Flexible TALEs for an expanded use in gene activation, virulence and scaffold engineering

Bacterial growth conditions

Escherichia coli strain Top10 (New England Biolabs) was grown at 37°C using LB media. Agrobacterium tumefaciens strain GV3101 was cultivated at 28°C using YEB media, while Xoo strains BAI3 and BAI3ΔtalC and their derivatives were grown at 28°C using PSA media.

Plant growth conditions

Nicotiana benthamiana plants were grown under greenhouse conditions with 16 h of light, relative humidity of 40–60% and temperatures of 23 and 19°C, during daytime and night, respectively. After growing for 4–6 weeks, plants were used for inoculation experiments. Oryza sativa plants from the cultivars Zhenshan, Kitaake, Sadu Cho and IR24 were grown under greenhouse conditions with a relative humidity of 70% and temperatures of 28 and 25°C, during daytime and night, respectively. After 3–5 weeks of growth, plants were used for inoculation experiments.

Identifying TALE repeats of aberrant length

In order to identify novel aberrant repeat length variants, all available TALE-carrying Xanthomonas genomes as well as all available individual TALE sequences were analysed. For this, the AnnoTALE ‘TALE prediction’ and ‘TALE analysis’ functions (42) were combined with a manual evaluation process. All Xanthomonas repeats with <33 or >35 aa were classified as aberrant repeats.

Construction of novel aberrant repeat modules and artificial TALEs

In order to construct novel aberrant repeat modules, the required amino acid changes were introduced into existing repeat modules of standard length via PCR. Oligonucleotides are listed in Supplementary Table S1. Artificial TALEs were generated using the Golden TAL Technology assembly kit (43). In the first step, up to six single repeats were combined into one of the available assembly vectors. In the second assembly step, several of these multi-repeat modules were combined with the N- and C-terminal regions of a TALE and the desired expression vector to obtain the final TALE expression constructs. The N- and C-terminal regions were derived either from Hax3 (44) or from TalAG4 and TalAO3 from Xoo strain PXO83 (5). For the A. tumefaciens-mediated inoculation of N. benthamiana and the Xanthomonas-mediated inoculation of rice, the expression vectors pSKA2 and pSKX1 were used, respectively.

Reporter constructs and GUS analysis

To generate GUS reporter constructs, PCR-amplified promoter fragments from rice or artificial TALE target boxes in combination with the Bs4 minimal promoter were placed in front of a promoterless uidA reporter gene by inserting them into the vector pGWB3GG, a Golden Gate-compatible variant of pGWB3. Oligonucleotides are listed in Supplementary Table S1.

Nicotiana benthamiana infiltration and GUS reporter assay

Agrobacterium tumefaciens strains containing the reporter- and TALE-expressing constructs were resuspended in Agrobacterium infiltration media and adjusted to an OD₆₀₀ of 0.8. For the inoculation of N. benthamiana plants, one reporter- and one TALE-expressing construct were mixed in a 1:1 ratio and then infiltrated using a needleless syringe. Two days after inoculation, two leaf discs (diameter 0.9 cm) were harvested from each inoculation spot and frozen via liquid nitrogen. The samples were stored at −80°C until further use. All GUS analyses were performed as previously described (23). Experiments were repeated at least twice.

Oryza sativa virulence assay

The two youngest leaves of 3–5-week-old plants were inoculated with X. oryzae strains BAI3 and BAI3ΔtalC or one of their derivatives using a needleless syringe. Prior to the inoculation, the bacterial suspensions were adjusted to an OD₆₀₀ of 0.5 using 10 mM MgCl₂. Each strain was inoculated on three different rice plants, using two leaves per plant by placing three to four spots within the first 5 cm of each leaf tip. Five days after inoculation, leaves were harvested and symptoms were documented.

Prediction of TALE target genes

TalBK2 target genes were predicted using the ‘Predict and Intersect Targets’ tool of AnnoTALE version 1.4.1 (42,45). For this analysis, promoter sequences from 300 bp upstream to 200 bp downstream of the transcriptional start site or the start codon were utilized. The promoter sequences were extracted from the O. sativa genome MSU7.

