Multiple gene expression in plants using MIDAS‐P, a versatile type II restriction‐based modular expression vector

Abstract MIDAS‐P is a plant expression vector with blue/white screening for iterative cloning of multiple, tandemly arranged transcription units (TUs). We have used the MIDAS‐P system to investigate the expression of up to five genes encoding three anti‐HIV proteins and the reporter gene DsRed in Nicotiana benthamiana plants. The anti‐HIV cocktail was made up of a broadly neutralizing monoclonal antibody (VRC01), a lectin (Griffithsin), and a single‐chain camelid nanobody (J3‐VHH). Constructs containing different combinations of 3, 4, or 5 TUs encoding different components of the anti‐HIV cocktail were assembled. Messenger RNA (mRNA) levels of the genes of interest decreased beyond two TUs. Coexpression of the RNA silencing suppressor P19 dramatically increased the overall mRNA and protein expression levels of each component. The position of individual TUs in 3 TU constructs did not affect mRNA or protein expression levels. However, their expression dropped to non‐detectable levels in constructs with four or more TUs each containing the same promoter and terminator elements, with the exception of DsRed at the first or last position in 5 TU constructs. This drop was alleviated by co‐expression of P19. In short, the MIDAS‐P system is suitable for the simultaneous expression of multiple proteins in one construct.

To date, a common strategy for transient expression of multiple recombinant proteins in plants is co-transformation with individual Agrobacterium strains each harboring a vector coding for a single gene of interest (Giritch et al., 2006;Roy et al., 2010). This approach can be problematic as it is not possible to ensure that the transformed plant cells simultaneously receive all of the constructs.
There are a number of reports of multiple recombinant genes expressed in tandem from a single expression vector using transient expression (Sarrion-Perdigones et al., 2013;Teh et al., 2014;van Dolleweerd et al., 2014) (Chen et al., 1998) by particle bombardment. For the insertion of multiple genes arranged in tandem in a single vector, binary bacterial artificial chromosomes (BIBAC) (Hamilton et al., 1996) and yeast artificial chromosomes (YAC) (Mullen et al., 1998) have been used. The highest number of linked genes arranged in tandem and expressed in planta using a single vector is 10 genes, with a single 33.6 kb T-DNA, utilizing Cre/loxP site-specific recombination and a transformationcompetent artificial chromosome (TAC) based vector (Lin et al., 2003).
However, classical cloning of linked transgenes in one vector to create large T-DNAs can be an overly complex process, as the finite number of available restriction enzymes becomes a limiting factor.
In recent years, a plethora of different strategies allowing the assembly of different "bioparts" such as promoters, terminators, and transcription factors have been reported (Engler et al., 2008;Knight, 2003;Rebatchouk et al., 1996;Shetty et al., 2011). More recently, a modular idempotent DNA assembly system (MIDAS) was reported (van Dolleweerd et al., 2018). This is a hierarchical cloning assembly toolkit based on the Golden Gate use of type IIS restriction enzymes to generate non-palindromic overhangs that ligate upon addition of a ligase in a "one-pot" reaction. This allows the assembly of genes from basic, reusable parts and the assembly of plasmids containing multiple genes. Using the MIDAS system, the group was able to successfully assemble seven genes from 21 modules in a single plasmid and demonstrate expression in Penicillium paxilli (van Dolleweerd et al., 2018).
We have generated a simple, non-hierarchical version of MIDAS, named Modular Idempotent DNA Assembly System for plants (MIDAS-P), to investigate expression of multiple genes in planta.
The TUs are first prefabricated in entry vectors, followed by a strategy using type IIS restriction sites and alternating blue/white screening that can arrange these multiple TUs in a binary destination vector. We used MIDAS-P to assemble plasmids designed to express a cocktail of anti-HIV biologics as a test case. The cocktail of anti-HIV therapeutics include (1) VRC01, a potent and broadly neutralizing antibody that has been shown to neutralize 91% of HIV-1 isolates by targeting the CD4 binding site of the virus (Wu et al., 2010); (2) the lectin Griffithsin (GRFT), which targets the HIV envelope glycans and has subnanomolar activity against CXCR4-and CCR5-tropic strains of  , GRFT (Hoelscher et al., 2018), J3-VHH (McCoy et al., 2012) with a 6xhis-tag followed by a C-terminal KDELtag, and the codon optimized gene from Discosoma sp. fluorescent protein FP583 R2G mutant (DsRed) (AF168419) with a six amino acid (SATGSA) chloroplast-targeting N-terminal transit peptide from potato starch granule-bound starch synthase (GBSS) were amplified using relevant templates and primers (Table S1). The genes were domesticated to be free from NcoI, XbaI, BsaI, and BsmBI sites. The exception is the DsRed where NcoI was present. The P19 gene silencing suppressor of Tomato bushy stunt virus (ACV49953.1) was amplified from pEAQ-HT-DEST3 (Sainsbury et al., 2009) and the internal BsaI site was removed using Quickchange II mutagenesis kit (GGTCTC to GCTCTC) as per manufacturer's instructions (Agilent, USA). The 5ʹ and 3ʹ ends of the sequences were flanked with NcoI and XbaI restriction sites, respectively, and cloned into either pWHITE or pBLUE depending on their assembly position in the destination vector pMIDAS (Table 1). For DsRed, the forward primer had a BsaI followed by an NcoI restriction site to generate compatible cohesive ends to the NcoI site in pWHITE or pBLUE. Therefore, the DsRed PCR products were digested with BsaI and XbaI (BsaI site was removed after digestion). Ligated constructs were transformed into E.
The entry vector pWHITE harboring the first transcription unit (TU) was ligated into the pMIDAS destination vector using BsaI.
Briefly, 100 ng of entry vector were mixed with 100 ng of destination vector together with 10 units of BsaI (NEB), 400 units of T4 DNA ligase (NEB, USA) and 1x T4 DNA ligase buffer. The mixture was cycled 50 times between 37°C for 2 min and 16°C for 5 min before ending with 37°C for 5 min. The mixture was then transformed into Escherichia coli DH10B (Thermo Fisher) and spread onto LB selection plates containing 50 µg/ml carbenicillin (Apollo Scientific), 1 mM IPTG, and 20ug/ml X-Gal. White colonies were selected.
In the next step, the second TU, which is cloned in the pBLUE entry vector, was assembled in the same way into the destination vector pMIDAS + TU1 using BsmBI, transformed into DH10B and plated onto selection plates. Blue colonies were selected. The third, fourth, and fifth TUs were consecutively ligated into the destination vectors using BsaI (for pWHITE TUs) or BsmBI (for pBLUE TUs). The constructs were then transformed into Agrobacterium tumefaciens GV3101::pMP90(RK) and plated onto Yeastextract mannitol (YM) medium (0.04% w/v Yeast extract, 1% w/v Mannitol, 1.7 nM NaCl, 0.8 mM MgSO4, 2.2 nM K 2 HPO 4 , pH7) with 100 µg/ml rifampicin, 50 µg/ml kanamycin, 50 µg/ml gentamycin, and 50 µg/ml carbenicillin.  (Strasser et al., 2008) were either transformed using syringe-mediated infiltration or vacuum infiltration as described by Kapila et al. (1997). The plants were then further grown in containment at 25°C with a 16/8-h light/dark cycle. Leaves were harvested at 6 days postinfiltration (dpi). Plant crude extract was obtained by grinding the leaves using pestle and mortar, or 3 mm chrome steel ball bearings and a Mixer Mill MM400 (Retsch) with 3 ml 1xphosphate-buffered saline (PBS) (2.7 mM KCl, 8 mM Na 2 H-PO 4 , 137 mM NaCl, 2 mM KH 2 PO 4 ) per 1 g leaf fresh weight. Total soluble protein (TSP) of the crude extract was measured at A 280 with a Nanodrop 2000 (Thermo Fisher).
Purified plant-made DsRed protein (gift from Fraunhofer IME) was used as a standard and/or positive control. Leaves infiltrated with Agrobacterium harboring pMIDAS empty vector were used as a negative control. Readings were carried out at 590 nm using the Infinite F200 Pro plate reader (TECAN). Each point was measured in triplicate.

