Proximity Induced Splicing Utilizing Caged Split Inteins

Abstract

Naturally split inteins drive the ligation of separately expressed polypeptides through a process called protein trans splicing (PTS). The ability to control PTS – so-called conditional protein splicing (CPS) - has led to the development of powerful tools to modulate protein structure and function at the post-translational level. CPS applications that utilize proximity as a trigger are especially intriguing as they afford the possibility to activate proteins in both a temporal and spatially targeted manner. In this study, we present the first proximity triggered CPS method that utilizes a naturally split fast splicing intein, Npu. We show that this method is amenable to diverse proximity triggers and capable of reconstituting and locally activating the acetyltransferase p300 in mammalian cells. This technology opens up a range of possibilities for the use of proximity triggered CPS.

Graphical Abstract

Inteins, or intervening proteins, spontaneously remove themselves from the host protein in which they are embedded.¹ While this protein splicing reaction is performed in cis by the more common contiguous inteins, a subset of these autocatalytic proteins are expressed as two separate polypeptides and undergo protein trans splicing (PTS) upon association.² Because protein splicing results in a significant change in protein primary structure - the intein is removed and the flanking protein sequences (exteins) are joined through a native peptide bond – there has long been interest in harnessing the process to modulate protein function at the post-translational level.³ A key challenge in this area relates to the autocatalytic nature of the splicing reaction, which makes temporal control of activity difficult. To address this issue, a variety of conditional protein splicing (CPS) methods have been developed in which intein activity is activated only upon addition of a specified trigger. Generally, these triggers induce splicing by introducing either a splicing favorable conformational change in an otherwise structurally locked intein,^4–10 deprotecting chemically caged residues that inhibit catalytic activity,^11–18 or bringing intein fragments with low inherent affinity into close proximity, referred to herein as Proximity INduced Splicing (PINS).^{3, 18–24}

PINS methods are an especially attractive variation of CPS because they afford the possibility to sense and respond to myriad cellular stimuli; in principle, any process that results in the intein fragments being brought together should be capable of triggering splicing. Despite this potential, the PINS approach has been limited by its reliance on the artificially split VMA intein³, which suffers from poor stability²⁵ and stringent extein dependence.^26–28 In contrast, many naturally split inteins are stable over a wide range of conditions and have greater flexibility with regards to extein context.^29–32 For example, the fast splicing split DnaE intein from the cyanobacterium Nostoc Punctiforme (Npu) maintains activity up to 65°C and functions in 2M Guanidine or 6M Urea³¹. Furthermore, a rationally engineered promiscuous version of Npu has recently been reported that requires only the native +1 cysteine to effectively splice.³⁰ However, fast splicing inteins such as Npu have yet to be developed into a PINS method, as this would require overcoming the remarkably rapid and tight association of their intein fragments (Npu, K_d = 1.2 nM³³).

We recently reported the development of a CPS system based on inhibiting the association of a variety of naturally split inteins, including Npu.⁹ By extending a split intein fragment with the appropriate binding region from its complementary partner, it becomes locked in an intermediate folded structure (Figure S1). As a consequence, the caged intein fragments (NpuN^cage and NpuC^cage) cannot associate until the caging sequences are cleaved off by a specified protease. We wondered whether this split intein zymogen system could serve as a starting point for PINS. Specifically, we imagined that the caging sequences appended to each intein fragment likely dissociate and re-associate at a given rate. Thus, forcing the caged inteins together should increase the probability of a splicing productive binding event (involving a reorganization of the initial complex) between the full-length intein pairs (Figure 1a, S2).

Figure 1.

Proximity induced splicing (PINS) using caged split inteins. (a) Schematic depicting PINS of NpuN^Cage and NpuC^Cage through rapamycin induced dimerization of FRB and FKBP. (b-d) Coomassie stained SDS-PAGE gels of splicing reactions (37°, 1μM each protein construct) monitored over time with or without 10μM rapamycin. SP = splice product (MBP-eGFP, MW~ 69kDa). (b-c) CC = FRB-NpuC^Cage-eGFP. (b) NC = MBP-PINS-NpuN^Cage-FKBP. (c) NC = MBP-ePINS-NpuN^Cage-FKBP, Cas3 = caspase3 (cleaves the cages off the split inteins). (d) N = MBP-VmaN-FKBP, C = FRB-VmaC-eGFP. (e) Schematic depicting PINS-mediated FRET changes between Cy3 and Cy5 dyes. (f) Graph depicting observed FRET efficiencies in the presence or absence of rapamycin between FRB-NpuC^Cage-Cy5 and either Cy3-PINS-(C1A)-NpuN^Cage-FKBP or Cy3-ePINS-(C1A)-NpuN^Cage-FKBP. Errors = s.e.m. (n=3)