Naturally occurring aberrant repeats

To understand the variety of natural length polymorphisms in TALE repeats, we analysed all available TALE sequences from Xanthomonas spp. All repeats that deviate in length from standard repeats (34 aa) or from commonly occurring length variants (33 and 35 aa) were classified as aberrant repeats. Within 276 completely sequenced TALE-containing Xanthomonas genomes and multiple individual TALE sequences, we identified 195 repeats of aberrant length in 180 TALEs with 77 unique RVD sequences. One hundred sixty-five TALEs carry a single and 15 TALEs carry two aberrant repeats (Supplementary Figure S1; Supplementary Tables S2 and S3). Sixty-seven TALEs share small N- and C-terminal deletions that mark them as part of the group of truncated TALEs (truncTALEs)/interfering TALEs (iTALEs) (40,41), while four other TALEs harbour larger deletions and are annotated as putative pseudogenes. Most TALEs with aberrant repeats are from Asian Xoo or Xoc strains with four exceptions identified in X. axonopodis pv. citri and one in the X. theicola strain CFBP4691. Every completely sequenced Asian X. oryzae strain carries at least one TALE with an aberrant repeat (Supplementary Figures S1 and S2). According to the AnnoTALE nomenclature (42), the aberrant repeat-carrying TALEs are categorized into 24 different TALE classes (Supplementary Figures S1 and S2; Supplementary Table S2). The 195 aberrant repeats themselves can be categorized into 10 different length variants with 42, 40, 37 (versions 1, 2 and 3), 36 (versions 1, 2 and 3), 30 and 28 aa, respectively (Figure 1A; Supplementary Tables S2 and S3). For repeat variants with 42, 40 and 30 aa, it was shown earlier that they mediate a flexible recognition of a perfectly matching sequence and a −1 nucleotide frameshift sequence (14), while an obligatory frameshift recognition has been described for the 28 aa variant (41). In contrast, none of the other variants have been analysed so far.

Not all aberrant repeats facilitate a flexible frameshift recognition

To analyse the impact of the novel aberrant repeats on the DNA-binding mode of TALEs, we introduced a single aberrant repeat of each type into an artificial TALE. To build artificial TALEs, all aberrant repeat variants were constructed as modules (Supplementary Table S4) for our Golden TAL Technology assembly kit (43). The TALEs contained 17.5 repeats with an alternating DNA-binding specificity, to maximize frameshift impact. As control, a TALE containing only 34 aa standard repeats was used. We also included two different TALE repeats with 35 aa since they were never analysed for their ability to confer frameshift tolerance. All aberrant repeats were placed at the same position (position 7) and contained the same RVD (HD) (Figure 1A–D). The TALEs were tested in combination with two different target TALE boxes in a GUS reporter assay. One TALE box was the optimal sequence based on the RVDs (op), and the other one was a frameshift derivative with a single nucleotide deleted at the position corresponding to the aberrant repeat (−1 p7). If a TALE successfully binds to the target box, the GUS reporter gene is induced and activity can be detected.

The control TALE with only 34 aa repeats can activate transcription at the optimal box but is unable to address the frameshift derivative (Figure 1E). Similar observations were made for the two repeat variants with 35 aa, the variant 36 aa v2 and the variant 37 aa v1, making the latter two the first natural aberrant repeats unable to confer a frameshift recognition (Figure 1E). In contrast, the TALEs containing one of the aberrant repeats with the known ability to confer frameshift recognition (30, 40 and 42 aa) and the variant 36 aa v1 are able to flexibly address both boxes (Figure 1E). The TALE containing the 28 aa repeat shows a clear and reproducible preference for the frameshift box while still being able to address the optimal box. This observation partially contradicts earlier reports in which the 28 aa variant conferred obligatory frameshift binding (41). The TALE containing the newly identified aberrant repeat 37 aa v2, however, showed activity only on the frameshift box, rendering it the only natural repeat variant that loops out constitutively (Figure 1E). The last novel aberrant repeat, the variant 36 aa v3, is based on a repeat with 35 aa. While it contains only a single amino acid duplication, this duplication, an additional G, is placed right after the RVD, and thus at a position with crucial impact on repeat functionality. As a consequence, the activity of a TALE with even a single 36 aa v3 repeat is barely detectable (Figure 1E).

