanced upon a paper by Chin et al. (2021) which described incorporation of p-azido-L-phenylalanine (AzF) in E. coli. This ncAA is compatible with copper-catalysed azide-alkyne cycloaddition (CuAAC), a widespread click chemistry reaction that occurs under biologically-relevant conditions, produces a stable product and is orthogonal. The process was reported to be efficient in several contexts and the reagents were inexpensive.

We therefore cotransformed BL21 E. coli with TAG-RebA or TAG-RebB and pEVOL AzF (the plasmid encoding the AzF-tRNA synthetase/tRNA^CUA pair) and cultured in medium with AzF. Purified R bodies were conjugated with sulfo-Cy5 alkyne dye.

Test and Learn

R body purification yielded appreciable pellets which implied amber codon suppression had been successful. However, subsequent CuAAC conjugation with dye was unsuccessful. As we were confident in ncAA incorporation, we troubleshooted the CuAAC reaction by adding a water-soluble ligand, THPTA, which coordinates Cu(I) and accelerates the reaction by preventing Cu(I) oxidation into Cu(II). This afforded excellent results (Figure 9). Demonstrating that R bodies are amenable to ncAA incorporation adds powerful site-specific optionality to our conjugation chassis. We report observing TAG-RebA and TAG-RebB extension under powerful brightfield microscopy but due to time constraints are unable to present DIC images in this wiki.

Next Cycle

Once again, as no biochemical tests exist to quantify R body yields or concentrations we could not quantify the efficiency of conjugation. This will be an important consideration in future applications. Attempting AzF incorporation at the C-terminus of RebA or RebB is difficult to validate, but if R bodies can be at least partially dissociated into monomer peptides, MS/MS could confirm the presence of AzF at the correct position. However, now that we have confirmed that amber codon suppression is successful at the N-terminus, we could verify AzF incorporation at the C-terminus by simply reacting a fluorophore like sulfo-Cy5 alkyne to verify its presence. Lastly, experiments with different ncAAs would help with expanding the R body conjugation repertoire.

---

Cycle 2.3

Design and Build

Modification of native proteins by 2-pyridinecarboxaldehyde (2-PCA) derivatives affords excellent site specificity, mild reaction conditions and no inherent requirement for genetic engineering (MacDonald et al., 2015). This imine condensation reaction occurs specifically at peptide N-termini, yields a stable junction (imidazolidinine) and shows no cross-reactivity with lysine ɛ-amino groups (Figure 10). Moreover, the reaction is largely independent of the N-terminus residue identity, meaning 2-PCA conjugation should be compatible with our N-terminus engineered constructs. In this cycle we therefore aimed to simply replicate the 2-PCA reaction on wildtype R bodies using an alkyne derivative, 5-ethynylpicolinadehyde, reported by Bridge et al. (2021) to enable CuAAC-mediated fluorescent tagging. We used sulfo-Cy5 azide dye because its deep blue colour creates a simple visual output to verify conjugation success.

Although this is a two-step process (5-ethynylpicolinadehyde conjugation, then CuAAC tagging) best monitored by tandem MS, R bodies are incompatible with mass spectrometer as they cannot be decomposed into monomer peptides (and the extreme conditions required would likely create undesirable chemical modifications). Therefore our iterative design cycle relied on empirically testing our product and careful observation. Our optimisation and success with CuAAC in Cycle 2.2 gave us confidence that poor results could be attributed to the 5-ethynylpicolinadehyde reaction step, and that is where we would focus our troubleshooting.

Alternatively, we could have performed the CuAAC-mediated 2-PCA reaction with sulfo-Cy5 azide first, identified the product via NMR or MS, then proceeded to conjugation with R bodies. However we opted against this approach as not only would the sulfo-Cy5 azide group reduce the water solubility of 5-ethynylpicolinaldehyde, but we were also concerned that the resulting product from such a direction would yield a 2-PCA derivative with decreased reactivity, owing to the different electronic properties of the CuAAC triazole product (as opposed to the alkyne) which could be undesirably less electron-withdrawing upon the carbonyl carbon. Future cycles would determine whether these concerns hold ground and help understand the effect of reaction order on conjugation efficiency.