To test this idea, we fused the heterodimerization domains FKBP and FRB to the C- and N-terminus of NpuN^cage and NpuC^cage, respectively, allowing rapamycin to be used to force the caged inteins into close proximity (Figure 1a, S3). Initial results were encouraging in that rapamycin-dependent splicing was observed at 37°C, albeit at low efficiency (Figure 1b). We reasoned that the rate-limiting step in the overall process was likely isomerization of the initial rapamycin-induced heterodimer to afford an active intein complex. With this in mind, the PINS system was redesigned to favor this rearrangement. Informed by our previous work in which the split intein caging interface was optimized through an iterative design process⁹, we effectively retro-engineered the system to weaken slightly the interaction between NpuN and its caging sequence (Figure S4, S3). Gratifyingly, this exercise afforded an enhanced PINS-Npu system (ePINS-Npu) in which efficient rapamycin-dependent splicing was observed at 37°C over the course of a few hours, albeit slower than protease mediated splicing (Figure 1c). This represents a dramatic improvement over our previous PINS method based on the split VMA intein³, which while functional at lower temperatures, fails to undergo efficient PINS at the physiological relevant 37°C and, indeed, slowly precipitates out of solution (Figure 1d, Figure S5, S3).

Additional studies were conducted to better understand the relationship between splicing efficiency and the presumed structural reorganization within the rapamycin-induced complex. Specifically, PTS-based bioconjugation techniques were used to site-specifically install Cy3 or Cy5 dyes into the two components of the PINS system (Figure S3, S6). Importantly, a mutant version of NpuN was employed (NpuN-C1A) that cannot support PTS and that therefore afforded us the opportunity to follow the dimerization and subsequent isomerization process using FRET (Figure 1e). This biophysical experiment was conducted for both the original inefficient PINS-Npu system and the improved version thereof, ePINS-Npu. The complementary dye-labeled constructs were mixed and FRET efficiency monitored over time as a function of added rapamycin (Figure 1f). In both cases, we observed a rapid (i.e. within 1 minute, the earliest time-point taken) rapamycin-dependent increase in FRET efficiency, likely due to dimerization of FRB and FKBP. Thereafter, the behavior of the two systems diverged markedly; whereas FRET efficiency increased very slightly over the course of several hours for the original PINS-Npu system, the more active ePINS-Npu version exhibited a large increase in FRET over the course of the experiment. We interpret this second phase as reflecting the structural re-organization within the complex to give the active intein, which we note should bring the FRET pair into closer proximity based on their location. Consequently, the fluorescence data correlate well with the differing splicing activity of PINS-Npu and ePINS-Npu and support our contention that the overall kinetics of the PINS process are dictated by the efficiency of the internal isomerization within the complex.

An appealing feature of the PINS strategy is the potential to exchange the triggering mechanism^{3, 18–20}. In this regard, we were especially interested in the idea of using DNA as a substrate for PINS²⁰, particularly given the availability of programmable DNA binding proteins such as the nuclease dead version of CRISPR Cas9 (dCas9)³⁴. A model system was designed to explore this idea in which dCas9 from Streptococcus pyogenes was fused to the C-terminus of NpuN^cage while the NpuC^cage construct was fused to the N-terminus of histone H3 within a nucleosome (Figure 2a, S3). Proximity induced splicing, in this case triggered by a single stranded guide RNA (sgRNA) directed against a DNA sequence immediately upstream of the assembled nucleosome, should result in the splicing of an extein sequence (a myc epitope) to the histone (Figure 2a). Consistent with this design, we observed dose-dependent binding of the dCas9 fusion to the nucleosome in the presence of the on target gRNA, but not with an off target gRNA (Figure 2b). Importantly, sgRNA-dependent binding was accompanied by splicing to give the expected myc-tagged histone product (Figure 2c). The success of this proof-of-principle experiment illustrates the potential of using genomic targeting to mediate PINS and, more specifically, the possibility of carrying out localized protein tagging in a chromatin setting.

Figure 2.

PINS on a chromatin substrate. (a) Schematic depicting the dCas9-mediated PINS of an extein sequence (myc) on to the N-terminus of histone H3 within a nucleosome. sgRNA = single stranded guide RNA directing dCas9 to DNA adjacent to the nucleosome. (b) Gel shift assay in which increasing concentrations of myc-ePINS-NpuN^Cage-dCas9, 0.1μM NpuC^Cage-H3 nucleosomes (containing 5’-fluorescein tagged DNA) and on-target or off-target sgRNA were incubated for 16hrs. (c) Western blot against indicated epitope tags of reactions in which 0.5μM of myc-ePINS-NpuN^Cage-dCas9 and 0.1μM of NpuC^Cage-H3 nucleosomes were incubated with sgRNA that were on or off target for the nucleosome. NC = myc-ePINS-NpuN^Cage-dCas9, SP = splice product (myc-H3, MW~17kDa), CC-H3 = NpuC^Cage-H3.