We can thus categorize aberrant repeats also based on their functionality. The first category does not affect TALE binding, the second one facilitates a facultative frameshift recognition, the third is completely unable to participate in DNA recognition, likely by looping out of the repeat array, while the fourth and last category impairs overall TALE function.

The novel aberrant repeats themselves loop out

TALEs with an aberrant repeat of 42, 40 or 30 aa show the highest activity on a TALE box with either no deletion or a deletion exactly corresponding to the position of the aberrant repeat (14). This leads to the conclusion that the aberrant repeats loop out themselves when challenged with a frameshift sequence, instead of forcing out one of the adjacent repeats. Based on these data, we wanted to clarify how the novel identified aberrant repeats with 37, 36 or 28 aa behave. To answer this question, we tested our TALEs with the various aberrant repeats at position 7, as well as our control TALE with exclusively 34 aa repeats, on TALE boxes with a one-nucleotide deletion at position 6, 7 or 8 (Supplementary Figure S3A–D). All TALEs containing an aberrant repeat that is able to facilitate a facultative frameshift recognition showed highest activity either on an optimal box or on a box with a −1 nucleotide deletion corresponding exactly to the aberrant repeat (Supplementary Figure S3F and G). Furthermore, the TALE containing the variant 37 aa v2 could only activate the box with a −1 deletion at position 7, supporting our previous observation that this repeat variant loops out constitutively. In contrast, TALEs with 37 aa v1, 36 aa v2 or 35 aa behaved exactly like a TALE solely constructed with standard 34 aa repeats and recognized only the optimal sequence (Supplementary Figure S3F and G). These results show that all aberrant repeats that confer recognition of a frameshift sequence prefer to loop out of the repeat array themselves.

We further investigated whether aberrant repeats can mediate binding to a −2 nucleotide frameshift sequence, for example by causing disturbances in the stabilizing inter- and intra-repeat interactions so that one of the adjacent normal repeats is forced out together with them. For this, we used target boxes containing a −2 nucleotide deletion at the positions 6 and 7 or 7 and 8 (Supplementary Figure S3E) and tested them together with TALEs harbouring one of the natural aberrant repeats at position 7. In contrast to a 34 aa standard repeat, all aberrant repeats with the ability to confer a −1 nucleotide frameshift recognition were also able to allow for the recognition of a −2 frameshift box; however, the observed activity was extremely low (Supplementary Figure S3H and I). Nonetheless, this demonstrates that aberrant repeats do have a weakening effect on the inter- and intra-repeat interactions of a TALE and that this effect can lead to multiple repeats being excluded from the DNA binding simultaneously. This observation is the first indication that a TALE is able to loop out multiple repeats to fit to a target sequence (Supplementary Figure S3J).

Multiple aberrant repeats in tandem can loop out at once

Based on the observation that an artificial TALE with two 40 aa aberrant repeats in tandem showed activity in our previous study (14), we wondered whether it is possible to create a functional TALE with multiple aberrant repeats. We therefore generated TALEs with zero to six of the 40 aa aberrant repeats placed in tandem. Those TALEs were tested on an optimal box and on target sequences that had one to six nucleotides deleted corresponding to the positions of the aberrant repeats (Figure 2A and B). We wanted to limit the number of necessary boxes to a minimum, while on the other hand gain full readout on how many aberrant repeats are tolerated and how many repeats can loop out. Therefore, we designed the centre of the repeat region (repeats 7–12) containing only NN repeats, which can bind to any of the Gs in the target sequence, while placing six repeats with mixed RVDs as ‘specific anchors’ in the beginning and the end of the TALEs to enforce an exact positioning of both TALE ends (Figure 2A and B). With this set-up, we were able to test all TALEs on the same set of reporter boxes. This experiment demonstrated that a TALE can in principle tolerate multiple aberrant repeats (Figure 2C). In our set-up, a TALE with as many as four aberrant repeats in tandem still showed full activity (Figure 2D), while the activity of TALEs with more aberrant repeats dropped significantly (Figure 2C). Since TALEs with multiple aberrant repeats yielded only high activities on boxes with several nucleotides deleted, we conclude that multiple aberrant repeats in tandem tend to loop out simultaneously (Figure 2C and D). This offers further insight into the structure and function of TALEs, indicating that aberrant repeats weaken the repeat-to-repeat interaction such that a looping out is the preferred conformation. At the same time, it appears that at least four repeats looping out at once is tolerated without penalty for the overall activity, possibly, because the remaining repeats have enough binding affinity to keep the TALE connected to the DNA target sequence.

