Spider silk has unique physical properties that make it attractive to utilise in applications that require lightweight materials with high tensile strength. Its properties include not just exceptional strength and toughness, but also high flexibility, unique anisotropy, and in situ biological inertness (Heim et al., 2009; Vollrath et al., 2013; Blamires et al., 2020; Blamires, 2022; Linklater et al., 2023). The successful replication of spider silk’s properties within fibres spun in the laboratory would undoubtedly open exciting opportunities for the production of extremely high-performing materials for applications in medicine, textiles, wearables, and materials engineering (Kluge et al., 2008; Leal-Egaña and Scheibel, 2010; Humenik et al., 2011; Heidebrecht et al., 2015; Blamires, 2022, Blamires, 2024).

Spiders are solitary and cannibalistic, and they need to be anaesthetised for silk to be spooled from their spinnerets for several hours at a time. Moreover, individual spiders yield proportionally smaller masses of silk than can be conventionally attained from an individual silkworm cocoon. Accordingly, it is not feasible to harvest spider silk by the same methods as silkworm silk is harvested (Blamires, 2022). The only way to possibly generate large amounts of spider silk, or spider silk-like material, is by the recombinant production of spider silk proteins (so called spidroins) followed by purification and the utilisation of advanced wet or dry fibre spinning methodologies to facilitate spinning the proteins into robust fibres (Lewis et al., 1996; Scheller et al., 2001; Wong Po Foo and Kaplan, 2002).

Recombinant spider silk proteins (or more strictly speaking proteins inspired by spider silk sequences) are produced using techniques that involve, firstly, the synthesis of a gene encoding the spidroin-like construct, insertion of this construct into a synthetic expression vector (such as plasmid DNA) and then transforming an expression host (such as Escherichia coli, yeast (Saccharomyces cerevisiae), plants (e.g. potato, tobacco) or animals (e.g. silkworms, hamsters, goats) (Scheller et al., 2001; Menassa et al., 2004; Miao et al., 2006; Chung et al., 2012; Teulé et al., 2012; Mi et al., 2023). Spidroins are produced by the host’s protein expression system, which can then be isolated and concentrated into a solution called dope. Generating recombinantly engineered synthetic fibres with physical qualities that resemble native spider silk fibres, however, remains one of material science’s grandest challenges (Humenik et al., 2011; Jastrzebska et al., 2016; Blamires, 2024). Some specific issues associated with producing recombinant spider silk proteins, and the spinning of fibres from those proteins, include inducing the hosts to express recombinant spidroins, and obtaining a sufficient quantity of proteins at concentrations high enough to spin the resulting dope into exceptionally robust fibres (Copeland et al., 2015; Harris et al., 2016)

Inducing host organisms to express full-length spidroins necessitates the use of sophisticated techniques, some of which we overview herein. Recent advances in DNA sequencing, DNA synthesis and improved protein structural prediction have enabled an acceleration in the construction of genes encoding recombinant proteins inspired by spider silk proteins. However, many difficulties remain in terms of expressing and purifying these constructs in a scalable manner (Scheller et al., 2001; Schacht and Scheibel, 2014; Rising and Johansson, 2015; Koeppel and Holland, 2017; Edlund et al., 2018).

Difficulties associated with expressing the large repetitive amino acid sequences of the spidroins include plasmid stability, difficulties in host-cell translation, the potential of protein misfolding that accompanies the expression of large and repetitive proteins and the potential toxicity of certain sequences to the host cells (Heim et al., 2010; Rising and Johansson, 2015). Purification of adequate quantities of the proteins yielded is another challenge. Indeed, efficient purification of high-quality samples of recombinant proteins is critical for the implementation of all of the downstream applications (Oliveira and Domingues, 2018; Tripathi and Shrivastava, 2019). This is all the more critical for the synthesis of spider dragline, or major ampullate, silks (i.e.,those silk whose toughness has been compared to that of Kevlar and carbon fibres; Blamires, 2022) because the quality of the fibres spun from these proteins is a product of the concentrations of the spidroins and factors associated with spinning protocols (Humenik et al., 2011; Liu et al., 2014; Doblhofer et al., 2015; Peng et al., 2016). Thus, we detail herein how the gene construction, protein expression and purification, structural characterisation protocols, and downstream fibre fabrication methodologies are performed in vitro to generate native-like spidroin-like fibres.

