Spider silk types

Spider silk is considered to be one of biomaterials of the future, which is already used to solve modern problems in innovative and sustainable ways (Fang et al. 2025; Pulkkis 2020; Ramezaniaghdam et al. 2022; Spiess et al. 2010). Being stronger than most kind of steels or Kevlar, ultra-light weight, biodegradable and highly elastic at the same time makes dragline silk a highly promising candidate for biotechnological innovations (Gosline et al. 1999; Vollrath und Knight 2001). Furthermore, applications of spider silk span a broad range of industries and branches like biomedicine, aerospace and biosensors to name just a few (Ramezaniaghdam et al. 2022). To date most of the research in this area has focused solely on dragline silk of spiders, which is known for its extreme tensile strength, even though spiders possess multiple other types of silk as well, which they use for specialized tasks as shown in Figure 1. These different silk types possess different properties, potentially useful in different applications.

Pyriform silk, which is also called the attachment cement of spiders, is used for anchoring the web to surfaces and to join different silk types together. This silk type consists of two spider silk proteins (which are also called spidroins), a dry fibre (pyriform silk protein 1 (PySp1) and a wet glue (pyriform silk protein 2 (PySp2); Geurts et al. 2010; Wolff et al. 2015). The dry fibre forms a micro framework, as shown in Figure 2, in which the glue hardens (Greco et al. 2020).

Such an organized matrix could substantially improve construction materials, for example innovative cements based on MICP-driven biomineralization, by supplying additional nucleation sites, creating surfaces for bacterial colonization, and minimizing gaps between particles. Until now, only two complete coding sequences (CDS) of pyriform 1 silk genes (pysp1) have been annotated (Araneus ventricosus, Accession number: MH376748 (Wang et al. 2019) and Argiope argentata Accession number: KY398016.1 (Chaw et al. 2017)). Other submissions in the NCBI GenBank database consist only of partial fragments or unvalidated whole-genome shotgun entries (NCBI GenBank, as of April 15, 2025). Both genes are divided into five distinct domains, as illustrated for the pysp1 from A. ventricosus in Figure 3. Since the pyriform silk gene from A. ventricosus is significantly smaller (11,935 bp) than that of A. argentata (17,280 bp) it was chosen as template for further work in this project.

Bacillus subtilis as ideal expression host

B. subtilis was selected as ideal expression host as inclusion body formation is less frequently observed compared to E. coli; its efficient secretion systems (Zhang et al., 2020); its genetic tractability (Radeck et al., 2013); it is exotoxin and endotoxin-free and generally recognised as safe (GRAS; de Souza et al., 2021); many tools and strategies are available for its genetic and metabolic engineering (Cruz Ramos et al., 2000; Radeck et al., 2013; Newman et al., 2020; Brockmeier, Wendorff & Eggert, 2006); can be grown on inexpensive media (Zhang et al., 2020). Typical approaches for stable heterologous protein production in B. subtilis often include the creation of single-copy transformants. For genomic integration, the amyE locus was selected because it is non-essentiality and ease for selection against successful integrants.

To overcome the challenges of cloning a highly repetitive gene, the design and construction of a simplified synthetic pyriform silk gene was aimed at. First, the pysp1 gene sequence from A. ventricosus was analysed and split into three domains to ease downstream workflows and cloning. All three domains were harmonized towards the codon usage of B. subtilis W168, possibly circumventing tRNA depletion and the loss of translation pauses, some of which may be important for correct protein folding. Consequently, this might have increased the chance of obtaining correctly folded protein and better yields. A modular design of the single domains, including the utilization of MoClo and RFC25 cloning standards and corresponding restriction enzymes, allowed the generation of constructs with increasing complexity.

AlphaFold predictions hypothesis domain related structural independency

To date, no complete resolved structures of pyriform silk proteins have been published. To gain a better understanding of the protein′s structure, we used AlphaFold v3 (Abramson et al. 2024) to generate structure predictions of synthetic constructs with varying numbers of repeat units. The impact of increasing repeat numbers on the protein structure was subsequently evaluated in ChimeraX (Pettersen et al. 2021). In Figure 4, the predictions of synthetic pyriform silk proteins with one, four and eight repeat units are shown.

The construction and cloning of highly repetitive genes present significant challenges. Flexible assembly techniques like Gibson Assembly are impractical, since they require long single stranded DNA overhangs, which would pose a risk for undesired recombination or hybridisation (Torella et al. 2014). Standard cloning methods based on restriction digestion and ligation lack the scalability, flexibility and modularity required to rapidly assemble large synthetic operons or expression units with fine-tuned architectures (Marillonnet & Grützner 2020). Moreover, PCR-based strategies are often error-prone and inefficient when applied to highly repetitive sequences (Hommelsheim et al. 2014).

To overcome these limitations, we developed a novel PCR free cloning strategy termed Pyricloning. This approach combines elements of the type IIP-restriction-based assembly standard RFC25 (Grünberg et al. 2009) for repeat oligomer generation with the type IIS-dependent MoClo (Modular Cloning) assembly system (Weber et al. 2011), providing a flexible and robust framework for DNA assemblies.

The RFC25 BioBrick standard (also known as Freiburg Standard) is a refinement of the original BioBrick assembly method, developed by the 2007 iGEM team from Freiburg (Grünberg et al. 2009). It enables in-frame fusion of protein coding sequences, ideal for the buildup of repeat oligomers. It employs AgeI and NgoMIV sites flanking each part, in this case the repeat elements (Figure 6). Here the AgeI and NgoMIV sites are between the coding region and the SapI sites, which are needed for the MoClo assembly. These create compatible overhangs, preserving the reading frame during ligation and leaving a stable but non-cleavable AgeI/NgoMIV scar. Each scar encodes two amino acids (threonine, glycine), but the prefix and suffix remain intact, allowing iterative extensions without frame shifts or stop codons.

In Pyricloning, RPE vectors are digested in two parallel reactions, one with SapI/NgoMIV and the other with SapI/AgeI. This generates two repeat units with defined overhangs. After enzyme inactivation, a SapI-digested entry vector and T4 ligase are added. The specific overhangs ensure directional ligation: the NgoMIV-flanked fragment integrates at the vector’s left site, while the AgeI-flanked fragment ligates to the right. The resulting AgeI/NgoMIV scar stabilises the fusion, yielding a repeat dimer.

This dimer can be recursively combined with monomers or other dimers to generate higher-order oligomers (trimers, tetramers, etc.). Ultimately, assembled repeat blocks are integrated into complete coding sequences (CDS) through the MoClo hierarchy (RFC1000), enabling seamless progression toward functional pyriform silk expression units.

To be able to express synthetic CDS (level 0), a full expression construct (level 1) consisting of a promoter, a ribosome binding site and a terminator needed to be formed around them, enabling ideally controllable transcription. For the assembly of the different constructs across the levels, the respective parts together with a suitable entry vector were combined in a one-tube MoClo reaction.

The entry vectors for all MoClo reactions, as well as all other MoClo parts, which were not generated by us, were kindly provided by Georg Fritz (University of Western Australia). Next we edited our sequences with respect to the codon usage bias to increase our chances of obtaining spider silk proteins later.

Heterologous protein production can be challenging, particularly for complex proteins such as spider silk. Their repetitive nature leads to a high abundance of specific amino acids. Especially serine (937 amino acids – 23.6%), alanine (709 amino acids – 17.8%) and glutamine (539 – 13.6%) alone make up over 50% of the wild-type protein’s amino acid composition. To minimize resulting risks like tRNA depletion, misfolding or degradation, codon harmonization was employed to all three parts of the synthetic pyriform silk gene, aligning it with the codon usage of B. subtilis W168.

Info Box - Why does codon harmonisation matter?

How codon usage bias affects recombinant protein production

Codon usage bias significantly affects heterologous protein production in microbial systems. Different organisms prefer certain synonymous codons to encode the same amino acid (Gustafsson et al. 2004). When genes with a codon usage profile that differs significantly from the host’s are expressed, translation can be hampered by insufficient levels of the corresponding tRNAs (Angov et al. 2011). As illustrated in Figure 7, the presence of rare codons (left panel) can slow down translational speed. In the case shown in the left panel, only a few correct tRNAs are available to bind the ribosome, while many incorrect tRNAs can compete for the same site. In contrast, a frequently used codon has a much larger tRNA pool, increasing the chances of correctly binding to the ribosome (right panel; Ikemura 1981).

Codon optimisation seeks to adapt rare codons to the host’s preferred synonyms, improving compatibility with its translational machinery (Plotkin und Kudla 2011). It is important to note that increased translation speed due to codon optimisation is not universally beneficial. In some cases, slower translation at specific regions – caused by rare codons – can assist in proper co-translational folding of complex proteins (Angov et al. 2011). This is exemplified in Figure 8. As it takes more time for the tRNA to come to the ribosome in a rare codon situation, the translation speed is slowed down, allowing the already formed peptide chain to fold into a defined structure before other domains are translated (Liu 2020). To account for this effect in a heterologous system, the gene’s codon frequencies can be adapted to the codon usage of the host organism. By replacIng rare codons with synonymous codons that are also rarely used in the host, one can approximate the native translation speed and maintain co-translational folding (Angov et al. 2011).

Summarized, codon optimization can alleviate translational inefficiencies by adapting the gene sequence to reflect the tRNA availability of the host, but it does not fully address downstream problems such as misfolding, aggregation, or degradation (Gustafsson et al. 2004). For complex proteins, like spidroins with modular repeat structures, intracellular folding may fail or tRNA pools of highly used codons deplete, resulting in inclusion body formation and a loss of functional protein (Ramezaniaghdam et al. 2022).