Neighbourhood community: the nature of an aberrant repeat controls its impact on either flanking repeats

By comparing TALEs containing two or three aberrant repeats in tandem of either 40 or 42 aa length, we observed that they do not activate the same reporter constructs (Supplementary Figure S4). Based on this observation, we generated TALEs containing each aberrant repeat placed at positions 7 (HD) and 8 (NN) (Supplementary Figure S5). We used for each of the two aberrant repeats RVDs with a different nucleotide specificity to be able to distinguish which of the two aberrant repeats participates in DNA binding and which loops out. We tested the TALEs on four different target boxes, with no deletion (op), the nucleotide deleted that corresponds to the first aberrant repeat (−1 p7), the nucleotide deleted that corresponds to the second aberrant repeat (−1 p8) or both nucleotides deleted (−2). For TALEs with two 28 and 37 aa v2 repeats, respectively, we saw exclusive activity on the −2 box, showing that both always loop out simultaneously (Supplementary Figure S5). In contrast, the TALE with the two 30 aa repeats in tandem activated only the optimal and the −1 p8 box, indicating that the first aberrant repeat cannot loop out. The 42, 40 and 36 aa v1 repeats showed activity on −1 and −2 boxes (Supplementary Figure S5). However, which of the two different −1 boxes was activated depended on the aberrant repeat variant present. The TALEs with two 40 or two 36 aa v1 repeats looped out the second or both aberrant repeats, whereas the TALE with two 42 aa repeats looped out the first or both aberrant repeats (Supplementary Figure S5). This suggests that the particular alteration in an aberrant repeat affects its interaction with the repeat either preceding or following the aberrant repeat, and its looping-out behaviour, accordingly. The duplication in the 42 aa repeat variant is placed within the first helix of the repeat, suggesting a weakened interaction with the preceding repeat. In contrast, the aberrant repeat with 40 aa has the duplication in the second helix, suggesting a compromised interaction with the following repeat. Based on these observations, we wondered whether the presence of both duplications in a single repeat would weaken the interactions with both adjacent repeats and result in a repeat that constantly leaves the repeat array. We placed such a synthetic 48 aa aberrant repeat at position 7 (HD) within a TALE with 17.5 repeats and tested the TALE for its activity on an optimal and a −1 frameshift box (Supplementary Figure S6). Indeed, the TALE containing the synthetic 48 aa variant showed only activity on a −1 frameshift box, indicating a strong bias of this repeat variant to loop out of the repeat array (Supplementary Figure S6). This supports a model in which the alteration in an aberrant repeat specifically affects the inter-repeat interactions with either one of the adjacent repeats.

All out: two aberrant repeats only function independently at a distance

The flexible looping-out behaviour of individual aberrant repeats raises the question how the distance between two aberrant repeats influences their behaviour and the overall activity of a TALE. For these experiments, we used the same TALE set-up with 17.5 repeats shown in Figure 2, and placed a 40 aa aberrant repeat at position 7 and another one at position 8, 9, 10, 11 or 12 (Figure 3A). The TALEs were tested together with a control TALE without an aberrant repeat on the series of target boxes with one to six nucleotides deleted (Figure 3B).

