Switchable inteins for conditional protein splicing

Introduction

Inteins are to proteins what introns are to mRNA: they are elements that occur within other proteins that eventually leave in a process called protein splicing (Hirata et al., 1990; Kane et al., 1990). Protein splicing is an autocatalytic process that concludes itself with the formation of a peptide bond between the two polypeptides originally flanking the intein (called N- and C-exteins depending on their location in respect to the intein) (Supplementary Figure S1A). Inteins may play regulatory roles at the DNA, mRNA and protein levels, but do not contribute to the function of the mature proteins formed after splicing.

In some cases, inteins are found naturally split in two parts, called N- and C-terminal intein fragments (later on indicated as Int^N and Int^C, respectively). Each intein fragment resides within a different polypeptide. After finding each other, the two fragments constitute a complex capable of trans-splicing just like a contiguous intein (Supplementary Figure S1A). Regardless whether contiguous or split, inteins use a common series of chemical reactions to splice themselves out of the host (Supplementary Figure S1A). The canonical splicing reaction requires a serine or cysteine as first amino acid at the N-terminus of the intein (called position 1) that engages in a N-S or N-O acyl shift with the peptide bond at the N-terminal splicing junction between position 1 of the intein and position −1 (i.e. the last residue in the N-extein just upstream of the intein). This results in the formation of a linear (thio)ester intermediate. A nucleophilic attack by the first residue in the C-extein (either a cysteine, serine or threonine; called position +1) is responsible for the trans-(thio)esterification that leads to the formation of a branched (thio)ester intermediate. This is then resolved by cyclization of the conserved C-terminal asparagine in the intein to form a succinimide and cleave the peptide bond at the C-terminal splicing junction. Finally, the (thio)ester between the liberated exteins undergoes an S-N or O-N acyl shift to form a stable peptide bond. The succinimide of the intein slowly hydrolyzes. The steps involved in the splicing mechanism rely on residues that lie close to the termini of the intein which are in most cases well conserved. Beyond the canonical one, alternative splicing pathways have also been reported (Southworth et al., 2000; Tori et al., 2010).

Interestingly, contiguous inteins can be artificially split, resulting either in intein fragments that require a renaturation step to become active (Mills et al., 1998; Southworth et al., 1998) or in intein fragments that are spontaneously active (Brenzel and Mootz, 2005; Brenzel et al., 2006). For synthetic biology purposes, split inteins are more interesting than contiguous ones, as not only they allow reactivation of a dysfunctional protein through reconstitution of the two inactive halves into a full-length, active molecule (which is what can be achieved when using a contiguous intein), but also to post-translationally modify a protein of interest fusing to it a functional tag such as a nuclear localization sequence (NLS), a degradation domain or a protein-protein interaction domain. Split inteins are also the basis of the so-called SICLOPPS technique for the creation of cyclic peptides or proteins (Tavassoli and Benkovic, 2007), which are valuable from a biotechnological point of view due to their higher resistance to enzymatic digestion and thermal inactivation (Williams et al., 2005; Waldhauer et al., 2015).

As synthetic biologists aim at controlling biological processes, a splicing reaction that spontaneously occurs at all times and in all places would be of limited use. If, on the other hand, one could specify when and where the splicing reaction should occur, a plethora of applications would become possible. The simplest way to gain control over inteins is by regulating their expression. Inducible promoters are good for this purpose: prior to the addition of the small molecule inducer, little intein is expressed thus the timing of splicing can be specified this way. When using split inteins, logic AND gates can be created by combining for instance two different promoters (Schaerli et al., 2014). Additionally, by selecting tissue-specific promoters, one can achieve spatial confinement of the splicing. However, recurring to the relative slow process of gene expression overall attenuates the gain of speed achieved by using the post-translational regulatory mechanism mediated by the inteins. Controlling the intein fragments at the protein level would be, therefore, a better option. Good news: methods to control inteins with diverse triggers such as small molecules, temperature and light are actually available (Figure 1 and Table 1). In this article, beyond briefly explaining what inteins are and their mechanism of action, we present the methods for achieving conditional protein splicing which were shown to work or have the potential to be applied in living cells.