Much of the reported work on recombinant spider silk protein expression has focused on the spider silk fibroins (known as spidroins). The first reported construction of a recombinant spider silk protein was in 1995, when constructs of varying length inspired by both Major Ampullate spidroin 1 and 2 (shortened conventionally to MaSp1 and Masp2) proteins from the golden orb web spider Triconephila clavipes were created using the gram-negative bacteria E. coli as the host (Prince et al., 1995). Since then, there have been many reports describing strategies to produce recombinant silks for research and commercial development (Tokareva et al., 2014). The primary advantage of spider silk protein expression by recombinant techniques is that it has a greater capacity for larger scale production than does collecting the silk directly from spiders.

Much of the work thus far has focused on Major Ampullate Spidroin Proteins (MaSp), MaSp1 and MaSp2 due to their contribution to the strength and elasticity of dragline silk fibres. The remarkable properties of MaSp1 and MaSp2 can be attributed to their composition, i.e. a non-repetitive and partially conserved N- and C-terminus that flanks a repetitive core domain (Ayoub et al., 2007a; Römer and Scheibel, 2008; Gaines et al., 2010; Andersson et al., 2017; Xu et al., 2017). Recombinantly-derived sequences have hence been generated using conserved N- and C-terminal domains from single or different spider species, creating so called “chimeric” structures. These constructs also commonly have varying repeating numbers of conserved glycine- and alanine-rich domains, which create polypeptides of shorter length than the native proteins ( Figure 1A ).

Many biotech companies are now developing recombinant spider silk fibres using genetically modified organisms such as bacteria, yeast, and silkworms with some success. However, there are challenge associated with the limitations of such hosts to express recombinant proteins at high levels and/or consistently produce quality fibres.

The solubility and spinning characteristics are heavily influenced by both the N-terminus (NT) and the C-terminus (CT) of the silk proteins. The NT and CT regions are not only conserved amongst species but also spidroin types (Dicko et al., 2004; Ayoub et al., 2007a; Kronqvist et al., 2014). It was determined that across 12 different species, and multiple fibre types, more than 50% of the amino acids were congruent (Ayoub et al., 2007a). Such similarity amongst N- and C-termini between species and spidroins suggest they are integral in the assembly of spider silk proteins which must be considered when designing constructs for recombinant expression (Dicko et al., 2004; De Oliveira et al., 2024). This has indeed been confirmed by studies inducing recombinant expression of the N- and C-terminal domains of major ampullate spidroins, whereby their role in fibre assembly solubilising effect and structural solving have been accomplished ( Figure 1A ) (Askarieh et al., 2010; Malay et al., 2020; De Oliveira et al., 2024).

While the terminal domains are responsible for polymerisation and drive a portion of the assembly process, the repetitive domain is predominantly responsible for the unique mechanical properties and intrinsically disordered behaviour of spidroin fibres in solution. As a result, repetitive domain sequences have been the source of extensive investigation due to their direct relationship to fibre mechanical properties.

Spidroin paralog and phylogenetic analyses have confirmed the convergence of key residues across sequences (Ayoub et al., 2007a; Craig et al., 2020; Arakawa et al., 2022). The abundance of comparatively smaller glycine, alanine, and serine residues within the low complexity repetitive region contribute most heavily to fibre structural flexibility as a consequence of disorder of the domains within the highly soluble regions of the silk gland, i.e., in and around the tail and sac ( Figures 1A, B ) (Oktaviani et al., 2018; Yang et al., 2023).

Consistent with many other proteins that have a high degree of intrinsic disorder, spidroins have demonstrated the ability to phase separate, a phenomenon where proteins separate into two phases: a proteinaceous dense phase and a buffer light phase (Malay et al., 2020; Mohammadi et al., 2020). The sticker-spacer model is often used to describe residue-residue interactions that drive liquid-liquid phase separation (LLPS). Stickers are residues which act as molecular glues and hold proteins together, whereas spacers are flexible residues that have low interaction strengths with other residues. In the context of spidroins abundant residues like glycine and alanine act as spacers, providing flexibility to the protein chain. Additionally, spacer residues also spatially control the interactions between less abundant but more reactive sticker residues like tyrosine, glutamine and proline which can enter π- π, π-cation and charge interactions ultimately driving phase separation. Understanding the relationship between sticker-spacers and their patterning provides a platform to develop de novo biomaterials for a diverse range of applications not limited to tissue engineering, synthetic organelles and encapsulants (Martin and Mittag, 2018; Martin et al., 2020; Maraldo et al., 2024).