All tested TALEs were functional, but differences in their behaviour were observed. Two aberrant repeats placed in direct vicinity preferentially looped out one or both aberrant repeats (Figure 3C). If one or two standard repeats were placed between two aberrant repeats, the respective TALEs were highly active on a wide range of boxes, indicating that both aberrant repeats can stay in (op), one of them can loop out separately (−1) or both of them can loop out simultaneously (−2) (Figure 3C and D). Strikingly, high activity was also observed on the −3 and −4 boxes, which can only be explained by a TALE conformation with three or four repeats not partaking in DNA binding. We hypothesize that two aberrant repeats drag out one or more standard repeats between them while they themselves leave the repeat array, resulting in an extraordinary flexibility of the TALE (Figure 3D). This shows that under specific circumstances also a standard 34 aa repeat can leave the repeat array and not partake in DNA binding. The likelihood to loop out more than two repeats dropped drastically, if three or more standard repeats were placed between the two aberrant repeats (Figure 3C). For two aberrant repeats to show an independent flexible looping-out behaviour, it is therefore advisable to position them with at least three, better four normal repeats in between. We further analysed whether a combination of 42 and 40 aa aberrant repeats together would enhance the cooperative excision of aberrant repeats and normal ones sandwiched between them by weakening the interaction with both outward neighbouring repeats. This was not the case, although the preferences of the TALEs for specific target boxes changed slightly in comparison to ones with two 40 aa repeats (Supplementary Figure S7). This demonstrates that the nature of the two aberrant repeats influences the overall binding behaviour of the respective TALEs.

Synthetic networks: TALEs for multi-readout expression of logic genetic gates

The independent flexibility of multiple aberrant repeats within one TALE prompted us to test the combinatorial possibilities as synthetic biology transcriptional switches. A series of TALEs with zero to two 40 aa aberrant repeats and an overlapping specificity was designed to match different combinations of highly related target sequences. The TALEs contained 17.5 repeats of alternating specificity and a single aberrant repeat at either position 4 or 12, or two aberrant repeats at positions 4 and 12. In addition, the repeat at position 4 or 12 was omitted (Supplementary Figure S8A–C). We tested the TALEs on a set of boxes lacking the base at either position 4 or 12, or both of them (Supplementary Figure S8A–C). The TALEs showed activity only on their optimal box or on boxes where the nucleotide deletions were placed corresponding to the positions of the aberrant repeats present in the respective TALE (Supplementary Figure S8D). This allowed activation of individual boxes, as well as combinations of two or four boxes. With careful fine-tuning and keeping in mind the requirement for a reasonable number of strong RVDs, it is now possible to address a set of different boxes individually and in multiple possible combinations. By using such a set of TALEs as master switches and inserting the different target boxes as control elements into target promoters, this set-up can be used as a powerful tool for the regulation of genetic circuits (e.g. artificial biosynthetic pathways) with multi-readout expression (Supplementary Figure S8E).

Natural flexibility: members of class TalBK contain two aberrant repeats and bind to multiple target sequences