Addressing Codon usage bias with Codon Harmonisation

As already mentioned rare codons can induce ribosomal pauses which facilitate co-translational folding of complex domains (Liu 2020). However, heterologous gene expression often introduces mismatches between donor and host codon usage, as seen in Bacillus subtilis, leading to low yields or misfolded proteins (Gustafsson et al. 2004). Codon harmonization represents a more nuanced strategy then codon optimization. Instead of uniformly replacIng rare codons with frequent ones in the host, harmonization aims to preserve the donor’s codon distribution relative to host usage (Angov et al. 2011). Rare codons in the donor are substituted with similarly rare codons in the host, while frequent donor codons are aligned with frequent host codons (Parvathy et al. 2022). This approach retains local pauses critical for folding while preventing depletion of tRNA pools (Hanson and Coller 2018; Quax et al. 2015). Despite these advantages, harmonisation is technically demanding. Accurate codon usage profiles from both donor and host are required, which can be difficult without well-annotated genomes. Available bioinformatics tools for harmonisation often require specific input formats and may not function reliably (Willems et al. 2023). Furthermore, unintended biases can be introduced if codon redistribution disrupts subtle regulatory elements within mRNA. In conclusion, codon usage directly impacts translational efficiency, folding dynamics, and host cell fitness (Gustafsson et al. 2004; Quax et al. 2015). While traditional optimisation enhances yields, it ignores the roles of rare codons in folding and translation regulation (Angov et al. 2011). Codon harmonisation, though more complex to implement, offers a refined strategy that preserves native translational behaviour while adapting constructs to the host. This is especially promising for aggregation-prone or structurally complex proteins, such as spider silk, where folding pausing mechanisms are essential (Parvathy et al. 2022). With ongoing improvements in bioinformatics, harmonisation may soon become a practical tool for synthetic biology and industrial biotechnology applications (Willems et al. 2023).

To assess differences in the codon usages (CU) between the donor organism (A. ventricosus) and the host organism (B. subtilis W168), codon usage tables were generated for both species. For B. subtilis W168, the annotated reference genome (ASM904v1) was utilised for the extraction of the codon usage. Although a reference genome exists for A. ventricosus, it is poorly annotated, with most of the entries consisting of hypothetical or predicted genes. This made it unsuitable for generating a reliable codon usage table. Furthermore, as a eukaryote, A. ventricosus contains introns within its genes, which would distort codon usage analysis on genomic sequences.

A screen of the NCBI GenBank database yielded 28 available cDNA entries for A. ventricosus. After sorting out redundant and synthetic entries, 14 cDNA sequences were selected for further analysis. To obtain the coding regions, open reading frames were extracted from these sequences and used to generate the A. ventricosus codon usage table.

These codon usage tables were then compared to identify codons with similar or differing usage patterns between the two organisms. This section illustrates the harmonisation process using proline codons as an example, although the same process was applied to all amino acids. Proline codon usage is shown in bars 1 and 2 of Figure 9 for A. ventricosus and B. subtilis W168, respectively. While the usage of the CCU codon is nearly identical, making the difference negligible, the codons CCC, and especially CCA and CCG, exhibit more substantial differences. These discrepancies are critical to address for ensuring proper protein production and folding.

To gain a better understanding of the relationship between host codon usage and codon frequency (CF) – defined as the relative occurrence of a specific codon for a given amino acid in the wild-type gene – the codon frequency of the wild-type sequence was extracted and compared to codon usage data. Bar 3 of Figure 9 illustrates the proline codon frequency in the wild-type gene, which closely mirrors the codon usage pattern observed in A. ventricosus, albeit with a slightly higher frequency of CCG, CCA, and CCC, and a lower occurrence of CCT. In our project, differences between codon usage and codon frequency are referred to as "distances." These distances for the proline codons - between the wild-type gene and A. ventricosus (bar 5), and between the wild-type gene and B. subtilis W168 (bar 7) - are clearly distinct. As with codon usage, a substantial discrepancy in distances is evident.

To design a well-harmonised synthetic pyriform gene for expression in B. subtilis W168, the goal was to achieve distance values comparable to those shown in bar 5. Thus, the distance between the codon frequency of the harmonised synthetic pyriform gene and the codon usage of B. subtilis W168 was approximated towards the wild-type situation. This might have led to a more proportional distribution of the burden on tRNA pools, while maintaining distances that may be involved in regulating translation speed through translation pauses.

In the case of proline codons, the simplest way to achieve a harmonised sequence with close to native distances, was to switch the CCC codons to CCA, the CCA codons to CCG and the CCG codons to CCC. Additionally, one CCG codon (codon 576) was changed to CCU to better match codon usage-to-frequency distance. These modifications resulted in new proline codon frequencies, shown in bar 4. As a result, the corresponding distances closely resembled those of the wild-type gene, as depicted in bar 6.

In contrast, bar 8 shows the distances of a hypothetical codon optimised synthetic pyriform gene, in which all proline codons were switched to CCG – the most used proline codon of B. subtilis W168. This resulted in even greater distances compared to the wild-type gene in B. subtilis W168 (bar 7), with a total value greater than one (distances were calculated as absolute values).

For amino acids where switching codons was not viable due to the different codon usage of A. ventricosus and B. subtilis W168, and single codons needed to be addressed manually, we wrote a programme to facilitate this step. In brief, the used code calculated the distance of each codon along the sequence and then highlighted it, if a certain distance threshold was surpassed. Thereby, codon hotspots and single codons with high distances could be directly identified. If an effect on translation of these hotspots could be ruled out, for example, if the codons were used far less than the average codon usage of A. ventricosus (which would also yield a high distance), they were distributed across synonymous codons to approximate the native distances, thus lowering the burden on single tRNA pools. Conversely, when codons were used far more frequently than the average, they were manually substituted one by one to best match the wild-type distances and preserve their relative distribution across the sequence.

However, as mentioned earlier, this process was only necessary if the codon usage pattern for an amino acid differed a lot between A. ventricosus and B. subtilis. In cases like CCU no adaptations were made, as the distances were already similar. Finally, the distances of the synthetic and the wild-type pysp1 were approximated across all codons, similar to Figure 9 bar 5 and 6, circumventing problems like tRNA depletion and protein misfolding.

Overall, the harmonisation of the synthetic pyriform silk gene minimised discrepancies in codon usage and codon frequency between the host organism and the synthetic gene construct, while preserving a close-to-native distribution of tRNA demand. This approach should increase the chances of obtaining sufficient expression levels and proper protein folding during heterologous production in B. subtilis.

First, multiple repeat oligomers were generated. For this purpose, a dedicated plasmid called pRPAV (repeat assembly vector) was designed, featuring an mRFP cassette as visual marker for empty vectors, a kanamycin resistance cassette for selection, and recognition sites for the MoClo enzyme SapI with predefined overhangs for repeat unit integration as shown in Figure 10A. As shown in Figure 10B the repeat oligomers were gradually built up from the monomer utilizing the same reaction protocol as described previously. Figure 10C illustrates a control digest (BbsI/SapI) validating the assembled repeat oligomers.

The cut backbone was observed at ~2203 bp across all lanes, including the control (empty pRPAV vector). Distinct bands corresponding to a repeat monomer (655 bp), dimer (1,300 bp), trimer (1,945 bp), tetramer (2,600 bp), and octamer (5,200 bp) appeared at the expected heights as indicated by the green arrows. Thereby, the successful generation of repeat oligomers up to an octamer was validated. Unexpected additional bands in the dimer, trimer, and octamer lanes suggest possible recombination events due to the use of a recA+ strain (NEB Turbo) for cloning. This is supported by the fact that these bands align with the expected sizes of repeat oligomers with lower repeat numbers. In the trimer lane, a faint band corresponds to the expected height of the repeat dimer. Similarly, in the tetramer lane, bands corresponding to the expected heights of the repeat trimer and dimer can be seen. In the octamer lane, multiple faint bands between the correct repeat band (5,200 bp) and the cut backbone might indicate the presence of tetra-, penta-, hexa-, and heptameric forms. Despite these artifacts, the majority of plasmids contained correctly assembled repeat oligomers, enabling the construction of synthetic pyriform silk genes with increasing repeat numbers.

Next, we used these repeats to build different CDS of synthetic spider silk genes. Initially, a basic synthetic spidroin construct (level 0) consisting of the standard NTE, a monomeric RPE and the standard CTE, was assembled to prove the concept and evaluate the protocol. In Figure 11A the level 0 CDS assembly is shown as a schematic. First, SapI cuts out all the parts from their vectors and the lacZα fragment from the entry vector. This allows the parts to be ligated in the entry vector in the specific order of NTE, RPE and CTE. While the SapI recognition sites of the entry vector are lost during the assembly, the BsaI recognition sites are maintained for further assemblies into full expression constructs (level 0). Following the same method, we successfully assembled a level 0 CDS construct with a tetrameric RPE, as depicted in Figure 11. Later, a level 0 CDS with an octameric RPE was assembled as well. Figure 11B shows the control digest (BsaI) of the synthetic pyriform silk gene construct with a tetrameric RPE, exemplifying the validation process for all assembled synthetic CDS (level 0).

In both lanes the cut backbone (2243 bp, black arrow) appeared at the same height around 2,800 bp running higher than expected. Since the backbone band of the control lane ran higher than expected too, this observation was attributed to the gel and a bending of the ladder bands in the upper region of the gel. The band of the lacZα fragment (602 bp, purple arrow) of the control digest appeared at the correct height around 650 bp, confirming that the digest worked as expected. Furthermore, a band in the lane of the CDS construct SD2 (spidroin 2) appeared around 4,400 bp corresponding to the correct CDS fragment (4406 bp, green arrow). Thereby, the correct and successful assembly of the synthetic CDS (level 0) with a NTE, a tetrameric RPE and a CTE was validated. In table 1, all successfully created synthetic CDS (level 0) are listed. These level 0 CDS assemblies were then brought into a full level 1 expression construct.