Test and Learn

As always, our testing phase aimed to verify extension post-conjugation and observe color change associated conjugation. Our first conjugation was unsuccessful, but after reviewing optimisations reported by Bridge et al. (2021) we concluded that the reaction was temperature sensitive and retried at higher temperatures. This time both our negative control and experimental pellets were blue (Figure 11A). We hypothesized that the sulfo-Cy5 azide dye might associate strongly and non-specifically with R bodies (e.g. due to π-π stacking) and washed thoroughly with 20% ethanol to better solubilise the dye. This yielded confirmation that we had conjugated wildtype R bodies with 2-PCA chemistry (Figure 11B). Once again we report observing 2-PCA conjugated R bodies to extend under powerful brightfield microscopy but due to time constraints are unable to present DIC images in this wiki.

Next Cycle

To better characterise this strategy changing the order of reactions would allow us to assess intermediate products by NMR or MS. As always, once a method for biochemical analysis of R body concentrations/yields is developed, characterising conjugation yields will be necessary. The purpose of the 2-PCA approach was to enable compatibility with all other N-terminal modifications (potentially enabling two orthogonal strategies to be applied at one site), so evaluating compatibility with, for instance, TAG-RebA or TAG-RebB, would be valuable in future cycles.

---

Cycle 2.4

Design and Build

Maleimide–thiol conjugation is one of the most commonly used methods for bioconjugation, and offers a very fast reaction to quantitatively conjugate proteins under mild conditions (Merten et al. 2019). The reaction employs a thiolate from a surface-exposed cysteine residue to undergo addition to the double bond of a maleimide, forming a thiosuccinimide or succinimidyl thioether junction (Figure 12A). Current FDA-approved ADCs Brentuximab and Trastuzumab employ a thiol-maleimide adduct for its site-specificity and junction stability (Fontaine et al., 2014). Reb A and Reb B monomers contain no cysteine residues: this allows our team to define conjugation sites by engineering cysteine residues onto the N-terminus or C-terminus of Reb A or Reb B (hereon named N RebA, C RebA, N RebB and C RebB). Since surface-exposure of the thiol group is essential, we once again appended the cysteine residue to a flexible GGGGS linker. Our payload was fluorescent sulfo-Cy5 maleimide dye.

Test and Learn

Despite being the most established approach, thiol-maleimide was the most difficult conjugation strategy for our team to perfect. Due to difficulties with transformations, work with N RebA and C RebA was suspended. The thiol-maleimide reaction was successful on N RebB but not C RebB (Figure 12B). Presently, our best theory for why this may be suggests that the density of exposed cysteine residues in close proximity favours disulfide bridge formation, though the success of N-terminus conjugation and presence of TCEP in reaction buffers renders this explanation unlikely. Moreover, while C RebB was observed to readily extend in acid, N RebB did not (Figure 12C and 12D), although an unusual pellet appearance leads us to believe this may be a result of a purification error. This result does demonstrate that there are differences in chemical reactivity or accessibility to different parts of the R body monomer. Prior to us demonstrating this difference in chemical functionality, only Cai (2023) had provided some deuterium exchange data which vaguely suggested there may be accessibility differences between monomer termini. Our finding recapitulates the importance of our site-specific conjugation design. If the site of conjugation has significant influence on chemical functionality, one could envision opportunities such as modulating accessibility to the chemical environment of the endosome and altering cargo cleavage kinetics through rational choice of conjugation site.

Next Cycle

The cause of Cys-N RebB's failure to extend needs to be determined. The first priority would be to replicate purification alongside a milder protocol and determine whether a purification error or harsh conditions are responsible for extension failure. If not, we would assume that the N-terminus linker impedes structural transition, which could be validated by testing whether other N-terminus constructs with identical linkers, like TAG-RebA or TAG-RebB, can extend.

---

Cleavable Linkers & In Vitro Application

Incredibly, we had been able to troubleshoot and successfully apply every conjugation strategy to an unusually large, insoluble protein. Our team now wanted to perform in vitro testing to 1) characterise the tendency of cells to endocytose R bodies, and 2) characterise the efficiency of endosomal rupture and drug release by R bodies.