Only two examples of natural TALEs carrying multiple aberrant repeats exist, classes TalHT and TalBK (Supplementary Figure S1). Class TalHT is a truncTALE/iTALE and unable to bind DNA. Class TalBK members were shown to induce the well-known susceptibility gene OsSWEET13 (19,46). All members of class TalBK share the aberrant repeat variant 36 aa v1 at positions 9 and 12 with the RVDs HD and NS, respectively (Supplementary Figure S9A). The general RVD set-up of these TALEs is extremely similar; however, TalBK1 and TalBK15 are six repeats shorter than the other members, while TalBK1, TalBK3, TalBK4, TalBK11, TalBK12 and TalBK14 contain one different RVD at position 8 (HD instead of NI) (Supplementary Figure S9A). To test the DNA-binding specificity of class TalBK, we constructed three artificial TALEs using the RVDs of TalBK1, TalBK2 and TalBK15, named artBK1, artBK2 and artBK15, respectively (Figure 4A; Supplementary Figure S9B–D). An optimal target box was generated based on the RVDs of TalBK2. Additionally, four different frameshift derivatives of this box were generated, carrying a one-nucleotide deletion at position 9 or 12, a two-nucleotide deletion at positions 9 and 12, or a four-nucleotide deletion from positions 9 to 12. GUS reporter studies in planta showed that TALE artBK2 is highly active on the optimal box, as well as on frameshift boxes with nucleotide deletions at either one or both positions corresponding to the aberrant repeats (Figure 4A and B). Furthermore, this TALE showed activity on the box with the four-nucleotide deletion, a box that can only be addressed by looping out the two aberrant repeats together with the two standard repeats between them (Figure 4A and B). ArtBK15, on the other hand, showed only low activity, while artBK1 displayed no activity at all (Supplementary Figure S9E). The lower activity of artBK15 in comparison to artBK2 is likely due to the reduced number of repeats, which results in a lower DNA-binding affinity, while artBK1 is further impaired by a mismatch at position 8 in our set-up. In nature (Xoo KACC10331), TalBK1 has an additional frameshift mutation in its C-terminal region, rendering it incapable of gene activation and possibly a pseudogene (Supplementary Figure S9B). We cannot conclude whether TalBK3, TalBK4, TalBK11, TalBK12 and TalBK14 would be able to induce gene activity on the optimal box. While they carry the same RVD change that renders artBK1 unable to induce the optimal box, they are six repeats longer, and thus should be less impaired by a single mismatch than their shorter class member.

An arms race: TalBK and TalAM activate different OsSWEET13 promoter variants

We used AnnoTALE v1.4.1 (42,45) to screen the promoterome of the rice cultivar Nipponbare for potential target sites of TalBK2. When we manually removed either one of the two aberrant repeats from the TALE DNA sequence to simulate its loop out, OsSWEET13 ranked at the top position with a perfectly matching target site (Supplementary Figure S10), which is consistent with previous reports (19,46). The same position in the OsSWEET13 promoter is also targeted by TalAM2 (PthXo2; Figure 4E) (6,14), which does not contain any aberrant repeat, and it coincides with known INDEL variations in the OsSWEET13 promoter (6,14). Such promoter variations function as natural resistances by prohibiting binding of TALEs and thereby blocking activation of key virulence genes (6,17,18). We aimed to understand the interaction of the different OsSWEET13 promoter variants and Xoo TALEs in the context of resistance. The OsSWEET13 promoter of several indica and japonica rice cultivars was sequenced and one new variant from cultivar Sadu Cho was identified in addition to the known ones from Nipponbare, Zhenshan and IR24. These promoters differ by a number of adenosines in a region that corresponds to the first quarter of the TalBK target site (Figure 4C). The target site in Sadu Cho resembles the one from the cv. Nipponbare, but it carries an additional A-to-T exchange at the position of the first repeat of TalBK. To test how well these promoters can be induced by the Xoo TALEs, 1-kb promoter fragments were cloned in front of the GUS reporter gene and tested in combination with artificially assembled versions (artBK1/2/15, artAM2) of the natural TALEs TalBK and TalAM2 in N. benthamiana. ArtBK2 was only able to activate the promoter from Nipponbare, whereas none of the other variants was recognized (Figure 4D). ArtBK1, on the other hand, showed no activity at all, while artBK15 induced the promoter fragment from Nipponbare, however, to a lower extend than artBK2 (Supplementary Figure S9F–H). ArtAM2 induced activity at the OsSWEET13 promoter variants from cultivars Zhenshan and IR24, but not Nipponbare or Sadu Cho (Figure 4C and D). The binding possibilities and resulting mismatches of the TALEs to the different promoter sequences are shown in Supplementary Figures S11–S14. Our results confirm OsSWEET13 from the cv. Nipponbare as natural target of TalBK2. They also indicate that the one additional mismatch at the critical position 1 in Sadu Cho is sufficient to block artBK2 recognition (Supplementary Figure S11). To demonstrate the potential of a flexible TALE, we designed a TALE to recognize all four promoter varieties (termed flexTALE). For this, we placed two aberrant repeats at positions 3 and 6, and combined those with an unspecific NS RVD at position 1 (Figure 4C and D; Supplementary Figure S14B). This flexTALE was active on all promoter variants (Figure 4D), demonstrating that TALEs containing aberrant repeats can recognize multiple allelic sequences.