Figure 1:

Existing methods to control inteins.

Schematic depiction of some of the methods currently available to control the splicing reaction mediated by inteins. Proteins are depicted as randomly curved lines for simplicity. (A) Methods based on photocaged amino acids. (B) Methods based on heterodimerization systems which respond to either light or small molecules. (C) Methods based on the LOV2 domain from Avena sativa. (D) Method based on proteases.

Methods based on pH, salt, divalent cations and reducing and oxidizing conditions are not discussed here, as they might serve as environmental sensors in a cellular setting but do not lend themselves to the external and general control of cellular processes. They have been reviewed elsewhere (Topilina and Mills, 2014; Wood and Camarero, 2014; Lennon and Belfort, 2017).

Methods to control contiguous inteins

The first method to control the splicing reaction dates back to 1995 (Cook et al., 1995). As serine (or cysteine) at position 1 of the intein is needed for the protein splicing pathway, the usage of a so-called photocaged serine – a serine with a light-removable protecting group (photolabile ‘caging’ group) attached to it – allowed inhibiting splicing by Tli pol-2 intein until UV light liberated the catalytic residue and activated the intein (Figure 1A). Importantly, UV light was applied for 10 min, which was not a problem only considering that the experiment was performed in vitro. While this work relied on ribosomal incorporation of an unnatural amino acid in an in vitro translation experiment, nowadays it is possible to incorporate such unnatural amino acids directly in living cells (Young and Schultz, 2018). Indeed, two recent studies undertook a similar approach, whereby in this case genetically encoded photocaged tyrosine and cysteine were incorporated in structurally and catalytically important positions of the Ssp DnaB and Npu DnaE inteins, respectively (Bocker et al., 2015; Ren et al., 2015). These two works proved the potential of the method for applications in living cells, for instance by activating the Src tyrosine kinase in mammalian cells using a 2 min irradiation time (Ren et al., 2015). Notably, photoremovable caging groups will not spontaneously leave in the absence of light, meaning that splicing activity in the dark state does not represent an issue here. The drawbacks of this method are the harmful nature of UV light, the necessity to do chemical synthesis and the need to transfect cells with constructs for the expression of the components required for the site-specific incorporation of the unnatural amino acid in the protein of interest.

An attractive alternative to the usage of chemically-modified amino acids is based on ligands that can be supplied to cells in the extracellular medium. In order to obtain responsiveness to the ligands, a directed evolution approach was used on the Mtu RecA intein into which the desired ligand-binding domain was incorporated (Buskirk et al., 2004; Skretas and Wood, 2005; Peck et al., 2011). The intein itself is interrupted by the ligand-binding domain – just like the intein interrupts the host protein – and this renders it splicing-incompetent. Upon ligand binding, a conformational change occurs that initiates protein splicing. Several rounds of mutagenesis and selection had to be made in order to find the appropriate mutants for which ligand binding triggers the needed conformational change. This method was applied in vivo (Yuen et al., 2006; Davis et al., 2015), on a variety of organisms. Interestingly, a version of Mtu RecA that was inhibited rather than activated by the ligand was also found (Skretas and Wood, 2005). Moreover, following the same approach used for Mtu RecA, the Sce VMA intein was made responsive to estrogen, in order to construct an estrogen biosensor in Escherichia coli cells (Liang et al., 2011). It is important to keep in mind, for methods based on small ligands, that these might have multiple functions inside the cells and activate alternative pathways that may render data interpretation difficult.

Finally, temperature-sensitive inteins have been identified or positively selected for after mutagenesis (Adam and Perler, 2002; Zeidler et al., 2004). However, with temperatures between 15° and 23°C permissive for splicing, these tools are restricted to cell types or organisms compatible with the necessary temperature shifts. Nonetheless, the power of this method has been demonstrated by regulating Notch activity with the temperature-sensitive Sce VMA intein in wing imaginal discs in Drosophila melanogaster (Zeidler et al., 2004).