---

Cycle 3.1

Design and Build

To characterise endocytotic efficiency we chose to use RebA-LPETGGG-mNeonGreen and RebB-LPETGGG-mNeonGreen constructs. mNeonGreen retains its fluorescence in the acidic endosomal environment (Steiert et al., 2018) and is attached to the R body via a non-cleavable linker. This would allow us to track the number of R bodies endocytosed. For cell culture we used EXPI293 cells in suspension, which are derived from a neoplastic human embryonic kidney (HEK) cell line. We made this choice after our interview with Dr. Kristina Cook, who suggested that delivery of R bodies to leaky tumour vasculature/interstitial fluid via trans-arterial chemoembolisation (TACE) would be an optimal implementation strategy. Dr. Cook advised that it could limit premature drug release into systemic circulation, minimising patient side effects, and would work synergistically with our stable platform by allowing them to stay longer in the vicinity of the tumour before (presumably) eventual breakdown. A suspension EXPI293 culture allows us to most closely mimic the delivery of R bodies to tumour cells in interstitial fluid.

Drug release from the R body delivery platform requires cleavable linkers. The cargo must be released prior to R body extension in the endosomal milieu, since endosomal contents will be rapidly diluted after membrane rupture. We considered two approaches: acid-labile groups and protease-labile sequences (such as for cathepsin B). Acid-labile groups, such as hydrazones or oximes, must exhibit favorable kinetics for most cargo molecules to be released prior to endosome rupture, and might suffer from stability issues in vivo. Protease-labile sequences, however, might suffer from incomplete cleavage if membrane rupture proceeds prior to endolysosomal fusion, and cleavage will likely vary significantly between cell types in vivo.

We ultimately chose to characterise our delivery platform with aldoxorubicin as a proof-of-concept. Aldoxorubicin is a prodrug of doxorubicin with an acid-labile hydrazone group and maleimide-functionalised linker. In vivo it is bound to an exposed cysteine on circulating serum albumin and concentrated in neoplasms via the enhanced permeability and retention (EPR) effect. The acidic microenvironment of the tumour facilitates hydrazone hydrolysis and doxorubicin release into the tumour interstitial fluid (Cranmer, 2019; Mita et al., 2014). We selected aldoxorubicin due clinical precedent (it has undergone Phase I clinical trials) and immediate compatibility with our thiol-maleimide conjugation strategy. Another practical advantage of aldoxorubicin is that it fluoresces at around 600nm. We hoped to demonstrate that internalising aldoxorubicin-conjugated R bodies into the acidic endosomal environment achieves targeted intracellular cleavage, and that the cytotoxic effect is enhanced by endosomal rupture. Unfortunately, as Cycle 3.1 was undertaken in parallel with Cycle 2.4 we learned too late that Cys-N RebB did not extend at low pH, meaning our design did not examine the effect of endosomal rupture. We present our results here to only demonstrate that R bodies effectively enable aldoxorubicin entry into endosomes and presumably its hydrolysis therein. Subsequent tests with R bodies that have their extension function will characterise the contribution of endosomal escape to targeted cytotoxicity.

Test and Learn

In order to assess endocytic efficiency we calculated the average number of RebA-LPETGGG-mNeonGreen bodies per cell at different dilutions and incubation times. The mean number of R bodies per cell scaled with dilution factor and averaged 4 R bodies per cell (when a 2 mL pellet was diluted 1:8 times). However, the mean number of R bodies yields a misleading impression: as seen in Figure 13 the number of R bodies per cell follows a bimodal distribution. At high concentration (1:8 dilution), a majority of cells internalised 5-10 R bodies and the remaining proportion internalised 0-1 R bodies (see Figure 14D). The cause and therapeutic implications of this heterogeneity merit further investigation. Notably, RebA-LPETGGG-mNeonGreen constructs did not exert cytotoxic effects (Figure 14D) and up to 20 R bodies were observed accumulating inside cells in a vesicular pattern without notable deleterious effect (the cells continued doubling every 24h, they appear morphologically healthy, and absence of trypan blue staining indicates the cell membrane is intact). Thus we conclude that our drug delivery system is intrinsically non-cytotoxic, matching reports by Pond et al. (1989), and that its cytotoxic activity can be modulated by its cargo. Further, its average intracellular accumulation is ostensibly dose-dependent, which is highly promising for therapeutic application. We also believe the ability of R bodies to facilitate internalisation of membrane-impermeant molecules is highly compatible with the development of large macromolecules such as cyclic peptides into therapeutic options. Cyclic peptides, such as those developed using the RaPID platform, have picomolar binding affinities to target proteins but are limited by cellular impermeability.