To analyse the activity of TalBK2 in a more natural context, we expressed it in Xoo strain BAI3ΔtalC. This strain is an avirulent derivative of the highly virulent Xoo strain BAI3 that lacks the major virulence factor TalC (TalBS1). Its pathogenicity can be restored by the introduction of an OsSWEET-inducing TALE (11,12). The TalBK2-expressing variant of this strain was inoculated into rice plants from the four rice cultivars Zhenshan, Nipponbare, Sadu Cho and IR24. Disease symptoms developed only in plants from the cv. Nipponbare, while a TalAM2-expressing BAI3ΔtalC strain caused symptom formation on plants from the cv. Zhenshan and IR24, thus mirroring the results previously obtained in our GUS assay (Supplementary Figure S15). At the same time, the avirulent strain BAI3ΔtalC did not result in any disease development, while the BAI3 wild-type strain, due to its TalC-dependent induction of OsSWEET14, caused disease phenotypes on all four cultivars.

Taken together, our results support the finding that OsSWEET13 is the natural target of some if not all TalBK members (19,46). The binding of these TALEs to the promoter is most likely achieved by looping out either one of the two aberrant repeats since only these two confirmations result in a situation where the TALE encounters a perfect matching binding sequence in the OsSWEET13 promoter from the rice cv. Nipponbare.

A fragile balance: how a TALE repeat becomes aberrant

We observed that not all natural occurring aberrant repeats can confer frameshift binding. This indicates that the position of the modification within the repeat and its amino acid composition control the impact of aberrant repeats on the interaction of TALEs with DNA. To investigate this further, we first focused on 36 aa repeat variants. The natural 36 aa v1 repeat (duplicated positions 27–28) loops out facultatively and allows standard as well as frameshift binding, whereas the 36 aa v2 repeat (duplicated positions 32–33) is fixed within the array in a normal recognition mode (Figure 1). By introducing a 2 aa duplication at several other positions of a TALE repeat, we generated four additional artificial 36 aa repeat variants (Supplementary Figure S16). The aberrant repeats (RVD NN) were placed at position 8 of a TALE with 16.5 repeats and the resulting TALEs were tested for their ability to bind to an optimal or a frameshift box (Supplementary Figure S16). Two variants (art1, art4) were fully flexible aberrant repeats, showing similar behaviour as the natural variant 36 aa v1. The two other variants (art2, art3), however, rendered the TALE completely inactive on any of the target sequences. In these variants, the duplication was inserted relatively close to the RVD loop. This demonstrates once more that even a single repeat with an unfavourable amino acid composition is able to generally block the activity of a TALE.

Linker scanning: amino acid composition and position of repeat insertions control their behaviour

To study how the amino acid composition of an aberrant repeat influences its behaviour, we created variants of the two natural 40 and 42 aa aberrant repeats. We replaced each of the duplicated parts present in the natural variants by a stretch of serines and glycines (SG) of identical length (Supplementary Figure S17). Chains of serines and glycines are often used as unstructured linkers to connect protein domains. These artificial aberrant repeats were placed at position 8 in a TALE with 16.5 repeats and tested for activity on an optimal and a −1 frameshift box. The two TALEs with artificial 40 aa variants show opposite behaviour. While the variant with the first half of the duplication replaced strongly prefers the −1 frameshift box, the variant with the second half replaced prefers the optimal box (Supplementary Figure S17). If the SG linker replaced the first half of the duplication in the variant with 42 aa, the TALE showed activity on both target boxes. In contrast, if the second half was replaced, no activity at all was detected (Supplementary Figure S17). These striking differences between the natural repeats and artificial variants show that the amino acid composition of an aberrant repeat controls its impact on TALE activity.