Tab. 1: List of selected level 0 CDS constructs of synthetic pyriform silk genes generated in this study

Description of CDS 1	Composition 2, 3, 4, 5, 6
SD1	NTE-RPE monomer-CTE
SD2	NTE-RPE tetramer-CTE
SD3	NTE-RPE octamer-CTE
SD4	yoaWSP-mNTE-RPE monomer-mCTE-pep86
SD5	yoaWSP-mNTE-RPE tetramer-mCTE-pep86
SD6	yoaWSP-mNTE-RPE octamer-mCTE-pep86

¹SD = Spidroin construct. ²NTE = N-terminal element. ³RPE = Repeat element. ⁴CTE = C terminal element. ⁵SP_yoaW = secretion signal peptide sequence from yoaW. ⁶pep86 = C terminal fusion-tag required for the NanoLuc assay.

In summary, we successfully established a modular and robust strategy for assembling synthetic spider silk genes by combining repeat oligomer generation with CDS assembly. The pRPAV system enabled the seamless, PCR-free buildup of repetitive elements in an iterative manner, allowing the reliable generation of oligomers up to the octamer level. Despite minor recombination artifacts, the majority of clones contained correctly assembled RPEs, which could be directly integrated into CDS constructs. By coupling these RPEs with the NTE and CTE, we demonstrated the modularity and flexibility of our cloning framework. Overall, this system offers a powerful advantage for the assembly of highly repetitive sequences, overcoming the limitations of conventional restriction ligation and PCR-based cloning methods and providing a scalable route via the RFC1000 framework to synthetic spider silk gene construction. Having established the successful construction of the spidroin CDS modules, the next step was to evaluate regulatory element variants. Therefore, before building complete transcription units with our spidroin CDS, we systematically tested different combinations of promoters, ribosome binding sites, and terminators to determine the optimal expression strength for efficient spidroin production.

Info Box - Why we selected sfgfp as reporter gene?

sfGFP is a well-suited reporter as the protein matures rapidly with a folding time of 6 min in E. coli and is highly stable, thereby producing consistent fluorescence signals (Pédelacq et al., 2006; Shields et al., 2019). For that reason, sfgfp was selected as reporter gene (Shaner et al., 2013).

To be able to utilise sfgfp for the assembly with various promoters and RBSs using the MoClo system, the CDS must first be made compatible. As the MoClo system used is based on utilising restriction type IIS enzymes BsaI and SapI, the sfgfp sequence was first checked for unwanted recognitions sites of the latter enzymes. Fortunately, the CDS of sfgfp did not contain any BsaI or SapI recognition sites. Thereafter, the CDS of sfgfp was amplified using PCR with the primers SG1747 and SG1748. Each of the primers carried overhangs that comprised a SapI recognition sequence alongside with a four basepair sequence appropriate for directional assembly of the CDS in MoClo. Meaning, SG1747 comprised aATG, whereas SG1748 harboured GCTT. The 761 bp PCR product, the sfgfp CDS, was then cloned into a level 0 destination vector pICH41308 via a MoClo level 0 reaction. Colony PCR followed by Sanger sequencing validated the correct insertion of sfgfp into the vector (Figure 12).

Before the screening of promoter strengths and RBS activities, four promoters of varying strength and regulation, together with four synthetic RBSs, were selected for systematic analysis. The promoters included three inducible promoters PxylA, PbceA, and PliaI, and the constitutive PhpaII (Bhavsar, Zhao & Brown, 2001; Ohki et al., 2003; Toymentseva et al., 2012; Zhang et al., 2017). As synthetic RBSs, st4, st7, st11, and wk8 were chosen, while the terminator L3S1P47 was integrated based on a prior screening as the strongest available terminator (Newman et al., 2020). All selected parts originated from the B. subtilis MoClo library of Newman et al., except PhpaII, which was made available as MoClo part in our work (AG Fritz; Newman et al., 2020).

In an additional modification we replaced the lacZ cassette of MoClo destination vectors with an mrfp1-based red/white selection cassette, eliminating the requirement for IPTG and X-Gal in the screening for successful transformants. Using Gibson assembly, new destination vectors were generated: pMMS0 (Level 0), pMMS1 (Level 1), and pMMSM1 (Level M1).

Reporter constructs followed the modular architecture described by Weber et al. (Figure 13A; Weber et al., 2011). For promoter evaluations, transcriptional fusions of PxylA, PbceA, PliaI, or PhpaII with RBS st11, sfgfp, and L3S1P47 were assembled (Figure 13B). Conversely, RBS evaluation constructs employed PliaI with either st4, st7, st11, or wk8 combined with sfgfp and L3S1P47 (Figure 13C). MoClo level 1 reactions generated these transcriptional units in pMMS1, which were validated by colony PCR (cPCR).

To obtain single-copy genomic integrations, validated transcriptional units were subjected to level M cloning for insertion into the non-essential amyE locus (Dierksheide & Li, 2024). Level M constructs correspond to level 2 constructs in the RFC1000 standard and can be seen as equal. The assembly of integration constructs is described more in detail in sections below. Homology-flanked constructs were verified by cPCR (figure 13 E) and integrated into B. subtilis W168 (Figure 13D). Successful integrants were confirmed using a Lugol starch hydrolysis assay (Figure 13F), in which correctly disrupted amyE strains showed no halo formation, unlike the B. subtilis wild type.

The cloning strategy for various reporter constructs is depicted in the figure below (figure 13).

To compare promoter strengths, B. subtilis strains carrying transcriptional fusions of selected promoters PxylA, PbceA, PliaI and PhpaII with st11-sfgfp-L3S1P47 were analysed following induction with the appropriate stimuli. Expression was monitored by end point fluorescence measurements in a plate reader (Figure 14A) and by fluorescence microscopy, both qualitatively and quantitatively via single cell mean fluorescence determination (Figure 14B–C).

Plate reader data showed a strong variability at t0 due to low OD₆₀₀ values, reflected in large error bars. By 30 and 60 min post induction, PliaI and PhpaII consistently exhibited the highest fluorescence values, followed by PbceA, while PxylA remained at wild type levels. Fluorescence microscopy confirmed these findings. Notably, all inducible promoters were inactive prior to induction but produced clear sfgfp signals across all cells after induction, whereas the constitutive PhpaII promoter drove steady fluorescence throughout. Importantly, microscopy revealed an uniform promoter activity, with no detectable cell to cell heterogeneity.

The quantitative mean single cell fluorescence analysis at 60 min substantiated the bulk trends but additionally showed PliaI to be significantly stronger than PhpaII, a difference not resolved in plate reader data. PbceA showed intermediate fluorescence, while PxylA, although weakest, still displayed fluorescence above the wild type control. Statistical analysis confirmed the promoter ranking as PliaI > PhpaII > PbceA > PxylA, each significantly different from the next, with all tested promoters significantly stronger than the negative control.

To assess translational efficiencies, B. subtilis strains carrying translational fusions of four synthetic RBSs (st4, st7, st11, wk8) with the promoter PliaI, reporter gene sfgfp and terminator L3S1P47 (Figure 15A) were analysed following bacitracin induction. Expression was monitored via end point plate reader fluorescence (Figure 15B) and fluorescence microscopy (Figure 15C–F).

As observed in the promoter evaluations, plate reader measurements showed high variability at t0 due to low OD₆₀₀ values, with large error bars persisting at 30 min post induction. At both 30 and 60 min, fluorescence signals were only detected for st7 and st11 constructs, with st7 showing slightly higher bulk values. In contrast, fluorescence microscopy revealed that all RBS variants yielded detectable sfgfp fluorescence at 30 min, significantly above the negative control. Strongest induction was again seen for st7 and st11. Temporal differences were apparent: st11 reached near maximum fluorescence within 30 min, while st7 displayed delayed kinetics, reaching comparable levels only after 60 min. Accordingly, translational efficiencies at 60 min were ranked as st11, st7 > st4 > wk8.

The precise control of gene expression is essential for efficient heterologous protein production (Hershey, Sonenberg & Mathews, 2012), including recombinant spidroins, where fine‐tuned expression improves yields while reducing metabolic burden (Connor, Zha & Koffas, 2024). In metabolic engineering, promoter–RBS libraries have been successfully deployed to balance transcription and translation (Mutalik et al., 2013). To establish a framework for controlled gene expression in B. subtilis, we evaluated promoter strengths and ribosome binding site (RBS) activities using sfgfp as a MoClo-compatible reporter gene (Figure 12 and Figure 13).

Bacitracin-inducible promoter PliaI as promising candidate for fine-tuned protein production in B. subtilis

Four promoters were tested, PxylA, PbceA, PliaI, and PhpaII. A clear ranking emerged, with PliaI > PhpaII > PbceA > PxylA, reflecting distinct transcription initiation strengths. At the pre-induction time point (t0), fluorescence microscopy confirmed the absence of sfgfp expression in inducible constructs, consistent with tight regulation during exponential growth. Following induction, all inducible promoters rapidly reached maximum activity within 30 minutes and maintained stable expression until 60 minutes, showing robust responsiveness to inducers. Notably, fluorescence microscopy revealed no cell-to-cell heterogeneity in expression, supported by uniform intensity distributions and a Shapiro-Wilk test for normality, indicating tight and reliable regulation. This aligns with previous analyses of inducible promoters including PxylA, PbceA, and PliaI (Radeck et al., 2013; Fritz et al., 2015; Ohki et al., 2003; Toymentseva et al., 2012). For instance, PliaI has been reported to display a strong activation within 10 min of induction (Toymentseva et al., 2012).

By contrast, PhpaII produced fluorescence at all time points, consistent with its constitutive activity and previous reports (Guan et al., 2015). A technical variation was observed in bulk fluorescence data at t0, likely resulting from normalisation artifacts at low OD₆₀₀ (≈0.5). Here, even small measurement inaccuracies disproportionately affected calculated fluorescence values. In contrast, microscopy showed no expression, indicating that microscopy provides more reliable baseline measurements.