Considering the repetitive nature and low complexity of the repetitive domain, several motifs can be observed and used to classify the subspecies of MaSp to which they belong (Ayoub et al., 2007a; Arakawa et al., 2022). Most MaSp constructs are primarily formed with two major subunits; a glycine rich GGX or GPGXX where “X” denotes any amino acid and often a sticker residue (Tokareva et al., 2013). These motifs are then typically followed by a poly-alanine sequence spanning 4-7 residues on average ( Figure 1B ) (Gaines et al., 2010). The spatial arrangement of these residues ensures that sticker-sticker interactions are finely balanced preventing premature aggregation, and the formation of highly concentrated dope.

As fibre formation proceeds, the role of these residues transforms. These repetitive motifs in response to stimuli changes like pH, shear and kosmotropic buffer exchange in the duct begin to form defined secondary structures that are responsible for the unique mechanical properties spidroins are well-known for (Dong et al., 1991; Gosline et al., 1999; Hayashi et al., 1999; Gu et al., 2020). The comparatively smaller size of the most abundant residues enables the formation of highly compact β-sheet content due to minimised interchain distance (Van Beek et al., 2002; Keten et al., 2010). Evidence suggests that poly-alanine sequences form hydrophobic β-sheet crystalline domains provide silk with high tensile strength, whereas the glycine-rich regions form hydrophilic extensible 3₁₀-helical structures and account for the elasticity of silk– ( Figure 1B ) (Dong et al., 1991; Kümmerlen et al., 1996; Ayoub et al., 2007a). Recent work comparing over 1000 spider silk species transcriptomic data has provided unparalleled insight into the relationship between the sequence-property relationship and serves as a platform for designing and modifying the aforementioned motifs to create fibres with tuneable properties (Arakawa et al., 2022).

Native spidroins are polypeptides of approximately 200–350 kDa (Kümmerlen et al., 1996; Hayashi et al., 1999; Van Beek et al., 2002; Chaw et al., 2017). Due to their length and the repetitive nature of MaSps, several challenges arise when utilising heterologous hosts for expression. When considering the two most common subclasses, MaSp1 repetitive sequences have been shown to be more homogenous than those of MaSp2 (Ayoub et al., 2007a; Craig et al., 2020). As a result, recombinant forms of MaSp1 have historically been more difficult to develop compared to MaSp2, in part because of limitations on producing highly repetitive synthetic gene sequence constructs.

Thankfully, several molecular biology techniques have been developed to enable the assembly of full length spidroin proteins from a single motif repeat, namely recursive directional ligation (McDaniel et al., 2010). Alternatively, recent circular RNA production serves as another promising avenue for generating native-length templates without the requirement for duplicating a single repeat, instead translational units loop over a single construct and insert fragments to generate multimeric repeats (Lee et al., 2021; Liu et al., 2022).

However, despite the ability to reliably produce vectors for recombinant expression, the main challenge lies in the lack of specialised tissue and translational machinery in the commonly used heterologous hosts (Candelas et al., 1990; Xia et al., 2010; Sonavane et al., 2024). As residue diversity is low within MaSp sequences, spiders have evolved to overexpress tRNAs for highly abundant residues like glycine and alanine to prevent translational bottlenecking (Candelas et al., 1990; Ayoub et al., 2007a; Xia et al., 2010; Tokareva et al, 2013). Comparatively, when expressing “native like” spidroins in heterologous hosts, a decrease or complete lack of expression compared to truncated forms is observed due to the exponential increase in the translational stress it places on the host (Xia et al., 2010).

Solutions for heterologous expression systems particularly within E. coli systems, have taken inspiration from spider physiology. Co-expression of recombinant forms of MaSp with a plasmid that supplies additional glycine and alanine t-RNAs showed significant improvement in expression (Xia et al., 2010; Cao et al., 2017). This co-expression strategy along with incubating the bacteria in media with elevated glycine and alanine levels allows expression of large “native like” spidroins with comparable yields to those of smaller-sized repeat proteins (Cao et al., 2017). Proteomic analyses have confirmed such expression patterns (Ramezaniaghdam et al., 2022). In conjunction with further codon optimisation, challenges pertaining to sequence repetition and size, on recombinant expression, can be mitigated when attempting to produce recombinant MaSps.