Aldoxorubicin conjugated onto Cys-N RebB (aldoxorubicin-Cys-N RebB) produced similar data. It is important to keep in mind that Cys-N RebB does not extend under pH 5, and we verified that aldoxorubicin-Cys-N RebB too did not extend. After 24 h of cell incubation we observed a vesicular distribution of brightly fluorescent particles which we assumed to be R bodies, and a diffuse cytoplasmic gradient (Figure 14A and 14B). It is likely that the R bodies localised to the endosome, and as the endosome matures and acidifies aldoxorubicin is cleaved off and diffuses through the vesicular membrane into the bulk cytoplasm. The relatively high permeability of cell membranes to aldoxorubicin is well known (Yousefpour et al., 2019). This might be why we observed a relatively lower number of R bodies in living cells and a high density of aldoxorubicin in apoptotic cells (Figure 14A and 14C). Such an interpretation already highlights the efficient ability of R bodies to deliver intracellular cargo. However, we cannot rule out that the source of aldoxorubicin observed within cells is slow cleavage from R bodies present in solution (i.e. diffusion through the plasma membrane from the extracellular environment).

Overall, this experiment yields little concrete data due to suboptimal design. It confirms the ability of R bodies to deliver aldoxorubicin into cells, and suggests that they assist intracellular accumulation. However the endosome-rupturing function has not been examined and no cytotoxicity measurements can be made given the starting concentration of R bodies cannot be quantified.

Incidentally, as we were forced to sterilise the R bodies with 100% ethanol to prevent cell culture contamination, we confirmed that RebA-LPETGGG-mNeonGreen is stable in 100% ethanol and preserves its extension function (see Figure 8). Ethanol typically induces protein aggregation and unfolding by weakening non-local hydrophobic forces and enhancing local polar interactions, resulting in destabilisation of the native hydrophobic core and increased formation of local secondary structures (Singh et al., 2010). Our conjugation handles are engineered at exposed surface microenvironments that are most likely not involved in protein folding (indeed, most of our designs rely on exposed, flexible, terminal glycine-rich linkers averse to forming secondary structures that will expose functional groups like thiols or AzF to the chemical environment). The stability of RebA-LPETGGG-mNeonGreen is very likely a function of intrinsic R body stability, which has been reported elsewhere (Pond et al., 1989). We are confident that future investigation will reveal that all conjugated Reb constructs are stable in 100% ethanol. The ability to sterilise a protein with 100% ethanol without structural damage should enable more efficient purification, sterilisation for therapeutic application and a greater variety of compatible chemical reactions. It is a novel and promising discovery for a therapeutically targeted protein construct.

Next Cycle

This round of characterisation convincingly demonstrates that R bodies are non-cytotoxic and efficiently deliver cargo into endosomes. The next cycle of in vitro testing must characterise the endosomal escape function using a cleavable linker. For instance, engineering a cathepsin B cleavage sequence into RebA-LPETGGG-mNeonGreen would allow us to test for diffuse fluorescence in the bulk cytoplasm. As mNeonGreen does not cross membranes this experiment would serve as a proof of concept that R bodies liberate cargo from endosomal entrapment. We could also try attaching a fluorescent cyclic peptide onto a TAG-RebA/B construct. Cyclic peptides are emerging therapeutic options for a variety of diseases but are not membrane-permeant. They could easily incorporate chemical modifications like a hydrazone junction or fluorophore group, which would enable us to evaluate the benefit our system brings (in temrs of endosomal escape) to real therapeutic options.

---

Conclusion

Our team has almost brought a novel intracellular delivery system to experimental completion. We developed a novel purification method that can be used with any R body construct, leveraging the innate stability of the R body platform. We engineered 9 new basic parts and 8 new composite parts which are modular and orthogonal and can be easily interchanged to customise our delivery platform. We characterised each composite part by validating conjugation on every R body construct, bar one. We have demonstrated that our product is intrinsically non-cytotoxic and effectively endocytosed at dose-dependent rates. Only one step remains: proving our system enables membrane-impermeant molecules to cross into the cytoplasm. Our platform offers a promising solution to the challenge of endosomal entrapment that has plagued the therapeutics field since its inception.

Engineering Philosophy

Below, we have documented the philosophical principles that guided our engineering success. We also intend for this to be a resource for future iGEM teams, as an example of how they can apply engineering philosophy to their own work! Learn more on our Contribution page.