Both microscopy and plate reader analyses identified PliaI and PhpaII as the strongest promoters, though only single-cell quantification revealed a statistically significant strength difference in favour of PliaI. No difference between PxylA and the negative control was detected in plate reader assays. However, fluorescence microscopy confirmed a weak but significant PxylA activity. These results highlight the reduced sensitivity of bulk measurements compared to single-cell fluorescence quantification, which better captures weak expression events (Miyashiro & Goulian, 2007). Collectively, PliaI emerged as the most promising promoter for recombinant spidroin expression in B. subtilis, combining tight regulation with strong induction potential.

st7 and st11 are promising RBSs for fine-tuned gene expression in B. subtilis

Translational efficiencies were assessed for four synthetic RBSs (st4, st7, st11, wk8). Bulk plate reader measurements again showed high variability at t0 due to low cell densities, but after induction, only st7 and st11 displayed fluorescence significantly above the control. Among them, st7 yielded slightly higher values than st11 in the bulk assay, suggesting a stronger translation.

Single-cell fluorescence analyses, however, provided a more detailed perspective. All four RBS constructs yielded measurable sfgfp expression at 30 min, with st7 and st11 markedly outperforming st4 and wk8. Importantly, st11 reached near-maximal expression within 30 min, while st7 exhibited delayed kinetics, reaching levels comparable to st11 only by 60 min. Statistical analyses confirmed significant differences between st7 and st11, supporting the robustness of these observations. Based on the fluorescence microscopy’s higher sensitivity compared to population-averaged assays, these data more reliably represent true translational differences (Miyashiro & Goulian, 2007).

The slower activation of st7 may derive from differences in Shine–Dalgarno complementarity, which strongly influences the ribosome recruitment efficiency, or from inhibitory mRNA secondary structures near the RBS that reduce the accessibility for translation initiation (Rodnina, 2016; Hershey, Sonenberg & Mathews, 2012; Englund, Liang & Lindberg, 2016). Even subtle differences in secondary structure stability can substantially lower translation rates (Rodnina, 2016).

Taken together, st11 emerges as the most effective RBS for rapid and robust protein production in B. subtilis. However, the delayed but strong activity of st7 presents a valuable alternative when a slower translation initiation rate may be advantageous for correct protein folding which is a common challenge in recombinant production of complex or aggregation-prone proteins (Rosano & Ceccarelli, 2014).

Conclusions from Expression construct screening

This subproject confirms the importance of systematically evaluating regulatory elements at the transcriptional and translational levels. The promoter evaluation identified PliaI as the strongest and most tightly regulated candidate for high-level heterologous expression in B. subtilis. Similarly, comparative testing of RBS variants highlighted st11 and st7 as functional and strong, but with distinct kinetic behaviours affecting timing of protein accumulation. Together, these findings provide a foundation for constructing finely tuned expression systems tailored to recombinant spidroin production in B. subtilis.

The first of these, transcription unit TUC1, was built with the well-characterized promoter PxylA (Heiss et al. 2016; Schmiedel et al. 1997), the strong synthetic ribosome binding site st11 (Salis et al. 2009), the SD1 coding sequence, and the strong synthetic terminator L3S1P47 (Chen et al. 2013). Identical reaction conditions were used as for the Level 0 assemblies, but with BsaI instead of SapI. Subsequently, transcription units containing the SD2 coding sequence were generated with either PliaI (TUC2) or PxylA (TUC3) as promoters, while all other elements remained identical to TUC1. As an example representative of all Level 1 transcription unit assemblies, Figure 17A illustrates the assembly of TUC3, and Figure 17 shows the corresponding control digest (SapI and NgoMIV) used to verify successful construction.

A distinct band pattern correlating to the backbone fragments can be seen in both lanes. Since the lacZα fragment (purple arrow) and one of the backbone fragments (black arrows) have almost the identical size (600 bp and 591 bp respectively), these two bands cannot be distinguished in the gel. However, the band appearing around 600 bp showed a strong signal indicating the presence of both bands. While the band pattern of the TUC3 lane missed such a strong signal around 600 bp, two additional bands around 3300 bp and 1600 bp indicated the presence of the transcription unit fragments (green arrows) at the correct heights. By this the successful generation of the level 1 transcription unit was confirmed.

Several level 1 transcription units with different promoters and synthetic CDS were generated during this project which are listed in table 2. Validation was done the same way as described above. Selected constructs were further used in the next step to assemble level M transcription units.

Tab. 2: List of selected level transcription units with synthetic pyriform silk genes assembled during this project

Description of the transcription unit	Composition
TUC1	PxylA-st11-SD1-L3S1P47
TUC2	PliaI-st11-SD2-L3S1P47
TUC3	PxylA-st11-SD2-L3S1P47
TUC4	PliaI-st11-SD1-L3S1P47
TUC13	PxylA-st11-SD4-L3S1P47
TUC14	PliaI-st11-SD4-L3S1P47
TUC15	PxylA-st11-SD5-L3S1P47
TUC16	PliaI-st11-SD5-L3S1P47
TUC18	PliaI-st11-SD6-L3S1P47

With the successful assembly of level 1 transcription units, the next step was to add homology flanks for the genomic integration (level M) into B. subtilis W168. Consequently, level M transcription units, incorporating homology flanks for the amyE locus and a chloramphenicol resistance cassette were assembled as depicted in Figure 18A. In general, the same reaction conditions and principles as in level 0 were utilised, using different parts and entry vectors. Figure 18B presents the control digest (BsaI and SacI) of TUC2 level M, exemplary for all level M assemblies.

Both the control and the TUC2 lane showed the backbone band (black arrows) at the expected heights around 5,000 and 4,600 bp, respectively. Also, all bands in the TUC2 lane appear at heights corresponding to the level M transcription unit fragments (green arrows). Since the repeat monomer sequence has a SacI recognition site, three of the repeat monomers were cut out by SacI, corresponding to the band (blue arrow) appearing around 650 bp. Moreover, the lacZα fragment also contained a SacI recognition site, which, together with the BsaI recognition site, yielded a 244 bp fragment that was too faint to be visible in the control lane. As the expected and correct band pattern appeared for the TUC2 lane, while those bands could not be detected in the control, the assembly of the level M transcription unit was confirmed. To further verify the assembly process, the plasmid containing the TUC4 level M transcription unit was analysed through whole plasmid sequencing, which served as a representative example for all other level M transcription units. The resulting sequence confirmed that the assembly worked as expected. Several level M transcription units were constructed during this thesis, listed in table 3, which were validated the same way as described above. These transcription units were next used for the genomic integration into B. subtilis W168.

Tab. 3: List of slected level 2/M transcription units habouring pysp1 constructs and signal peptides

Description of the level M transcription unit	Composition 7, 8
TUC1 level M	amyE LF – cmr - TUC1- amyE RF
TUC2 level M	amyE LF - cmr - TUC2- amyE RF
TUC3 level M	amyE LF - cmr - TUC3- amyE RF
TUC4 level M	amyE LF - cmr - TUC4- amyE RF
TUC13 level M	amyE LF - cmr - TUC13- amyE RF
TUC14 level M	amyE LF - cmr - TUC14- amyE RF
TUC15 level M	amyE LF - cmr - TUC15- amyE RF
TUC16 level M	amyE LF - cmr - TUC16- amyE RF
TUC18 level M	amyE LF - cmr - TUC18- amyE RF

⁷amyE LF is the left homology flank targeting the amyE locus of B. subtilis W168. ⁸ amyE RF is the right homology flank targeting the amyE locus of B. subtilis W168.

The last step prior to test heterologous protein production in B. subtilis W168 was the genomic integration. Figure 19A shows a schematic of the integration process leading to the loss of the ability to hydrolyse starch. For the transformation, B. subtilis W168 was cultured in Paris medium, in which the cells become naturally competent (Harwood 1990; Anagnostopoulos & Spizizen 1961). Plasmids containing the level M constructs were added to these cultures, followed by selection with chloramphenicol. Since the level M plasmids lacked a replication origin for B. subtilis W168, only cells that successfully integrated the constructs into their genome were able to survive. Integration of the level M constructs into the amyE locus was facilitated by the previously added homology flanks through two homologous recombination events. Because the native amyE locus enabled B. subtilis W168 to utilise starch as a carbon source, and the construct was targeted to this locus for integration, it was tested whether the transformed strains had lost the ability to metabolise starch. Clones that grew in the presence of chloramphenicol were restreaked on starch plates to assess their ability to hydrolyse starch. After one day of incubation, Lugol’s solution was added to detect the presence of starch and thereby confirm successful disruption of the amyE locus.

As shown in Figure 19B, B. subtilis W168 wild-type colonies, which have an intact amyE locus, can hydrolyse starch, resulting in a clear halo around the colony. This halo forms because starch is degraded and cannot be stained by Lugol’s solution. In contrast, colonies with a properly integrated transcription unit in the amyE locus lose this hydrolytic ability. As the starch remains undegraded, Lugol’s solution stains the surrounding agar uniformly, and no halo is observed. No starch degradation halos were observed for all C1 colonies as well as for C2 of TUC14 and TUC15 and for C3 of TUC13 and TUC16. Thereby, the successful integration of the respective transcription units into the genome of B. subtilis W168 was validated. Due to time constraints, the only other transcription unit assembled to level M, which was integrated into the B. subtilis W168 genome was TUC1. Depending on their design, these clones were used in various assays to evaluate whether B. subtilis W168 can produce and secrete the synthetic pyriform silk proteins.

As we found earlier in the reporter gene assay that among the tested combinations of promoters and RBSs, the liaI promoter in combination with the synthetic RBSs st11 and st7 exhibited particularly strong and consistent activity. Two different RBSs were selected as st11 and st7 showed comparable maximal fluorescence 60 min after induction. However, time dependent differences were observed between these RBSs, which can be particularly useful for tuning the expression of the spidroin genes.