Liquid–liquid phase separation (LLPS) of proteins is a mechanism for the organisation and regulation of the intracellular environment with underlying driving forces, which typically include hydrophobic, aromatic, and/or electrostatic interactions regulated by temperature, (osmotic) pressure, or ion concentration (i.e., osmolyte, pH, salt concentration) (Dai et al., 2022). Protein phase separation is also known to be sensitive to protein concentration, sequence, and length. The aqueous solvent based LLPS of biopolymers such as proteins, leads to a dense phase in thermodynamic equilibrium with a dilute phase. The dispersed dense phase domains (biopolymer rich phase) are often referred to as coacervates or biomolecular condensates.

The formation of spider silk fibrils from soluble spidroin precursors during spinning is a complex process involving spidroin proteins encountering multiple chemical and physical gradients (e.g., numerous ion fluxes, water dehydration, CO₂, shear force). There is mounting evidence that LLPS plays a critical role in mediating the self-assembly and aggregation processes involved in producing the unique molecular and hierarchical structures involved in spider silk fibre spinning (Yarger et al., 2018; Mohammadi et al., 2020; Perera et al., 2023). Much of our current understanding of spider silk assembly via LLPS comes from experimental studies of recombinant spidroins (Lemetti et al., 2019, Lemetti et al., 2022). For example, probing of a recombinant MaSp2 showed phosphate ion-induced LLPS at neutral pH which resulted in the formation of coacervates (Malay et al., 2020). Furthermore, if the pH is lowered during spinning, as happens within the silk gland of spiders, the coacervates become fibrous ( Figure 1C ). It has also been shown through the microfluidic spinning of recombinant MaSp2 that polyalanine blocks have limited influence on the occurrence of LLPS and hierarchical structure (Chen et al., 2024). The role of the N- and C-terminal domains and repetitive domains in LLPS and resulting fibril formation have also been explored using recombinant proteins, providing strong evidence that the C-terminal domain, repetitive domain and several key residues play a central role in LLPS, and the N-terminal domain is essential for subsequent fibril assembly (Malay et al., 2020; Watanabe and Arakawa, 2023).

LLPS is also being explored as a potentially useful mechanism to control protein solubility during expression and accumulation in the cytoplasm of E. coli as well as a property fit for developing non-canonical materials, enabling the further development of recombinant spidroin proteins for biomimetic spider silk materials and biomedical applications (Gabryelczyk et al., 2022; Dai et al., 2022; Maraldo et al., 2024). Evidence of such has already been shown through the generation of a synthetic organelle capable of performing enzymatically driven reactions within puncta of the cell using a recombinantly expressed MaSp2 (Wei et al., 2020). The importance of sequence design in both the terminal and repetitive domains is therefore integral in dictating the assembly behaviour and subsequent mechanical properties of formed fibres.

The methodologies involved in spinning spider silk fibres from recombinantly expressed spidroins are complex and multifaceted, particularly when aiming to replicate the natural spinning process of spiders. As stated, the success of any methodology at reproducing the properties of native silk is dependent on producing a spinning dope at appropriate concentrations. The spinning process itself aims to emulate the conditions under which spiders naturally spin their silk. This includes an aqueous medium, ambient temperature, low hydrostatic pressure, slow extrusion rates, and low post-spinning draw-down ratios ( Figure 1C ). Researchers have used various methods to spin silk fibres from recombinant spider silk proteins, including microfluidic spinning and post-stretching techniques. These methods aim to align the spidroins and induce the formation of β-sheet structures (Chung et al., 2012; Leclerc et al., 2013; Peng et al., 2016).

Studies have shown that factors such as temperature, pH, and ionic balance within the spinning column, and the speed at which the fibres are spooled, play significant roles in affecting the mechanical properties of the silk ( Figure 1C ). Higher reeling speeds tend to increase the orientation of spidroin molecules, leading to stiffer and stronger, but less extensible, fibres (Shao and Vollrath, 2002; Shao et al., 2003; Gronau et al., 2013; Oktaviani et al., 2023). Additionally, the acidification of the spinning dope during extrusion aids in the transition of spidroins from random coil and α-helix conformations to β-sheet structures (Gaines et al., 2010; Peng et al., 2016; Oktaviani et al., 2023). The incorporation of post-treatment steps, such as post-drawing in air and ethanol solutions, has been shown to improve the crystalline and molecular structure of the fibres, enhancing their mechanical properties (Hudspeth et al., 2012; Doblhofer et al., 2015; Peng et al., 2016).