Methods to control split inteins

The concept of photocaged residues can be applied to split inteins basically in the same way as to contiguous inteins. Taking advantage of their short size, some intein fragments can be obtained by total chemical synthesis. This was used to render the naturally split Ssp DnaE and the artificially split Ssp DnaB inteins light-sensitive by attaching a UV light-removable protecting group at a critical position within either Int^N or Int^C, respectively (Vila-Perello et al., 2008; Berrade et al., 2010; Binschik et al., 2011). Importantly, the method can be applied in complex biological fluids such as blood plasma where reconstitution of a coagulase led to the activation of prothrombin and thereby to blood coagulation after exposure to light (Binschik et al., 2011). In this case, the authors used a specifically induced cleavage reaction C-terminal to the intein, which is normally only a side-reaction of protein splicing. However, a great restriction of such chemically synthesized split intein fragments for cellular applications stems from the challenge to get these peptides across the cell membrane. This hurdle could be overcome by either using chemical or physical methods to deliver the synthetic intein fragment into the cells, such as commercial protein delivery reagents, cell permeabilization with streptolysin O and cell squeezing (Borra et al., 2012; Braner et al., 2016) or by recurring to cell penetrating peptides (Giriat and Muir, >2003; David et al., 2015).

Clearly advantageous from the synthetic biology perspective are intein tools that are fully genetically encoded. One such tool takes advantage of the fact that the two intein fragments must bind to each other to fold back into the active intein capable of catalyzing the splicing reaction. Here the two intein fragments obtained by artificially splitting Sce VMA, which fail to associate and reconstitute spontaneously (Ozawa et al., 2000), are brought in close proximity via rapamycin-induced heterodimerization of FKBP12 and FRB (Choi et al., 1996), each fused to one of the intein fragments (Mootz and Muir, 2002). Close proximity then allows the inteins to fold and activate splicing (Figure 1B). This approach, for which the term ‘conditional protein splicing (CPS)’ was actually coined, was applied in a variety of ways, for example in vitro for the control of an artificially autoinhibited protein kinase (Mootz et al., 2004), to activate tobacco etch virus (TEV) protease in yeast cells (Sonntag and Mootz, 2011), to induce apoptosis in mammalian cells via the RNAse α-sarcin (Alford et al., 2014) and to regulate firefly luciferase in living fruit flies, which showed detectable luciferase activity as early as 10 to 20 min after exposure to rapamycin-containing food (Schwartz et al., 2007) (see Table 1 for more examples of ligand-controlled Sce VMA intein).

Another beauty of the proximity-induced reconstitution of the Sce VMA intein fragments is its independence from the specific heterodimerization system. Indeed, beyond the small molecule rapamycin and its non-toxic derivatives, also red light has been used to control splicing of a model protein in yeast cells (Tyszkiewicz and Muir, 2008), because red (660 nm) light is responsible for the association of the two plant proteins PhyB and PIF3 (Quail, 2002) (Figure 1B). While theoretically a fully genetically encoded system employing the harmless and highly tissue-penetrating red light can be openly declared the best, the method based on PhyB-PIF3 suffers from some limitations, among which the most pressing is the requirement for an externally-supplied chromophore (phycocyanobilin or PCB, produced only in photosynthetic organisms). Furthermore, all methods based on the artificially split Sce VMA intein must deal with its limitations in terms of fragment solubility and foreign extein tolerance, which can result in lower splicing efficiencies. This is likely an inherent property of split intein fragments that do not spontaneously associate with each other, which is exactly what makes them amenable to the heterodimerization approach (Sonntag and Mootz, 2011).