Tradionally, the intracellular production of spidroins in bacteria, insect cells or plants is pursued (Jin et al., 2022). Building on these findings, the intracellular production of spidroin was in focus. Furthermore, as it is known that large, intrinsically disordered proteins tend to aggregate in the cytosol and pose significant challenges for heterologous production, we aimed to begin with the evaluation of the recombinant production of the simplest form of our synthetic pysp1 construct. An exemplary construct design is shown in figure 20.

The resulting recombinant strains of B. subtilis harbouring the gene construct coding for the SPM were tested for the ability to produce the synthetic SPM. To this end, protein production cultures were inoculated to an OD₆₀₀ = 0.05 and grown at 37°C in LB medium. Reaching mid-exponential phase, protein production was induced using 0.002% (w/v) of the peptide-antibiotic bacitracin. After 3 h, cultures were harvested and samples prepared.

Spidroin production has previously been shown to benefit from media supplementation with specific amino acids (Connor, Zha & Koffas, 2024). Furthermore, protein folding can be aided by growth temperature adjustments and thus enhance recombinant protein yields (Bhatwa et al., 2021). For that reason, we tested the protein production additionally in 2xYT medium and lowered the growth temperature after induction to 18°C. Deviating from the conditions when grown at 37°C, cultures grown at 18°C were harvested after 18 h to allow for sufficient protein production time. The protein production was then analysed in SDS-PAGE and Western Blot using whole cell lysate (figure 21).

No significant differences in the band pattern of the whole cell lysates from B. subtilis strains carrying the SPM constructs grown in different conditions compared to the wild-type were obtained in SDS-PAGE (figure 21 A). However, in the growth condition in 2xYT medium at 18°C, a band at ~60 kDa was more pronounced compared to the wild-type and when grown in LB medium. In addition, spidroin production could not be detected in Western Blot (figure 21 B). The positive control, His6-tagged mNG showed an apparent band at approximately 34 kDa, corresponding to the expected molecular weight of ~26.6 kDa. To summarise, spidroin production was not detectable under the tested conditions in SDS-PAGE and Western blot, whereas the positive control mNG-His6 was observable, confirming the validity of the detection approach.

Nevertheless, the comparison of the optical densities at 600 nm after 18 h of growth revealed notable differences between B. subtilis cultures harbouring the SPM constructs when cultivated in LB or 2xYT medium, respectively (table 4). When grown in LB medium, both recombinant strains reached OD₆₀₀ values comparable to the wild-type B. subtilis control. However, in the nutrient-rich 2xYT medium, the OD₆₀₀ values for the SPM containing strains were reduced relative to growth in LB medium. Moreover, there were differences observed between the two constructs. The B. subtilis strain containing the RBS st7 grew to an OD₆₀₀ of 4.12, whereas B. subtilis with RBS st11 grew to an OD₆₀₀ of 5.05. This indicates that the nutrient-richer 2xYT medium led to an impaired growth in the B. subtilis strains with SPM construct with the effect being stronger in the B. subtilis strain carrying the RBS st7.

Tab. 4: Optical density values (600 nm) of recombinant B. subtilis harbouring the spidroin monomer constructs show a decrease in growth after induction in 2xYT medium. B. subtilis grown in LB and 2xYT medium at 18°C after induction. Induced B. subtilis wild-type served as a control. n = 1

Condition	OD600 after 18 h cultivation
	WT	PliaI-st7-SPM-L3S1P47	PliaI-st11-SPM-L3S1P47
LB	6.4	6.14	6.34
2xYT	-	4.12	5.05

Since the spidroin production failed or was undetectable in SDS-PAGE and Western Blot under the tested conditions, we decided to optimise the protein production protocol. Overly strong induction is known to potentially lead to cellular stress and protein degradation, often resulting in low or undetectable yields of heterologous proteins. Hence, carefully tuning the induction to achieve a lower, optimal expression rate can improve recombinant protein yields (Sosa-Carrillo et al., 2023; Rosano & Ceccarelli, 2014). Based on these insights, we aimed to optimise the induction conditions in order to identify parameters that allow for successful spidroin production. 2xYT medium was selected for further cultivation of the B. subtilis cultures as an effect on growth was found in induced cultures compared to when grown in LB medium.

To this end, cultures of the B. subtilis strains carrying the SPM constructs were inoculated to an OD₆₀₀ = 0.05 and grown at 37°C in 2xYT medium. To investigate the effect of induction timing on protein production, protein production was induced at early-exponential phase (OD₆₀₀ = 0.3) and mid-exponential phase (OD₆₀₀ = 0.5) using 0.002% (w/v) bacitracin. Additionally, to assess the impact of a lower inducer concentration, protein production was induced at mid-exponential phase (OD₆₀₀ = 0.5) with a lower bacitracin concentration (0.0003% (w/v) bacitracin). Following induction, cultures were incubated at 18°C for 18 h, and the cells harvested and processed as described in the methods section. The protein production was then analysed in SDS-PAGE and Western Blot using whole cell lysate (figure 22).

Like the previous tested conditions, SDS-PAGE analysis (figure 22 A) showed no notable differences in the band patterns of whole cell lysates from B. subtilis strains harbouring SPM constructs compared to the wildtype. In addition, spidroin production remained undetectable in Western Blot (figure 22 B). In contrast, the positive control, mNG-His6, revealed a band at 34 kDa, confirming the functionality of the experiment.

The observed growth effect in 2xYT medium on B. subtilis strains harbouring the SPM construct was reproducible (table 5). Interestingly, when cultures were induced with a lower inducer concentration (0.0003% (w/v) bacitracin), the effect abolished and the strains reached an OD₆₀₀ comparable to the wildtype control. Additionally, the induction timing, namely induction in early-exponential compared to the mid-exponential phase, using the same inducer concentration did not affect the final optical density measured after 18 h of growth. Due to time constraints, no further attempts were made to optimise the protein production protocol or investigate alternative strategies for spidroin production.

Tab. 5: Effects on bacterial growth of different induction conditions on cultures of B. subtilis harbouring the spidroin monomer constructs. B. subtilis grown in 2xYT medium at 18°C after induction. Induction took place at either early- or mid-exponential phase (OD₆₀₀ = 0.3 or 0.5) with either 0.0003 or 0.002% (w/v) bacitracin.

Induction conditions:		OD600 after 18 h cultivation
		WT	PliaI-st7-SPM-L3S1P47	PliaI-st11-SPM-L3S1P47
At OD600 = 0.5	0.002% (w/v)	6.89	4.16	4.85
	0.0003% (w/v)	-	6.76	6.84
At OD600 = 0.3	0.002% (w/v)	-	4.26	4.93

A multi-copy strategy is useful in heterologous protein production since an increase in the number of gene copies encoding the target protein can boost overall protein yield and therefore enables high-demand applications. The basis for this multi-copy approach was the E. coli – B. subtilis shuttle vector pBSMuL1 constructed by Brockmeier et al. (2006). This is a suitable expression vector for producing heterologous proteins in B. subtilis. To adapt it for the use within the MoClo system and to meet the specific needs of this study, a MoClo-compatible version of the vector was created based on the pBSMuL1 backbone. Additionally, the mrfp1ΔlacO cassette was inserted. Upon successful cloning in the future, the cassette would be replaced by the target insert, resulting in white clones on the plate, while clones carrying the parent vector still show red colourisation. A pBSMuL1-like level 1 vector was constructed to provide the flexibility to change the coding sequence as well as the promoter, ribosomal binding site, secretion signal peptide sequence (sslipA; in case of ssyoaW in combination with a StrepII-tag, SUMO protein and a His6-tag), a C-terminal His₆-tag and terminator. This was achieved by implementing MoClo overhangs starting with GGAG and ending with CGCT. After Gibson assembly and validation via restriction digest and sequencing, this vector has been named pLIMO1 (LI: Lilli, MO: Modular Cloning) for simplification. The functionality of the created vector pLIMO1 was assessed by cloning of two reporter constructs harbouring the promoters PxylA and PhpaII used earlier in the promoter screening, and by assessing the sfgfp expression (figure 24).

B. subtilis W168 cells were transformed with the pLIMO1 PliaI-ssyoaW/sslipA-spider silk monomer (one single repeat) and tetramer (four repeats) constructs, respectively. After validation via an analytical restriction digest with BsaI, bacteria carrying the correct assemblies were used to inoculate LB medium cultures to a starting OD₆₀₀ of 0.05. Figure 25 shows the bacterial growth of the respective cultures with the WT as a reference.

Initially all cultures grew almost identical until induction. Then a clear difference between the cells carrying a spider silk construct to the ones with the WT were visible, resulting in an end-OD₆₀₀= 1-1.74 compared to the WT values of OD₆₀₀= 3.32-3.76 three hours post-induction. From all grown cultures, three samples for SDS-PAGE followed by Western Blot analysis were takes 3 h post-induction. This allowed us to examine five different fractions, i.e. whole cell lysate, secreted proteins, lysate after bead beating, cytoplasmic proteins and cell debris, to ensure a complete analysis of spider silk production. The corresponding protein gels with one representative example of a Western Blot are shown in Figure 26. The protein gels display the plethora of intracellular and secreted proteins. Although the spider silk proteins are labelled with secretion signals, it was expected, if production occurred, that some protein is still within the cell. But no gel showed clear bands for neither the monomer spider silk protein, which would be expected at 85 kDa (plus 15 kDa for ssyoaW-construct in all fractions except the supernatant or plus 8 kDa for sslipA), nor for the tetramer spider silk protein, which would be likely at 135 kDa (plus ssyoaW-construct or sslipA). A GFP-His₆ functioned as a positive control and showed a distinct band just above 26 kDa, corresponding to the expected molecular weight of 26.8 kDa (green arrow). As only the control, the GFP-His protein, was detected in the Western Blot, only one representative image is shown. Since no bands corresponding to the expected molecular weights of the synthetic spider silk constructs were visible in neither the protein gels nor in the Western Blots, the conclusion was that the proteins were either not produced or were present at levels too low to be detected by the methods used.