Many different hosts have been used for expressing recombinant spidroins ( Table 1 , Figure 2A ). The majority of reports describe the use of E. coli (Prince et al., 1995; Fahnestock and Bedzyk, 1997; Xu et al., 2007; Teulé et al., 2012). Reasons for the selection of E. coli include: (i) its ease of manipulation utilising plasmid transformation, (ii) extremely short regeneration times, and (iii) low cost and proven scale-up potential ( Figure 2B ). The methylotrophic yeast Pichia pastoris (now Komagatella phaffi) has an advantage over E. coli of being able to secrete the proteins and facilitate more complex post-translational modifications (Cregg et al., 2000; dos Santos-Pinto et al., 2014).

Many studies have reported a decrease in replication fidelity when expressing proteins with sizes larger than ~43 kDa (Xia et al., 2010; Teulé et al., 2012; Cao et al., 2017). However, for synthesising spider silks, expression of proteins of this size or greater is required to create fibres that approximate the mechanical properties of the natively spun fibres. While solutions we highlighted previously may address this problem, resolving expression remains a major challenge for recombinant spider silk production (Blamires, 2024).

For spinning spider silk fibres, a high level of purity is desirable for optimum downstream assembly. Contaminants from E. coli or other expression systems could either cause damage to the protein of interest, e.g., by degradation or obstruction of the assembly of fibres.

Before purification, the cells of the expression system, most commonly E. coli, must be lysed to release their contents. There are a variety of ways to go about this and the approach taken depends on the downstream use of the proteins ( Figure 2C ) (Islam et al., 2017). Commonly used methods include ultrasonication, which harnesses the cavitation of air bubbles, pressure homogenisation, multiple freeze/thaw cycles, or the use of chemical lysing agents (e.g. BugBuster) to “melt” the cells. Some methods may result in damage to the sample protein (e.g. when using ultrasonication due to sample warming) whereas others could be prohibitive when dealing with larger samples (e.g. bigger volumes may require more freeze/thaw cycles). It is therefore important to tailor the method used to the specific case and its requirements. The resultant cell lysate from the lysis step then undergoes a series of downstream purification steps to reach a desired level of purity. The overexpressed proteins may be soluble and isolated from the cell cytosol following cell lysis. Alternatively, during expression the protein may form insoluble “inclusion bodies” and require subsequent solubilisation and/or refolding (Tsumoto et al., 2003; Singh et al., 2015). The propensity to form these insoluble inclusion bodies is very protein and/or system dependent and thus a significant degree of optimisation and trial-and-error may be required during recombinant protein expression and purification experiments.

To achieve the required level of purity, a multistep purification protocol is generally recommended ( Figure 2D ). Most commonly this would begin with an affinity step, utilising a tag that has been incorporated into the expression construct. Many options are available (Pan et al., 2019), including the traditional 6-His tag or its newer equivalents HAT and NCTR₂₅ (based on a human protein to avoid immunogenicity in downstream applications; Young et al., 2012; Pan et al., 2019), which selectively binds to immobilised divalent metal cations like Ni²⁺ and Co²⁺ and can be released using an imidazole gradient. Others include the streptavidin-based tags eluted with biotin, maltose binding protein (MBP), which binds to amylose resin and is eluted using maltose, and glutathione-S-transferase, which binds to glutathione resin and is eluted with reduced glutathione. Recently a new generation of tags has been released including the Protein Select™ system by Cytiva™ (Clifford et al., 2024), in which the desired protein self-cleaves from its tag during purification leaving no residual amino acids. Alternative strategies to affinity chromatography include ion-exchange or hydrophobic interaction chromatography, which exploit the natural polarity of the protein construct in order to facilitate the separation of the target from the host cell contaminants. The initial steps, which involve either affinity, ion-exchange or hydrophobic interaction chromatography, all serve to achieve the bulk cleaning of the protein sample from contaminants prior to a later size-exclusion step (McCue, 2009; Cummins et al., 2017)

Spectroscopic analyses coupled with microscopy-based analytical methods offer a means for researchers to visualise protein structural organisation in silk fibres across hierarchical scales and spatial resolutions, thus, facilitating a greater understanding of how environmental factors can influence silk fibre properties.