References

Alvarez Dorta, D., Deniaud, D., Mével, M., & Gouin, S. G. (2020). Tyrosine Conjugation Methods for Protein Labelling. Chemistry – A European Journal, 26(63), 14257–14269. https://doi.org/10.1002/chem.202001992

Bartoschek, M. D., Ugur, E., Nguyen, T.-A., Rodschinka, G., Wierer, M., Lang, K., & Bultmann, S. (2021). Identification of permissive amber suppression sites for efficient non-canonical amino acid incorporation in mammalian cells. Nucleic Acids Research, 49(11), e62–e62. https://doi.org/10.1093/nar/gkab132

Bedingfield, G. W., Gibson, I., & Horne, R. W. (1984). A comparative study of the structure of isolated refractile bodies (R-bodies) from paramecium—I. The effects of treatment with EDTA or EGTA. Micron and Microscopica Acta, 15(4), 235–246. https://doi.org/10.1016/0739-6260(84)90037-6

Beerli, R. R., Hell, T., Merkel, A. S., & Grawunder, U. (2015). Sortase Enzyme-Mediated Generation of Site-Specifically Conjugated Antibody Drug Conjugates with High In Vitro and In Vivo Potency. PLOS ONE, 10(7), e0131177. https://doi.org/10.1371/journal.pone.0131177

Bridge, H. N., Frazier, C. L., & Weeks, A. M. (2023). An expanded 2-pyridinecarboxaldehyde (2PCA)-based chemoproteomics toolbox for probing protease specificity. https://doi.org/10.1101/2023.02.12.528234

Chen, L., Cohen, J., Song, X., Zhao, A., Ye, Z., Feulner, C. J., Doonan, P., Somers, W., Lin, L., & Chen, P. R. (2016). Improved variants of SrtA for site-specific conjugation on antibodies and proteins with high efficiency. Scientific Reports, 6(1), 31899. https://doi.org/10.1038/srep31899

Chin, J. W., Santoro, S. W., Martin, A. B., King, D. S., Wang, L., & Schultz, P. G. (2002). Addition of p-Azido-l-phenylalanine to the genetic code of Escherichia coli. Journal of the American Chemical Society, 124(31), 9026–9027. https://doi.org/10.1021/ja027007w

Cranmer, L. D. (2019). Spotlight on aldoxorubicin (INNO-206) and its potential in the treatment of soft tissue sarcomas: Evidence to date. OncoTargets and Therapy, Volume 12, 2047–2062. https://doi.org/10.2147/OTT.S145539

Fontaine, S. D., Reid, R., Robinson, L., Ashley, G. W., & Santi, D. V. (2015). Long-term stabilization of maleimide–thiol conjugates. Bioconjugate Chemistry, 26(1), 145–152. https://doi.org/10.1021/bc5005262

Lalucat, J. (1988). 4 Analysis of refractile (R) bodies. Methods in Microbiology (Vol. 20, pp. 79–90). Elsevier. https://doi.org/10.1016/S0580-9517(08)70048-5

Lang, K., & Chin, J. W. (2014). Cellular incorporation of unnatural amino acids and bioorthogonal labeling of proteins. Chemical Reviews, 114(9), 4764–4806. https://doi.org/10.1021/cr400355w

Lucas, A., Price, L., Schorzman, A., Storrie, M., Piscitelli, J., Razo, J., & Zamboni, W. (2018). Factors affecting the pharmacology of antibody–drug conjugates. Antibodies, 7(1), 10. https://doi.org/10.3390/antib7010010

MacDonald, J. I., Munch, H. K., Moore, T., & Francis, M. B. (2015). One-step site-specific modification of native proteins with 2-pyridinecarboxyaldehydes. Nature Chemical Biology, 11(5), 326–331. https://doi.org/10.1038/nchembio.1792

Meineke, B., Heimgärtner, J., Caridha, R., Block, M. F., Kimler, K. J., Pires, M. F., Landreh, M., & Elsässer, S. J. (2023). Dual stop codon suppression in mammalian cells with genomically integrated genetic code expansion machinery. https://doi.org/10.1101/2023.03.26.534279

Merten, H., Schaefer, J. V., Brandl, F., Zangemeister-Wittke, U., & Plückthun, A. (2019). Facile site-specific multiconjugation strategies in recombinant proteins produced in bacteria. In S. Massa & N. Devoogdt (Eds.), Bioconjugation (Vol. 2033, pp. 253–273). Springer New York. https://doi.org/10.1007/978-1-4939-9654-4_17

