Optimisation of Tet-On inducible systems for Sleeping Beauty-based chimeric antigen receptor (CAR) applications

Regulated expression of genetic elements that either encode polypeptides or various types of functional RNA is a fundamental goal for gene therapy. Inducible expression may be preferred over constitutive promoters to allow clinician-based control of gene expression. Existing Tet-On systems represent one of the tightest rheostats for control of gene expression in mammals. However, basal expression in absence of tetracycline compromises the widespread application of Tet-controlled systems in gene therapy. We demonstrate that the order of P2A-linked genes of interest was critical for maximal response and tightness of a chimeric antigen receptor (CAR)-based construct. The introduction of G72V mutation in the activation region of the TetR component of the rtTA further improved the fold response. Although the G72V mutation resulted in a removal of a cryptic splice site within rtTA, additional removal of this splice site led to only a modest improvement in the fold-response. Selective removal of key promoter elements (namely the BRE, TATA box, DPE and the four predicted Inr) confirmed the suitability of the minimal CMV promoter and its downstream sequences for supporting inducible expression. The results demonstrate marked improvement of the rtTA based Tet-On system in Sleeping Beauty for applications such as CAR T cell therapy.

Introduction of a G72V mutation in rtTA-M2 enhances the tightness of the Tet-On system. Roney et al. 16 reported that a GGG to GTG (G72V) missense mutation in rtTA mitigated basal gene expression in the absence of an inducer in S. cerevisiae clones. Because of the similarity of transcriptional machinery amongst eukaryote cells, we reasoned that this approach may give similar results in human cells. The G72V mutation was next introduced into pSBtet-2 and pSBtet-3 to create the pSBtet-4 and pSBtet-5 constructs (Fig. 1). The G72V mutation in rtTA-M2 decreased the background expression of TCE promoter in the absence of doxycycline at both the mRNA and protein level (P < 0.001, Fig. 3A,B). The G72V mutation also restored the maximal expression of pSBtet-2 and pSBtet-3 constructs following induction with doxycycline (Fig. 3C). As previously reported, G72V-rtTA-M2 appeared less sensitive to doxycycline compared to original rtTA-M2, though this was not statistically significant ( Fig. 3C, P > 0.9) 16 . A similar pattern of results was obtained after two weeks passage of cells to ensure stable integration of the pSBtet-5 (Fig. 3D,E). Note, the GFP expression of the transfected cell lines dropped from ~ 90 to ~ 70% after two weeks of culture, most likely due to a shift from transient gene expression to that from integrated cassettes. www.nature.com/scientificreports/ Investigation of autoregulatory strategy with G72V variant. We next investigated possible improvements in inducibility of multiple GOIs using positive feedback control in an autoregulatory cassette. Autoregulation can improve tetracycline-regulation in a retroviral vector 18 and in a bi-directional 19 or uni-directional lentiviral vector 7 . The bi-directional approach appears tight in transient transfection, however, high background was detected when cells were stably transduced 19 . Therefore, we utilised a uni-directional strategy with a P2A sequence in place of an Internal Ribosome Entry Site (IRES) sequence, to allow expression of GOI-P2A-G72VrtTA under TCE promoter (pSBtet-6, Fig. 1). We speculated that the leaky expression of the TCE would still allow sufficient levels of G72V-rtTA inducer to respond to doxycycline stimulation. Although the positive feedback system resulted in tight expression at the protein level (Fig. 4A), as previously reported 7,18,19 , the system was leaky at the mRNA level for Mcl-1 (Fig. 4B). Basal expression of luciferase in pSBtet-6 was lower than for pSBtet (P < 0.05) and showed a higher response upon induction (P < 0.05, Fig. 4C). Compared to the constitutive expression of G72V-rtTA, the autoregulatory system showed greater sensitivity to doxycycline induction (P < 0.05; Fig. 4C), although the basal expression was higher (P = 0.0066).