In summary, the production of synthetic spider silk proteins was tested using two different secretion signal peptides that have previously been tested for msfgfp secretion, namely sslipA and ssyoaW. In addition, the spider silk sequences were assembled as a monomer (one repeat) and as a tetramer (four repeats) to test different lengths of the protein. After several transformation attempts, it was apparent that no colonies could be obtained from any of the assemblies carried out with constitutive promoters such as PxylA and PhpaII. Only after a week of incubation at 30°C small colonies were visible on the LB agar plate. When these were inoculated into liquid LB media, they did not grow. This suggests that a constitutive transcriptional control probably causes such a high level of stress to the cell that it will lyse. Therefore, the inducible PliaI was used for each spidroin assembly. Colonies were visible after one day of incubation under non-inducing conditions. When B. subtilis W168 was grown with the PliaI-sslipA/ssyoaW-monomer/tetramer-constructs in LB medium, all cells grew almost identically pre-induction, suggesting tight control of PliaI, which was the reason for selecting this promoter. Post induction, all cells carrying a spidroin construct showed a strong reduction in growth rate compared to the WT strain, suggesting that the cells are at least trying to produce the synthetic spider silk proteins. However, no distinct (over-) production protein bands were seen in any fraction of the SDS-PAGE, neither for a monomer construct (85 kDa; if not secreted plus 15 kDa for the ssyoaW construct or plus 8 kDa for sslipA), nor for the tetramer spider silk protein, which would be expected to be visible at 135 kDa (plus signal peptide). Instead, a similar pattern was observed for almost all protein gels, except for the cell debris fraction and the secreted protein fraction precipitated by TCA. However, in the spidroin fractions of total cell lysate, cytoplasmic proteins and lysate after bead beating, a distinct band at around 41 kDa was visible, which could be DnaJ (40.7 kDa), a molecular chaperone that ensures protein folding and activates DnaK (65.8 kDa). In the lysate after bead beating, a distinct band at around 40 kDa was seen in all lanes except the non-induced WT sample, which could be LiaS (40.5 kDa), the sensor histidine kinase of the liaIHGFSR operon that senses the inducer bacitracin (Jordan et al., 2006). The other components such as liaI (13.2 kDa), LiaH (25.5 kDa), LiaG (25 kDa), LiaF (26.9 kDa), LiaR (23 kDa) cannot be clearly detected as the protein gel only shows distinct bands up to about 26 kDa.

Therefore, the lack of detection in this study may be due to the DNA copy number and therefore the tRNA pool being too low for the translation of these repetitive long proteins, despite the sequences being codon harmonised. Since there is still a huge knowledge gap in the area of pyriform silk sequences, there were not many data available for such a harmonisation process. Therefore, an additional vector with genes encoding tRNAs for the commonly required amino acids, such as glutamine, glycine and proline, could provide increased levels of these tRNAs and reduce the risk of shortages. Yang et al. (2016) showed that with a higher glycyl tRNA pool, the biosynthesis of glycine-rich silk proteins in E. coli was successful. Another bottleneck could be that the amino acid pool was too low and a targeted supplementation of amino acids could (0.3 g/L of each Gln, Glu, His, Ile, Phe, Pro, Tyr, Lys, Met) increase production, as shown by Connor et al. (2024). This resulted in a 135% increase in the production of a 16.1 kDa silk protein relative to both rich and minimal media, yielding secretory titres of approximately 100 mg/L in flask cultures. Further ideas for optimising the heterologous production of synthetic spider silk proteins are discussed later.

As the multi-copy expression attempt appeared to impose increasing cellular stress and did not yield detectable amounts of protein. We returned and focussed our optimisation steps towards the secretory expression of the synthetic pysp1 in a single-copy approach.

Following integration of the level M pysp1 constructs into the B. subtilis W168 genome, the protein production capabilities of the resulting strains were evaluated. Initially, the goal was to confirm protein production before optimising the setup, harvest timepoints or media composition. Production cultures were grown in LB medium, inoculated to an OD₆₀₀ of 0.05, and gene expression induced by the addition of 0.3% (w/v) xylose at an OD₆₀₀ of 0.4 0.5. Incubation was carried out at 37 °C under aerobic conditions. Cultures were harvested after 3 h of induction, and protein production was analysed by SDS-PAGE and Western blot across different cellular fractions. In parallel, OD₆₀₀ readings were taken to assess whether production of the synthetic silk proteins affected B. subtilis growth.

Accordingly, the TUC14 (monomeric RPE, BBa_255CP6QO), TUC16 (tetrameric RPE, BBa_25VXKHEZ) and TUC18 (octameric RPE, BBa_25AE0AHE) strains, which all have a His₆-tag within their yoaW secretion signal peptide, were cultured. Their corresponding growth curves, alongside those of controls strains, are presented in Figure 27.

Initially, all constructs and the wild types showed similar growth. Following induction after the measurements of T₂ (150 min) the uninduced TUC14 and wild-type control strains demonstrated a steeper rise in their final OD₆₀₀ compared to the induced production strains. Notably, the most pronounced impact on growth was seen in the TUC14 strain, which carried the shortest protein construct. In contrast, TUC16 and TUC18 reached final OD₆₀₀ values comparable to the induced wild-type control. However, both TUC16 and TUC18 already displayed slightly elevated OD₆₀₀ values at the point of induction, corresponding to a smaller increase in OD₆₀₀ post-induction compared to the controls.

This moderate growth inhibition in the induced production strains, relative to the uninduced controls, suggested active production of the synthetic pyriform silk proteins. Due to limited data points and a lack of biological replicates, these observations remain preliminary and require further validation. As time constraints prevented additional tests, direct protein detection via SDS-PAGE and Western blotting was proceeded with.

Cell pellets from the cultures were lysed using the previously described method. The whole cell lysate fraction was analysed first. Despite the inclusion of secretion signal peptides in the constructs, intracellular retention of the proteins was expected if production occurred. Figure 28A presents the SDS-PAGE gel of the whole cell lysates. Due to insufficient sample loading, no bands were observed in the uninduced wild-type control. Nevertheless, no significant differences were detected across all lanes. A His-tagged msfGFP served as a positive control (ctrl.) for the Western blot and produced a distinct band just below 34 kDa, corresponding to the expected molecular weight (yellow arrow, 26.8 kDa). Faint bands above 250 kDa were visible in both TUC16 and TUC18 samples but were absent in the TUC14 lane. However, these high-molecular-weight bands were also present in the induced wild-type control, making specific detection of synthetic spidroins in this region unlikely. No visible bands were detected at the expected molecular weights of the target proteins (green arrow: TUC14/SD4 – 102 kDa; pink arrow: TUC16/SD5 – 165 kDa; blue arrow: TUC18/SD6 – 249 kDa).

Given the high background observed in the whole cell lysate gel, the same samples were subjected to SDS-PAGE again, followed by electroblotting onto a polyvinylidene fluoride (PVDF) membrane for Western blot analysis. After blocking, the membrane was incubated with a conjugated anti-hexahistidine-peroxidase-antibody and treated with the colorimetric TMB Enhanced One Component HRP Membrane Substrate for specific detection of His-tagged proteins. The resulting blot is shown in Figure 28B.

Unfortunately, none of the synthetic protein constructs could be detected via Western blot, with the exception of the His-tagged msfGFP positive control, which produced a distinct and prominent band just below 34 kDa - corresponding to its expected molecular weight of 26.8 kDa (yellow arrow).

Subsequently, additional cellular fractions obtained during sample preparation were analysed to improve protein separation and increase the chances of detecting faint and distinct bands. Secreted proteins were isolated by precipitating them with TCA from the culture supernatant and resuspending the precipitate in SDS sample buffer. Furthermore, the whole cell lysate was fractionated by centrifugation to separate cytoplasmic proteins from the cell debris/membrane-associated protein fraction. However, SDS-PAGE and Western blot analyses of all these fractions yielded results comparable to those of the whole cell lysate.

Since no bands corresponding to the expected molecular weights of the synthetic protein constructs were visible in either the SDS-PAGE gels or Western blots, it was concluded that the pyriform silk proteins were either not produced or at levels too low to be detected by these methods. Due to time limitations, it was not possible to optimise the production protocol, which might have improved protein yields, or to test different constructs with alternative promoters. Instead, protein production was tested using the highly sensitive Nano-Glo® HiBiT Extracellular Detection System, which is capable of detecting proteins produced at low abundance (Hall et al., 2012; Miao, Zhou & Wang, 2024).

Detection via NanoLuc assay

The luminescence measurements are summarised and shown in Figure 29. Due to time constraints, most samples were measured in single replicates, however, two biological replicates were obtained for the induced wild type and TUC14 samples.

Induced TUC16 exhibited the strongest cell associated luminescence signal at endpoint, accompanied by a clear increase in supernatant luminescence, indicating robust pysp1 expression with measurable extracellular presence under 20 µg/mL bacitracin after LB blanking and normalisation to OD₆₀₀ = 1 with 1000 ms integration time. WT and TUC14 showed low signals in both fractions, with only modest induction effects, consistent with minimal reporter activity under these conditions and emphasising the contrast with TUC16. TUC18 displayed an elevated supernatant signal upon induction, suggesting expression and potential secretion, but the corresponding cell pellet measurement was unavailable due to a technical issue, so intracellular accumulation for TUC18 cannot be directly compared to TUC16 in this dataset. Measurements with error bars reflect two replicates and appear most consistent for induced TUC16, whereas single bars without error bars represent single measurements and should be interpreted with caution regarding variance.