Mita, M. M., Natale, R. B., Wolin, E. M., Laabs, B., Dinh, H., Wieland, S., Levitt, D. J., & Mita, A. C. (2015). Pharmacokinetic study of aldoxorubicin in patients with solid tumors. Investigational New Drugs, 33(4), 752–762. https://doi.org/10.1007/s10637-015-0232-1

Nödling, A. R., Spear, L. A., Williams, T. L., Luk, L. Y. P., & Tsai, Y.-H. (2019). Using genetically incorporated unnatural amino acids to control protein functions in mammalian cells. Essays in Biochemistry, 63(2), 237–266. https://doi.org/10.1042/EBC20180042

Polka, J. K., & Silver, P. A. (2016). A Tunable Protein Piston That Breaks Membranes to Release Encapsulated Cargo. ACS Synthetic Biology, 5(4), 303–311. https://doi.org/10.1021/acssynbio.5b00237

Pond, F. R., Gibson, I., Lalucat, J., & Quackenbush, R. L. (1989). R-body-producing bacteria. Microbiological Reviews, 53(1), 25–67. https://doi.org/10.1128/mr.53.1.25-67.1989

Popp, M. W., Antos, J. M., & Ploegh, H. L. (2009). Site‐Specific Protein Labeling via Sortase‐Mediated Transpeptidation. Current Protocols in Protein Science, 56(1). https://doi.org/10.1002/0471140864.ps1503s56

Rosa, M., Roberts, C. J., & Rodrigues, M. A. (2017). Connecting high-temperature and low-temperature protein stability and aggregation. PLOS ONE, 12(5), e0176748. https://doi.org/10.1371/journal.pone.0176748

Schrallhammer, M., Galati, S., Altenbuchner, J., Schweikert, M., Görtz, H.-D., & Petroni, G. (2012). Tracing the role of R-bodies in the killer trait: Absence of toxicity of R-body producing recombinant E. coli on paramecia. European Journal of Protistology,48(4), 290–296. https://doi.org/10.1016/j.ejop.2012.01.008

Singh, S. M., Cabello‐Villegas, J., Hutchings, R. L., & Mallela, K. M. G. (2010). Role of partial protein unfolding in alcohol‐induced protein aggregation. Proteins: Structure, Function, and Bioinformatics, 78(12), 2625–2637. https://doi.org/10.1002/prot.22778

Steiert, F., Petrov, E. P., Schultz, P., Schwille, P., & Weidemann, T. (2018). Photophysical Behavior of mNeonGreen, an Evolutionarily Distant Green Fluorescent Protein. Biophysical Journal, 114(10), 2419–2431. https://doi.org/10.1016/j.bpj.2018.04.013

Trivedi, R. N., Akhtar, P., Meade, J., Bartlow, P., Ataai, M. M., Khan, S. A., & Domach, M. M. (2014). High-Level Production of Plasmid DNA by Escherichia coli DH5α Ω sacB by Introducing inc Mutations. Applied and Environmental Microbiology, 80(23), 7154–7160. https://doi.org/10.1128/AEM.02445-14

Wals, K., & Ovaa, H. (2014). Unnatural amino acid incorporation in E. coli: Current and future applications in the design of therapeutic proteins. Frontiers in Chemistry, 2. https://doi.org/10.3389/fchem.2014.00015

Yousefpour, P., Ahn, L., Tewksbury, J., Saha, S., Costa, S. A., Bellucci, J. J., Li, X., & Chilkoti, A. (2019). Conjugate of Doxorubicin to Albumin‐Binding Peptide Outperforms Aldoxorubicin. Small, 15(12), 1804452. https://doi.org/10.1002/smll.201804452

Zhong, Y., Moghaddas Sani, H., Paudel, B. P., Low, J. K. K., Silva, A. P. G., Mueller, S., Deshpande, C., Panjikar, S., Reid, X. J., Bedward, M. J., Van Oijen, A. M., & Mackay, J. P. (2022). The role of auxiliary domains in modulating CHD4 activity suggests mechanistic commonality between enzyme families. Nature Communications, 13(1), 7524. https://doi.org/10.1038/s41467-022-35002-0