Removing cryptic alternative splice sites within rtTA reduces the background expression. Since the first description of the eukaryotic Tet-On inducible system, most optimisation studies have focused on rtTA mutations: for example, the removal of cryptic splice sites (flanking amino acids 8-144) in the TetR sequence 20,21 . Using recently-developed software 22 , we identified eight additional potential cryptic splice sites within the coding region of rtTA-M2 (Fig. 5A and Table S1). Because the G72V mutation resulted in the loss of a cryptic splice site at position 215 (Fig. 5A and Table S2, we determined if the success of the G72V mutation was due to the removal of the potential cryptic splice site at 215. These cryptic splice sites are located in two regions of rtTA; one in a surface residue (215 nt and seven in the dimerisation domains (320 nt, 326 nt, 367 nt, 392 nt, 408 nt, 456 nt and 541 nt; Fig. 5A, Figure S1). We therefore removed all eight cryptic splices sites by silent or conserved missense-mutations in the pSBtet construct ( Table 1). The removal of six cryptic splice sites modestly enhanced the tightness of Tet-On system 7.7-19.6 fold compared with original rtTA (P < 0.001, Fig. 5B). The remaining two mutations at position 320 (~ twofold, P = 0.8) and 367 (~ fivefold, P = 0.08) did not significantly affect Tet-On performance. The mutation at position 320 produced E106Q, while 367 (Q122) was a silent mutation (Table 1). It is possible these two splice sites are weak 5′ acceptor splice sites which are only used if other competing splice sites are removed 23 . Indeed, positions 320 and 367 have low score and confidence which represent strength and the probable occurrence of a splice site, respectively (Table S1). Combining all mutations together, improved the leaky background of the Tet-On system ~ 40 fold compared to original rtTA (P < 0.0001, Fig. 5B). However, superior results were still seen with G72V mutation (Fig. 5C).
Surprisingly, combining the G72V mutation and removing all cryptic splice sites abolished responsiveness and inducibility of Tet-On system (Fig. 5C,D). There are four altered amino acid positions within rtTA-M2 that result in a reverse activator phenotype, as compared to the original TetR: E71K, D95N, L101S and G102D 2 . In TetR, E71 is a surface residue amino acid, D95 connects the DNA reading head to the core domain, while L101 and G102 are crucial for dimerisation and the tetracycline response, respectively 2 . In TetR the E71 and G72 amino acids create the turn between α-helix-4 and 5 (Fig. 6). This region bridges the DNA binding domain to the tetracycline binding domain and the combination of both the E71K and G72V mutations might destroy the structure of this critical turn, causing a loss of rtTA-M2 activity. This may also explain the drop in tetracycline-induction observed with the position 320 mutant (E107 to Q107, Fig. 5D), since this residue is close to a 'high sensitivity www.nature.com/scientificreports/ region' ( Figure S2) 21,[24][25][26] . It is possible that the G72V mutation affected the secondary structure of rtTA 16 , rather than simply removing a cryptic splice site. It is interesting to note that further commercial algorithm-mediated codon optimisation of rt-TA-M2 attempted in pSBtet-1 (see Fig. 2) re-introduced 13 cryptic alternative splice sites with high score and confidence (Fig. 5A, Table S3) within the rtTA coding region. This may have contributed to the poor performance of the first pSBtet-1 construct analysed, since cryptic splice sites might be associated with poor performance of Tet-On systems 20,21 ( Fig. 2A,B).