Producing native spidroins through the farming or milking of spiders is unfeasible due to their solitary lifestyle and cannibalistic behaviour (Vierra et al., 2011; Ramezaniaghdam, Nahdi & Reski, 2022). For this reason, recombinant spidroin production has emerged as the only feasible industrial route to obtain spider silk. Yet despite decades of research, recombinant production has encountered major obstacles, with consistently low yields generally attributed to inefficient transcription and translation as well as the metabolic burden imposed on the host cell (Fahnestock & Irwin, 1999; Xia et al., 2010; Connor, Zha & Koffas, 2024).

Even though the coding sequence of pysp1 was codon harmonised and the intracellular and secretory pysp1 production attempted, and extensive optimisation, no detectable protein was observed by SDS–PAGE or Western blot. Only very low production signals were detected by the highly sensitive Nano-Glo® HiBiT luminescence assay (Hall et al., 2012; Miao, Zhou & Wang, 2024). The poor yields were hypothesised to result from secretion blockage via the SecYEG complex, possibly caused by premature folding of the synthetic pysp1 during translocation (Jiang, Wynne & Huber, 2021).

Despite employing promoter–RBS combinations that demonstrated strong reporter activity, spidroin production remained undetectable in SDS–PAGE and Western blot regardless of growth medium (LB or 2xYT) or induction temperature (37°C or 18°C). This aligns with recurring challenges in heterologous spidroin production: translational errors, metabolic burdens, or outright construct toxicity (Ramezaniaghdam, Nahdi & Reski, 2022). Similarly, Connor, Zha & Koffas were unable to detect spidroins in Bacillus megaterium, again pointing to translation and folding incompatibilities when expressed intracellularly. The authors of that work even speculated that certain spider silk proteins may intrinsically require a secretory environment for productive folding. Accordingly, the authors could show the secretory production of the same spidroins construct (Connor, Zha & Koffas, 2024).

The possible reasons for the lack or extremely low level of spidroin synthesis centre on transcriptional and translational challenges associated with the long length and highly repetitive composition of pysp1 (Connor, Zha & Koffas, 2024). Prokaryotic systems can suffer from premature ribosomal pausing and termination when tRNA pools become depleted in response to overused codons or amino acid requirements (Fahnestock & Irwin, 1999; Xia et al., 2010; Snoeck, Guidi & De Mey, 2024). Polyproline-rich motifs exacerbate this by inducing ribosome stalling (Park et al., 2024; Hong et al., 2024). When stalling occurs, rescue pathways recognise incomplete peptides and transcripts, targeting both for degradation (Buskirk & Green, 2017). While the synthetic spidroin coding sequence here had been codon-harmonised to better match B. subtilis codon usage, spidroin length and repetitive content still made premature terminations and stalling likely. Due to time and resource constraints, the harmonised versus original sequences were not directly compared, leaving unresolved how effective harmonisation truly was in avoiding these issues. Altogether, truncation, misfolding, and low-level expression appear plausible outcomes under these translational stresses (Connor et al., 2023; Xia et al., 2010; Snoeck, Guidi & De Mey, 2024).

The hypothesis that aggregates might explain the lack of SDS–PAGE or Western blot detection seems less likely, given that whole-cell lysates were analysed and no spidroin signals were present. Instead, stress responses and proteolytic degradation are more consistent with the data. Previous studies have similarly reported spidroin misfolding in intracellular hosts followed by degradation (Snoeck, Guidi & De Mey, 2024; Jin et al., 2022).

Growth analysis provided parallel evidence for translational burden. Under intracellular production in LB medium, B. subtilis strains carrying spidroin constructs grew comparably to wild-type. However, in nutrient-rich 2xYT medium, recombinant strains showed clear growth impairments, strongly suggesting translational errors visibly impacted physiology (Connor, Zha & Koffas, 2024). However, the secretory production of spidroin constructs similarly led to growth impairments in LB medium. Overproduction of silk proteins is documented to retard bacterial cell growth due to errors in translation (Heidebrecht & Scheibel, 2013). Intriguingly, spidroin expression driven by the st7 RBS caused stronger growth defects than st11, despite st11’s higher predicted initiation rate. Since RBS function depends strongly on downstream coding context (Mutalik et al., 2013), and stress burdens result from multiple interacting processes, the data suggest a nonlinear and complex relationship between translational initiation and cellular stress.

The regulatory framework contributed further complications. The liaI promoter driving expression is controlled by the LiaRS two-component system, which responds to bacitracin (Toymentseva et al., 2012; Radeck et al., 2013). Upon sensing bacitracin, LiaS activates LiaR, which induces via PliaI liaIH expression. Its promoter activity is tightly regulated, concentration-dependent, and transient, falling after about two hours due to native stress response dynamics (Castillo-Hair et al., 2019). Kesel et al. showed that reducing bacitracin from 20 µg/mL to 3 µg/mL halved expression rates, and in this study, lowering the inducer indeed improved growth of recombinant strains. This suggests that spidroin-induced growth inhibition was directly proportional to expression levels. Timing of induction, whether in early or mid-exponential phase, made no difference, further supporting a direct toxicity effect. As production in this study was shifted to 18°C post-induction, prolonged promoter activity was hypothesised, though unconfirmed empirically.

Another major explanation for failure lies in proteolysis. B. subtilis maintains protein quality through multiple cytosolic and extracellular chaperones, including DnaK, GroEL, and trigger factor (Moliere & Turgay, 2009), and at least seven AAA+ proteases (Matavacas & von Wachenfeldt, 2022). These systems prevent misfolded protein accumulation but may degrade misfolded spidroins. Moreover, spiders employ specialised chaperones in their silk glands for spidroin folding (Miserez, Yu & Mohammadi, 2023). In recombinant contexts, without such chaperones, disordered spidroins may overwhelm folding machinery, induce stress, and undergo degradation (Li, Zhou & Lu, 2004; Tjalsma et al., 2004). Notably, in 2xYT medium at 18°C, recombinant B. subtilis showed an SDS–PAGE band at ~60 kDa, more pronounced than in control strains. This band likely corresponded to the cytosolic chaperone DnaK (66 kDa; UniProt: P17820), confirming chaperone upregulation during stress. While nutrient-rich media provided more amino acids favouring spidroin translation, they simultaneously worsened protein folding stress responses.

The final part of the study investigated constructs of different repeat lengths, monomer (single repeat) and tetramer (four repeats), under secretion signals sslipA and ssyoaW, each controlled by either constitutive promoters (PxylA, PhpaII) or the inducible PliaI. No colonies formed with constitutive constructs, indicating lethal stress from unregulated expression. After prolonged incubation at 30°C, a few tiny colonies appeared but failed to grow in liquid culture. By contrast, colonies with PliaI constructs grew normally pre-induction, consistent with tight promoter regulation. Once induced, growth rates declined sharply in all recombinant strains compared to wild-type, implying that cells attempted spidroin expression but could not sustain it. Expected protein bands (~85 kDa for monomer constructs with signal peptides, ~135 kDa for tetramers) were absent from SDS–PAGE. Instead, distinct bands at ~41 kDa likely reflected DnaJ (40.7 kDa) or LiaS (40.5 kDa), direct evidence of molecular chaperone or cell envelope stress responses (Jordan et al., 2006). Bands near 32 kDa may be PrsA, a foldase anchored to the membrane where it refolds proteins post-translocation (Neef et al., 2020). This strongly suggests folding and secretion stress induction but degradation of spidroins before detection. Quantitative RT PCR of stress-related genes would be useful in future to verify upregulation of these networks (Livak & Schmittgen, 2001).

Western blot experiments confirmed these observations: GFP-His₆ or mNG-His₆ controls were detected, confirming the experimental validity, but neither monomer nor tetramer spidroins produced bands. Possible explanations include (i) low expression below detection limits, (ii) instability and degradation due to absent protease inhibitors during sample prep, or (iii) losses during fractionation. Mass spectrometry was proposed as a more sensitive detection method capable of confirming extremely low expression by quantifying peptide fragments (Connor et al., 2023).

Switching our detection strategy to the more sensitive HiBiT pep86 luminescence tag system, showed that spidroins were indeed produced at very low levels in PliaI-driven monomer, tetramer, and even octamer constructs, expressed from single-copy chromosomal integrations. Increased luminescence in both cell pellet and supernatant fractions suggested limited secretion also occurred (Dixon et al., 2016; Pereira et al., 2019). Accordingly, the lack of detection in SDS–PAGE or Western blot here may be due to low copy number, insufficient translation resources, or unstable expression.

The translational resource limitations are highly plausible. Even codon-harmonised constructs continued to rely heavily on glutamine, glycine, and proline. High demand for tRNAs corresponding to these amino acids likely overwhelmed natural pools. Consistent with this, Yang et al. (2016) showed that supplementing glycyl-tRNA enabled E. coli to produce glycine-rich silk proteins, and Connor et al. reported that supplementing amino acids improved silk production by 135%, achieving titres of ~100 mg/L for small silk proteins. In the present study, neither tRNA supplementation nor amino acid feeding was employed, making these limitations critical bottlenecks.

In conclusion, the results from this study show that recombinant B. subtilis initiates spidroin gene expression, as evidenced by growth impairments and stress protein upregulation, but fails to produce detectable spidroins in gels or blots. Both intracellular and secretory approaches produced consistent outcomes: severe cellular stress, growth inhibition, chaperone upregulation, and either degradation or non-accumulation of spidroins. These findings mirror other unsuccessful attempts in prokaryotic systems and confirm that full-length spidroins present extraordinary translational and folding challenges. The combined data strongly suggest that yield was limited by translational stalling, tRNA depletion, amino acid resource constraints, premature folding during secretion, and subsequent proteolytic degradation. Sensitive reporter assays confirmed trace expression, but titres were orders of magnitude below the levels required for downstream applications. The findings underscore the importance of translational resource engineering, codon and amino acid alignment, co-expression of chaperones, and alternative detection methods. Until these limitations are addressed, large repetitive spidroins like pysp1 remain unsuitable for reliable recombinant production in B. subtilis.