To overcome having to externally supply the chromophore, an attractive alternative is to adopt the single-component, blue light photoreceptor AsLOV2 whose light sensing is achieved by a chromophore (flavin mononucleotide or FMN) that is naturally produced by any cell type. Indeed, two methods have been developed to control the very fast Npu DnaE intein (Iwai et al., 2006; Zettler et al., 2009) based on the second light-oxygen-voltage (LOV) domain from Avena sativa phototrophin 1 (AsLOV2) (Wong et al., 2015; Jones et al., 2016). Both harvest, albeit slightly differently, the conformational change that AsLOV2 undergoes upon absorption of blue light whereby its N- and C-terminal helices unfold and undock from the core LOV domain (Harper et al., 2003). In the dark, in its ‘closed’ conformation, AsLOV2 is supposed to interfere with the binding between Int^N and Int^C, while light should lead to a new conformation permissive of trans-splicing (Figure 1C). However, given the very high affinity between Npu DnaE Int^N and Int^C (Shah et al., 2011), dark activity remains an issue with these methods despite the attempts made to reduce it, namely truncation of Int^C that also reduced the splicing rate (Wong et al., 2015; Jones et al., 2016), and introduction of the C450M mutation to AsLOV2 (Wong et al., 2015), which moreover brings about the undesirable requirement of having to shine light for many hours onto the cells. Indeed, despite being less toxic than UV light, blue light is still to be administered with caution to cells to avoid photodamage. Moreover, its tissue penetration is rather poor rendering applications in intact animals on non-superficial areas very complicated.

Finally, split inteins have been conditioned to function after addition of proteases (Vila-Perello et al., 2008; Gramespacher et al., 2017). As mislocalized, overexpressed or hyperactive proteases are often a hallmark of diseases, having them trigger the splicing reaction has potential for the design of novel treatments based on intein-mediated protein reconstitution. For instance, a toxin that would kill the cell could be reconstituted using this conditional method, given that high activity of the protease would represent a disease condition and would justify the intervention.

The common idea is to use the protease to release a ‘factor’ introduced to inhibit trans-splicing, being it a protecting group on a chemically synthesized C-intein (Vila-Perello et al., 2008) or genetically-encoded cages made of truncated pieces of the complementary intein fragments fused to these latter ones (Shah et al., 2013; Gramespacher et al., 2017) (Figure 1D). The second approach was shown to be very generally applicable as it could regulate several split inteins, among which the fastest intein known to date, gp41-1 (Carvajal-Vallejos et al., 2012). Additionally, its strength for applications in living cells was demonstrated creating a sensor of viral infection in mammalian cells (Gramespacher et al., 2017). However, such an approach is not generally applicable as it is not always feasible to overexpress a protease in a healthy living cell without causing unwanted collateral effects.

Conclusions

It is tempting to claim that inteins evolved as a natural mechanism to regulate protein function, given that intein excision is required for reconstitution (=function) of the host protein. Indeed, at least few examples of naturally switchable inteins have emerged (Belfort, 2017). However, so far evidence for a well-defined regulatory role for inteins on their host proteins or for a link between presence of an intein and host organism fitness is missing (Frischkorn et al., 1998; Papavinasasundaram et al., 1998). Despite this, chemists, protein engineers and synthetic biologists are actively trying to create the perfect switchable intein that would allow controlling complex biological processes in space and time. We have discussed here several existing methods that have proven to be very useful and showed promising results. However, each of these tools has limitations mainly when combined with living cells, arguably explaining why they are not as wide-spread as it would be desirable. As a matter of fact, even when trying to use ‘just’ inteins – without any regulation of the splicing reaction – there are likely issues to solve. Indeed, inteins can display varying levels of optimization for their native host protein, in particular with regard to the so-called local exteins (typically a few amino acids immediately upstream and downstream the intein). Thus, splice sites need to be selected carefully and mutagenesis may be necessary. Splitting the target protein in two dysfunctional parts may also be associated with folding and solubility problems, which may be exacerbated when fusing them to poorly folding split intein fragments. Despite these hurdles, inteins are so powerful that the efforts invested in making them work are well justified.

We eagerly wait for the advent of new methods that fully meet the requirements on our wish list thus inviting cell and synthetic biologists to include inteins in their repertoire of regulatory tools for proteins alongside those for nucleic acids, such as CRISPR/Cas, RNAi and inducible gene expression.