Bright-field and confocal microscopy, in particular, have been deployed to effectively characterise spidroin phase separation (Zhang et al., 2023). Firstly, serving as a binary method to determine whether a two-phase state is observed, image analysis techniques such as circle detection offer valuable insights into droplet formation, growth, and coacervation kinetics, shedding light on the crucial LLPS pre-assembly step in fibril development ( Figure 3 ).

An extension of these methods is Fluorescence Recovery after Photobleaching (FRAP). Unlike image analysis methods, which assess overall droplet characteristics such as size, morphology, and fusion rates, FRAP is specifically used to determine the diffusion properties of proteins within the dense proteinaceous droplets they form, providing insights into the relationship between these proteins and their environment ( Figure 3 ). In this technique, a high-intensity laser bleaches a specific region within the imaging field, removing the fluorescence of a conjugated fluorophore or fluorescently expressed tag. The time it takes for fluorescence to recover indicates the diffusion rate and, ultimately, the mobility of proteins within the droplet (Meyvis et al., 1999; Sprague et al., 2004). Recent work has utilised FRAP analysis of a recombinantly expressed MaSp2 protein to determine physical property shifts when buffered into a series of environments with progressively decreasing pH (Lemetti et al., 2019; Malay et al., 2020; Feng et al., 2023). Samples with lower pH environments showed a significant decrease in fluorescence recovery compared to higher pH environments, characterising the transition from pre-assembly droplets into more rigid nanofibril structures (Malay et al., 2020). Methods like FRAP will become increasingly utilised to effectively measure changes in the physical properties of spidroins at a molecular level, allowing inferences on residue-residue interactions that drive structural development to be made.

In contrast to the dynamic, in situ observations provided by confocal microscopy scanning electron microscopy (SEM) and transmission electron microscopy (TEM) offer high-resolution, static imaging that enables precise quantification of the structural arrangement and fine details of nano to microscale entities allowing for a more detailed analysis of their architecture and organisation at the nanoscale (Henini, 2000; Putthanarat et al., 2000). TEM, which utilises electrons passing through a sample, has enabled the characterisation of spidroin protein arrangement in the pre-fibril phase, providing a better understanding of the structural organisation of various domains and the interactions between monomers within the network (Parent et al., 2018). Conversely, SEM provides high resolution images of the surface of a sample by scanning samples with an electron beam. Imaging fibre topography is extremely useful as it can rapidly determine stress, fractures and void formation in fibres (Stark et al., 2007; dos Santos-Pinto et al., 2018; Arakawa et al., 2022; Craig et al., 2022). SEM can be implemented as a quick and effective quality control step to ensure fibre uniformity during production.

Protein structure prediction programs are frequently employed to aid the design of protein constructs for expression and purification. As these predictions provide critical information about domain boundaries and allow researchers to determine the effect of mutations, it is imperative that these programs can predict protein structures with a high degree of accuracy.

The release of AlphaFold2 in 2021 has revolutionised the field of protein structure prediction with an increase in accuracy of the predicted structures even in instances where there are no known structures of homologous proteins (Jumper et al., 2021). Figures 4A, B provides an example of its use for predicting the structure of spider silk proteins. The accurate and reliable protein structure predictions afforded by AlphaFold2 have proven to be invaluable as a replacement or complement to X-ray crystallography and cryo-EM (Gupta et al., 2021; Pereira et al., 2021), thus enabling researchers to make predictions about protein structure by combining experimental data with integrative structural modelling.

ColabFold (Mirdita et al., 2022) further enhanced the accessibility of fast, accurate protein structure prediction by opening up AlphaFold2 and RosettaFold to the wider scientific community (Mirdita et al., 2022). ColabFold is five times faster than for single predictions than AlphaFold2 and AlphaFold-Colab. These improvements, coupled with AlphaFold multimer that opens up the possibility of investigating macromolecular complexes, meaning that complex and previously intractable protein systems – including spider silk proteins and fibres – are more explorable than ever (Evans et al., 2022). AlphaFold3, which was released during the preparation of this manuscript, offers even more exciting improvements, with increased speed and ease of use (Abramson et al., 2024).