Dissection of the tce proximal promoter. Modification of the minimal CMV promoter can affect TCE
promoter performance 27 . Removing elements downstream of the TATA box can reduce the maximal expression, whereas deleting the upstream elements can decrease the leakiness 27 . We therefore revisited the design of pTIGHT to ensure optimal performance in our setting. Core promoter elements were identified using the YAAP program (Fig. 7A). It is possible that the presence of alternative initiator element (Inr) might lead to a loss of control of the TCE-promoter. We therefore removed these elements in single or combinatorial mutation fashion from pSBtet and monitored the tightness and maximal expression of TCE promoter. Removal of each, or all, downstream elements of the TATA box (Inr-1, -2, -3 and -4 sites, and the DPE element) markedly decreased the tightness of the TCE promoter (Fig. 7B), and also reduced the maximal expression, as previously reported 27 (Fig. 7C). Specifically, removing Inr-3 increased the background expression remarkably www.nature.com/scientificreports/ (P < 0.01). Inr-3 contains two CTF/NF1 binding sites that bind to DNA as dimers 28 (Fig. 7A). CTF/NF1 is an enhancer-blocker element which specifically blocks the interaction of other enhancers with the promoter 29 . A general explanation would be that deletion of the CTF/NF1 binding site could increase the interaction of neighbouring enhancers to the TCE promoter, resulting in high background. The only core promoter element found upstream of the TATA box is B recognition element (BREu). As shown in Fig. 7D,E, removing BREu did not improve the tightness and the basal expression of the TCE promoter (P > 0. 19) and optimal transcription through the TCE promoter was dependent on the TATA box. However, deleting the BREu site increased the response of the TCE promoter to doxycycline (Fig. 7E, P < 0.01).

Discussion
Several strategies have been proposed to reduce the leakiness and enhance the inducibility of Tet-On systems, with only some tested in human cells. Such approaches include: (1) increased expression of rtTA using a strong promoter and codon optimisation 21,30,31 , (2) mutation of rtTA to increase binding to doxycycline or DNA 16,20,21,26 , (3) autoregulatory systems 7,18,19,32 , (4) removing a cryptic splice sites in the rtTA coding region 20 , and (5) alteration of the core promoter elements within the proximal region of the TCE promoter 27 . We revisited these strategies for use in the SB-based Tet-On system in a human cell line for future investigation in CAR T cell therapy.
The introduction of a single mutation G72V, gave the optimal induction results at both mRNA and protein level, as reported in S. cerevisiae 16 . Future studies may explore the use of a G72P instead of G72V in our system as a candidate amino acid at position G72, though G72P appeared to result in a small loss in sensitivity, as compared to G72V 16 . It is interesting that independent efforts into the rtTA structure have resulted in distinct amino acid changes in different studies, but with similar outcomes. For example, mutations introduced into the rtTA-M2 gene used here are present in distinct positions, as compared to the original four mutations in rtTA 6,20 . Moreover, introducing sensitivity enhancing (SE) mutations [24][25][26] (V9I, F67S, G72P, F86Y, and R171K) could further increase the sensitivity to doxycycline, without increasing the background, as demonstrated in yeast 16 .
Autoregulatory systems have recently generated interest, with both the rtTA and GOI transcribed by a single TCE promoter, using either a bi-directional promoter 19 or an IRES sequence 7,18,32 . However, our constitutive expression of G72V-rtTA gave tighter expression, but was less sensitive to doxycycline compared to the autoregulatory system. The autoregulatory system may be preferred for controllable expression of a toxic rtTA or toxic GOI in mammalian protein production [32][33][34][35] .
Next, our analysis found evidence of cryptic splice sites within an rtTA, a sequence that was previously optimised for mammalian expression by Urlinger et al. 30 Removing these splice sites reduced the basal expression and further increased the maximal expression. Unfortunately, using the combination of the G72V mutation with all splice sites removed (using predominantly silent, but with two necessarily non-silent, mutations) created a non-responsive system. It appears likely that the combination of the E71Q and G72V mutations disrupted the turn between two critical α-helixes 4 and 5.