Next, we wanted to see how the position of an alteration within an aberrant repeat affects its behaviour. We therefore introduced a 7 aa linker sequence (SSGGGGS) at different positions throughout a TALE repeat (Figure 5A) and inserted those repeats into a TALE at position 7 (Figure 5B and C). We then tested those TALEs on an optimal box (op) and a −1 nucleotide frameshift box with a deletion at position 7 (−1 p7). We noticed that the positions close to the end of a TALE repeat (positions 26–34) tolerate the insertion of the 7 aa linker without any change in the binding behaviour, leading to activity only on the optimal box (Figure 5E–G). Insertions at positions 4–5 within the first α-helix (α1) or positions 19–24 within the second α-helix (α2) of the repeat, on the other hand, resulted in a compulsory loop out of the repeat (Figure 5E–G). The only repeat variant that conferred flexible binding to both the optimal and the frameshift box was the one with an insertion at position 25. If the linker was placed in the region around the RVD loop (positions 9–18), it resulted in a complete or near-complete loss of TALE function (Figure 5E–G). Western blot analysis demonstrated that this loss of function is not attributed to loss of protein expression or stability (Figure 5H). A second experiment confirmed this pattern and showed that the insertion of even a single amino acid (serine, S) into or around the RVD loop of a TALE repeat is sufficient to impair the overall activity of a TALE drastically (Supplementary Figure S18). These results clearly show that amino acid insertions into the regular structure of TALE repeats have a differential impact on their behaviour. Disturbances of the regions flanking the RVD loop destroy overall function of the TALE, whereas insertions into either one of the helical regions disturb repeat-to-repeat interactions and favour looping out of the aberrant repeat. Finally, insertions near the loop region between repeats (e.g. the end of a repeat) are tolerated without impact on the DNA-binding behaviour of the TALE.

Functional fusions: TALEs tolerate insertions of large domains between repeats

Because the insertion of a 7 aa linker close to the end of a repeat was well tolerated, we focused on placing inserts between two repeats. First, we inserted two serine/glycine linkers of 7 and 18 aa, respectively, between repeats 7 and 8. The resulting TALEs showed activity only on the optimal box (Supplementary Figure S19A–C). Next, we tested whether it is possible to insert larger protein domains between two repeats. We combined a 238 aa long GFP protein with a 7 aa SG linker on each end and placed it between the repeats at positions 7 and 8 (Figure 6). To determine whether the TA–GFP–LE fusion protein is still able to bind to a specific target sequence, a GUS assay on an optimal box (op), on a −1 frameshift box (−1 p8) and on a +1 frameshift box was performed. The TA–GFP–LE resulted in a comparable activity to the control TALE with standard repeats on the optimal box (Supplementary Figure S19D-F). Both had no activity on a +1 box, indicating that the large size of the GFP protein does not impose a gap upon the standard repeat-to-repeat assembly of the regular TALE repeat structure, but likely loops out of the array, similar to an aberrant repeat. We also observed a low but reproducible activity for a TALE carrying a GFP insertion in the repeat array on a −1 frameshift box, indicating that sometimes one of the standard repeats adjacent to the GFP insertion loops out as well (Supplementary Figure S19D–F). The GFP insertion could be placed at different positions throughout the repeat array (position 2, 8 or 14) with similar results (Figure 6). However, a TALE with two or more GFPs inserted at different positions showed strongly reduced activity, which suggests that the protein cannot interact properly with target DNA sequences (Figure 6). To clarify whether all parts of the TA–GFP–LE are bound to the DNA, we designed three different mismatch boxes, harbouring three mismatches at position #3–5, #10–12 or #14–16 (Supplementary Figure S19G). This analysis confirmed that indeed all parts of the TA–GFP–LE are bound to DNA (Supplementary Figure S19H). Laser scanning microscopy showed that the GFP domain placed within the repeat array is properly folded and fluorescing (Figure 6C). These data show that it is possible to insert large functional domains in between individual repeats of a TALE without interfering with its specific DNA recognition in vivo. This opens exciting avenues to expand the TALE scaffold in the third dimension for new biotechnology applications.