As a result of these innovations, structural coverage of large proteomes has been estimated to have increased from 48% to 76% (Porta-Pardo et al., 2022). AlphaFold has also democratised the use of protein structures with millions of structure predictions for a range of species readily available via the AlphaFold database to researchers who had not routinely considered protein structure when presenting their findings (Varadi et al., 2022). Current limitations of AlphaFold include an inclination to overpredict folded regions. With the algorithm being largely trained on X-ray derived crystalline structures, there is an inability to include ligands in the predictions (although AlphaFold3 now offers the inclusion of certain ligands, ions, and post-translational modifications), and size limitations exist in the complexes that can be predicted. AlphaFold3 has improved this situation significantly but is still limited to 5000 ‘‘tokens’’ per run which equates to a maximum of about 14 MaSp1 units.

Many computational researchers have built on the foundations of AlphaFold introducing new algorithms that combine it with larger templating and other molecular dynamics-based methods to explore larger complexes and potential conformers and movements within them (Bryant et al., 2022; Mirabello et al., 2023). These experiments are producing vast, and difficult to interpret, arrays of data and offer opportunities for iterative and symbiotic marriages between experimental and computational work. The recombinant spider silk field seems ripe for such collaborations given the modular nature and unique properties of these materials.

Fold prediction tools are capable of estimating terminal domain structures and binding interactions of spidroins with high confidence. Their ability to confidently estimate the structure of the repetitive domain of spidroin proteins is, nevertheless, only capable of outputting low confidence and minimal structural conformations ( Figures 4A, B ). Recent developments, allow for comparative queries of sequences with residue-per-residue data output on the structure promoting and disorder promoting residues with a protein (Linding et al., 2003; Hu et al., 2021). Further to this, applying deep learning on proteome datasets has enabled accurate and efficient predictions of radius of gyration, end-to-end distance, polymer-scaling exponents, asphericity and phase separation propensity of proteins (Vendruscolo and Fuxreiter, 2022; Lotthammer et al., 2024) The ability to computationally determine these values provides an opportunity to compare experimentally compare native spidroins to recombinantly produced variants and allow more informed sequence design. While the surge in popularity of deep learning implementation at a generalised proteomic level has occurred, specialised spidroin tools have also been developed. These range from coarse-grained modelling of silk synthesis and behaviour throughout the gland to generative design of novel spidroin sequences with advanced mechanical properties trained on a dataset of over 1,000 silk sequences and motifs, annotated with physical properties such as elasticity, tensile strength, and water content ( Figure 4C ) (Lin et al., 2015; Ni et al., 2023; Lu et al., 2024). The wide range of in silico tools now available provides unprecedented potential to high-throughput screening and development of high-performing recombinant spidroins.

The production of recombinant spider silk proteins represents a promising yet complex endeavour that leverages advances in molecular biology, protein engineering, and materials science. While significant strides have been made in expressing and purifying the requisite spidroins using heterologous hosts, challenges remain in replicating the natural silk spinning process, particularly with regard to maintaining protein stability, scaling up production, and emulating the hierarchical structural features of native spider silk. Continued research into the roles of the N- and C-termini, as well as the repetitive domains of spidroins, will be vital in overcoming these obstacles. Further investigation into liquid–liquid phase separation and other pre-assembly processes offers exciting avenues for enhancing the quality and consistency of the fibres produced. Coupled with advances in analytical techniques such as mass photometry and spectrofluorimetric analysis, the future of recombinant spider silk production lies in optimising both expression systems and spinning methodologies to develop economically viable methods to create fibres with properties approaching those of their natural counterparts.

Addressing the urgency and direction for future research should prioritise understanding the unique silk gland environments across spider species, as this may hold key insights into the optimisation of synthetic spinning techniques. How factors such as the roles of pH shifts, shear forces, ion concentrations and duct physiology influence the silks mechanical properties requires closer examination. Comparative studies across species could reveal crucial differences in gland morphology and function, offering clues to refining recombinant processes. Additionally, advancements in bioengineering approaches, such as gene editing and expression systems, alongside improvements in synthetic spinning methodologies, are likely to drive advancements in fibre quality and production scalability.