It is also noteworthy that different programs identified other possible splice sites that needed to be investigated (Table S4). At least three independent approaches for codon optimisation of rtTA have been reported to enhance Tet-On function 21,30,31 . For example, Urlinger et al. modified the S. cerevisiae-developed rtTA-M2 sequence to remove potential hairpin, splice, and endonuclease sites, as well as codon optimising the sequence for use within mammalian systems 21 . www.nature.com/scientificreports/ In the proximal region of the TCE promoter, we confirmed that the TATA box was essential for the function of the TCE promoter 27 . Removal of the BREu element increased maximal expression, but did not markedly affect the tightness of the TCE. The deletion of the BRE site might enhance the elongation and / or reduce the TFIIB-rtTA sequestration. BRE plays a role in the preinitiation complex (PIC), leading to the dissociation of TFIIB from the promoter which is necessary for RNA polymerase II to initiate the elongation step [36][37][38] . Hence, interrupting the BRE-TFIIB interaction may enhance transcription via the enhancement of elongation 39 . Alternatively, direct sequestration VP16 on TFIIB has been reported 40,41 that may act to reduce VP16-mediated transcriptional activation.
Collectively, our results demonstrate marked improvements to the rtTA-M2 based Tet-On system in a Sleeping Beauty system through the yeast-optimised G72V mutation. The results especially highlight the necessity to investigate the placement of individual GOI and rtTA within an expression cassette. The use of the clinically relevant CAR cassette within this setting offer the possibility to enhance adoptive cell therapy though druginducible expression of cell-survival and memory enhancing genes, or death switches to conditionally ablate CAR T cells following the onset of cytokine release syndrome.

Material and methods
plasmid construction and cloning. The Tet-On SB (pSBtet-GP) contains the tetracycline-inducible pTIGHT promoter upstream of two asymmetric SfiI sites for cloning genes of interest (GOI), with a downstream RPBSA promoter driving GFP-P2A-rtTA-P2A-puromycin. pSBtet-GP and the SB-transposase vector (pCMV(CAT)T7-SB100) were purchased from Addgene. The pTIGHT promoter is a derivative of (Ptet-14) with shorter spacer (16-17 bp) sequences 27 between the TRE and the minimal CMV promoter (see Fig. 7) 9,27,42 . To generate the modified SB plasmids, a multiple cloning site (MCS) with Bsu36I and BstBI sites was cloned into pSBtet to remove GFP-P2A-rtTA-P2A-puromycin to create pSBtet-MCS. The codon optimised rtTA-M2 gene and FRP5 scFv Her2-CAR 43 were synthesised as gene blocks (IDT Singapore) and cloned into pSBtet MCS to create pSBtet-1. Other plasmids were generated by splicing by overlap extension (SOE) PCR to fuse the original rtTA2S-M2 (rtTA) 9,21 , GFP and Her2CAR in different combinations as illustrated in Fig. 1. Mutations into rtTA was introduced using inverse or SOEing PCR. Codon optimised-mouse Mcl-1 (Cop-Mcl-1) was synthetised as a gene block (IDT) with SfiI overhangs to replace the Firefly luciferase gene in the pSB-tet constructs. Both the Mcl-1 and firefly luciferase genes were used as GOI in this study. To modify the core promoters elements (Fig. 5A), the proximal promoter of TCE was PCR amplified from pSBtet and then subcloned into a pUC19 vector (Addgene) using conventional restriction fragment ligation method with EcoRI and NcoI enzymes. Inverse PCR with primers carrying point mutations were used to change the core promoter elements. Finally, each of the modified fragments were PCR amplified from pUC19 and cloned back to pSBtet using PshAI and NcoI restriction sites. To alter the cryptic splice sites within rtTA (Table 1), rtTA was sub-cloned into PUC19 and mutations introduced using inverse PCR.
Bioinformatics analysis. Analysis of the TCE proximal promoter for core promoter elements, including the initiation repeats (Inr1, 2, 3 and 4), TATA box, B recognition element (BRE) site and downstream promoter element (DPE), was carried out using YAPP Eukaryotic Core promoter predictor. TF binding sites were predicted using AliBaba 2.1 44 and PROMO 45,46 programs. The transcriptional start site (TSS) was predicted as reported previously for the minimal CMV promoter 47 . Screening of rtTA for cryptic splice sites was carried out using Alternative Splice Site Predictor (ASSP) software and Human Splicing Finder (HSF) 22,48 . The protein structure of TetR and the prediction of secondary structure were obtained from Protein Data Bank (PBD).