Biopolymers can serve multiple functions in MICP-based applications: they retain moisture, offer nucleation sites for calcite formation, bridge gaps between particles, and improve the distribution of bacterial cultures within the substrate (Jang, 2020; Soldo et al., 2020). Moreover, biopolymer-treated materials showed improved resistance to environmental conditions through improved compressive strength, tensile strength, and elasticity (Soldo et al., 2020).

Xanthan gum, in particular, is known for its ability to increase cohesion in soil, fill sand pores, and improve erosion resistance by enhancing water retention, all of which contribute to improved mechanical properties such as compressive and tensile strength (Gioia and Ciriello 2006; Jang 2020). Its gel-like matrix helps bridge gaps between particles, which is particularly valuable in bio-consolidated sand, where large pore spaces can weaken the material (Gioia & Ciriello, 2006; Lai et al., 2024). By supporting bacterial attachment and offering nucleation sites for calcite, xanthan may help overcome this limitation and improve the overall effectiveness of MICP (Jang, 2020; Lai et al., 2024).

As a proof of concept, the impact of xanthan, a widely used biopolymer produced by the bacterium Xanthomonas campestris (Bhat et al. 2022), on the biocementation of sand bricks via MICP, facilitated by the Solibacillus silvestris isolate CGN12, was investigated.

Initially, xanthan was produced from production cultures of X. campestris, which were incubated for over a week. Following incubation, xanthan was purified and subsequently lyophilised. The purified xanthan was then added to the biocementation medium (YAC medium) at varying concentrations.

A fixed volume of sand was added to individual wells of a silicone mold and mixed with a defined number of CGN12 cells resuspended in 5 mL of YAC medium, supplemented with xanthan concentrations ranging from 0 1% (w/v). The experimental setup is shown in Figure 31A. Over time, the sand bricks began to dry, as shown in Figure 31B. Each sand brick received at least one additional round of YAC medium (without xanthan) before being allowed to dry completely.

As the experimental setup was intentionally non-sterile, to better reflect potential real-world applications where sterile conditions cannot be guaranteed, microbial contaminations began to emerge after several days. This is visible in the right wells of conditions 2 and 3 in Figure 31B.

In the initial biocementation experiment, two conditions were compared: pure YAC medium and YAC medium supplemented with 1% (w/v) xanthan. Both conditions were inoculated with Solibacillus silvestris CGN12 cells to a final OD₆₀₀ of 2.7. Following inoculation, the sand bricks were left uncovered to dry overnight. On the following day, 3 mL of fresh YAC medium (without xanthan) was added to each well, just enough to form a thin liquid film. The bricks were again allowed to dry completely.

Once fully dried, the sand bricks were removed from the silicone molds for qualitative assessment of their mechanical properties. As shown in Figure 32A and B, the sand brick formed in the presence of xanthan was noticeably thicker and more structurally cohesive than the one formed without xanthan. Notably, the non-xanthan brick fractured during removal, indicating lower mechanical robustness. In contrast, the xanthan-containing brick retained its integrity, displayed sharper edges, and appeared more consolidated.

These observations were further supported by the residual sand left in the wells, as shown in Figure 32C. While the cylindrical parts of the bricks appear to be filled more in the well of the xanthan brick, the majority of the loose sand on the images in Figure 32 comes from the non-xanthan brick. Additionally, the solidified cylinders from the xanthan condition were sufficiently hardened to resist removal, whereas most of the sand from the cylinders of the non-xanthan condition disintegrated and fell out during extraction from the mold.

These results suggest that the addition of xanthan at the tested concentration enhanced the biocementation of sand bricks. However, it remained unclear whether the observed improvement was primarily due to increased MICP or simply the precipitation of salts from the medium in the presence of xanthan.

To differentiate between these effects, an additional experiment was conducted including a control condition in which only YAC medium supplemented with 1% (w/v) xanthan was added to the sand, without the addition of CGN12 cells. Replicates of the other conditions were also prepared, and the effect of a lower xanthan concentration (0.5% (w/v)) was tested as well. To emphasise the MICP-related effects over salt precipitation, this experiment involved covering the sand bricks with cling film after the medium was added, thereby reducing evaporation. After two days of incubation, the cling film was removed, and 4 mL of fresh YAC medium (without additional xanthan) was added. The bricks were then again covered with cling film for two days before being left to dry completely.

The resulting bricks are shown in Figure 33. Due to the modified incubation protocol, microbial contaminations had time to establish before the medium had dried, as indicated by the orange arrows in Figure 33A. Upon removal of the bricks, visible contaminations were also observed within the bricks themselves, as seen in the red dots in Figure 33B. Despite this, biocementation appeared to have taken place across all conditions.

Unexpectedly, the sand brick generated using only YAC medium supplemented with 1% (w/v) xanthan and no CGN12 cells (condition 1) was among the thickest and most consolidated, along with one of the replicates containing 0.5% (w/v) xanthan and CGN12 cells (condition 3, upper brick on the right). Notably, both of these conditions also showed strong signs of contamination. In contrast, the bricks produced using CGN12 cells but no xanthan (condition 2) were the weakest, with two breaking during removal from the mold and appearing the thinnest overall. The sand bricks produced with CGN12 cells and xanthan (0.5% or 1% (w/v)) were visually similar to each other and displayed greater mechanical integrity compared to those formed without xanthan.

These results further reinforce the findings of the initial experiment, indicating that xanthan enhances the biocementation of sand bricks in YAC medium at the tested concentrations. Furthermore, they proved the concept of introducing a biopolymer to enhance the mechanical properties of a biocemented sand brick.

Microbially induced calcite precipitation (MICP) is a promising alternative to cement-based binders for sustainable consolidation of granular materials (Achal & Mukherjee, 2015). However, MICP-treated materials often show limited mechanical performance (Castro-Alonso et al., 2019; Gebru et al., 2021). To address this, recent studies have explored biopolymer additives to improve matrix cohesion, bacterial immobilisation, and precipitation uniformity (Dubey et al., 2022). In this study, xanthan gum was evaluated as a model biopolymer additive in a non-sterile proof-of-concept MICP setup. Xanthan was chosen as a substitute for pyriform silk, which could not yet be obtained in sufficient quantities for testing, but which we ultimately aim to explore for biocementation applications.

Sand bricks formed with xanthan (0.5–1% w/v) were visibly stronger and more cohesive than those without it (Figure 32, Figure 33). Interestingly, even samples treated with 1% xanthan but no bacteria were among the most consolidated. Biopolymers such as xanthan are thought to enhance MICP products by distributing bacteria more evenly, filling interstitial spaces, and providing nucleation sites (Lai et al., 2024; Li et al., 2022; Xu et al., 2019). In this case, however, the observed consolidation may have been largely abiotic, possibly due to salt crystallisation or polymer crosslinking during drying. As this was only a single qualitative experiment, replication and quantitative testing are needed. Still, the results agree with previous studies showing that biopolymers can significantly improve compressive strength and cohesion in MICP-treated soils (Chang et al., 2016; Dubey et al., 2022).

The chemical properties of xanthan provide additional mechanistic insights. Its carboxyl groups interact with divalent cations like Ca²⁺ (Groves & Chaw, 2015; Maity & Sa, 2014), dynamically altering viscosity (Fjodorova et al., 2022) and forming a gel-like matrix (Figure 34A). Such a matrix increases contact between bacteria, substrates, and soil particles, promoting localized supersaturation (Mendonça et al. 2021). Xanthan gels have also been used for bacterial entrapment in industrial processes (Dzionek et al., 2021) and shown to enhance bacterial adhesion during MICP (Li et al., 2022). Furthermore, calcite precipitation could shift equilibrium to release bound Ca²⁺, enabling slow ion release and more uniform mineralization. Binding Ca²⁺ may also buffer against inhibitory concentrations for strain CGN12 (Seidel et al., 2025). These mechanisms support more efficient nucleation and crystal growth (Hoffmann et al., 2021; Lai et al., 2024).

At higher xanthan levels (>1% w/v), however, negative effects are likely. Excessive Ca^²⁺ sequestration can reduce free ion availability, while blocked pores may prevent nutrient diffusion and bacterial migration (Lai et al., 2024; Rebata-Landa, 2007). Thus, xanthan’s role appears concentration-dependent, with benefits at moderate levels but inhibition at higher ones (Figure 34B).

The non-sterile experimental setup introduced another layer of complexity. Because MICP is widespread among soil microbes (Lowenstam, 1981), contaminating organisms may have altered calcite precipitation. Still, even the most consolidated bricks showed contamination (Figure 32), suggesting that MICP biocementation is resilient against microbial competition.

The non-sterile experimental setup introduced another layer of complexity. Because MICP is widespread among soil microbes (Lowenstam, 1981), contaminating organisms may have altered calcite precipitation. Still, even the most consolidated bricks showed contamination (Figure 32), suggesting that MICP biocementation is resilient against microbial competition.

While xanthan served here as a stand-in for pyriform silk, parallels with other protein-based polymers are promising. For example, silk fibroins from Bombyx mori cross-linked with genipin enhanced both biomineralisation and compressive strength in MICP-treated sand (Li & Achal, 2023). Likewise, spider silk proteins have proven effective substrates for calcite biomineralisation in biomedical studies (Dmitrovic et al., 2016; Kiseleva et al., 2020; Mehta & Hede, 2008). These findings strengthen the rationale for testing pyriform silk once sufficient quantities are available.

Overall, xanthan addition improved sand biocementation and highlighted the importance of biopolymer chemistry in shaping MICP processes. Optimising polymer concentration, ion dynamics, and microbial activity will be crucial to maximise performance. Having established xanthan as a functional stand-in for pyriform silk and demonstrated its potential to enhance MICP-driven biocementation, we next turned to curdlan, a structurally distinct β-(1,3)-glucan. Unlike xanthan, curdlan forms strong thermo-reversible gels and has been proposed as a stabilising matrix in composite materials.