Encouraged by the in vitro experiments, we next tested the functionality of the PINS strategy in mammalian cells. As part of this, we were also keen to take advantage of the minimal extein dependence of NpuC by reconstituting an active enzyme in a splicing-dependent manner. Previous research has shown that non-functional split fragments of the acetyl-transferase p300 can be re-ligated to restore activity using native chemical ligation (NCL)³⁵ and that a catalytically active p300 can be targeted to promoters via dCas9 to induce gene expression³⁶. Therefore, we wondered whether split p300 could be targeted via dCas9 to a user defined promoter and then reconstituted and activated utilizing Npu^cage in a PINS-dependent manner to induce transcriptional activation. Using the p300 core domain (amino acids 1,048–1664) which has been shown to be sufficient for catalytic activity³⁶, as well as the previously described split site³⁵ (Figure S7), we fused the C-terminal fragment of p300 (p300^C) to FRB-NpuC^cage and the N-terminal fragment of the enzyme (p300^N) to NpuN^cage-FKBP (Figure S8a, S9). Importantly, all residues flanking the intein were maintained as native p300 residues to ensure the absence of a splicing scar in the product (Figure S7, S9). Lastly, we appended a degradation tag to the C-terminus of p300^N-NpuN^cage-FKBP to help mitigate high protein concentrations and eliminate any risk of background splicing. As was hoped, co-expression of these PINS constructs in HEK 293T cells resulted in rapamycin-dependent generation of the p300 splice product based on western blotting (Figure S8b). To integrate this p300 activation system with genomic targeting, dCas9 was fused to the N-terminus of p300^N-NpuN^cage-FKBP (Figure 3a, S9). Transfection of this construct into 293T cells along with FRB-NpuC^cage-p300^C and a luciferase reporter plasmid led to a robust increase in luciferase activity in the presence of rapamycin and the appropriate sgRNA (Figure 3b). Importantly, luciferase expression was dependent on both protein splicing and FRB-FKBP dimerization as constructs that were either catalytically dead or missing FKBP, respectively, did not show any change in activity in the presence of rapamycin (Figure 3b). Notably, the corresponding constructs containing VmaN and VmaC in place of NpuN^cage and NpuC^cage displayed no luciferase activation when incubated with rapamycin (Figure 3b, S9), despite equitable expression (Figure S10).

Figure 3:

Rapamycin-mediated PINS in mammalian cells. (a) Schematic depicting constructs transfected into mammalian cells for PINS mediated Luciferase expression. (b) Graph depicting luciferase activity of HEK 293T cells transfected with indicated constructs in the presence or absence of 10μM rapamycin (+rapa) for 48hrs. Relative luciferase activation represents the fold-increase in signal observed for a sample in which 3 on-target sgRNAs are used relative to when a scrambled/off target sgRNA is used. N = dCas9-p300N-ePINS-NpuN^Cage-FKBP-Deg, C = FRB-NpuC^Cage-p300C, FKBP Del = N with FKBP deletion, C1A = N with a NpuN-C1A mutation, Vma = N and C where NpuN^Cage and NpuC^Cage are replaced with VmaN and VmaC respectively. Errors = s.e.m. (n=3)

It has long been appreciated that protein trans splicing provides unique opportunities for controlling protein function at the post-translational level¹. Progress in this area has, however, been constrained by the inability to control the activity of naturally split inteins using the induced proximity paradigm. Herein, we provide a solution to this problem by exploiting recently developed caged versions of these proteins. Our PINS strategy overcomes many of the limitations of previous tools in this area. Indeed, by replacing the unstable split VMA intein by the more robust naturally split Npu intein, we were able to achieve temporal control over the key transcriptional co-activator, p300, in mammalian cells. This application also illustrates the ability to integrate our PINS strategy with other cellular control elements, in the present case genomic targeting using dCas9. While the current PINS method relies on a caged version of Npu, it should be noted that similarly caged versions of several other naturally split inteins are available⁹. Since these split inteins are all functionally orthogonal to each other and to Npu, it follows that integrating these into our PINS manifold will extend the scope of the technology, including offering the intriguing possibility of multiplexing-type applications. We imagine that the work described herein will serve as a template for such studies.