| Genomic DNA extractions and insert amplification
Genomic DNA was extracted from leaf samples using DNeasy plant mini kit (QIAGEN) according to the manufacturer's instructions. PCR was carried out using HF Phusion master mix (NEB) with PCR cycle according to manufacturer guidelines using gene-specific primers detailed in Table S1. Negative controls were leaves infiltrated with pMIDAS alone and positive controls were leaves infiltrated with pMIDAS harboring VRC01 HC + LC, GRFT, J3-VHH, or DsRed depending on the experiment.

| Taqman qPCR
RNA was extracted using RNeasy plant mini kit according to the manufacturer's instructions and treated with DNase using RNase free DNase set (all QIAGEN). Following DNase treatment, RNA was quantified with a NanoDrop 2000 (Thermo Scientific) and 200 ng of RNA were used to synthesis cDNA using the LunaScript RT supermix kit (NEB). The synthesized cDNA was diluted 1:50 and 5 µl were used for a qPCR reaction with the addition of 15 µl mixture of GoTaq probe qPCR master mix (Promega), target specific forward primers (250 nM), reverse primers (250 nM), and 6FAM/BHQ1 internal Taqman probes (900 nM). The ribosomal protein L25 was used as a reference gene and qPCR thermocycle protocol was carried out according to qPCR master mix manufacturer's instructions; reactions were performed using the CFX-connect real time PCR detection system (Biorad). All primers were designed using Primerplus and were purchased from Sigma. Primers used for Taqman qPCR are summarized in Table S2.

| Statistical analysis
Normality of all data was tested and null hypothesis rejected if p < 0.05 using the Shapiro-Wilk test; if found to be normal, a one-way ANOVA test was carried out. If data was significantly skewed, a non-parametric Kruskal-Wallis test was used for data analysis. The homogeneity of variance was also tested for each sample data set using the Levene test and the null hypothesis was rejected if p < 0.05 resulting in the analysis of a data set either with the Brown-Forsythe and Welch correction or ordinary one-way ANOVA test depending on Levene test outcome. Post hoc analysis of the different sample groups was carried out either using the Dunn Bonferroni or the Tamhane T2 post hoc multiple comparison test depending on the outcome of the Levene test. All graphs were drawn and analyzed using the GraphPad Prism 8 software (GraphPad, USA).

| Design of MIDAS-P entry and destination vectors
The MIDAS-P assembly system for expression of multiple genes in plants comprises two entry vectors, pWHITE and pBLUE, which contain transcription units (TUs) based on the pTRAk system (Sack, 2007), and a destination expression vector, pMIDAS for accepting the TUs containing the genes of interest from the entry vectors ( Figure 1).
The entry vectors pWHITE and pBLUE each contain an identical cassette which comprises the Scaffold Attachment Region (SAR) of the tobacco Rb7 gene, a cauliflower mosaic virus F I G U R E 1 Schematic representation of the MIDAS-P assembly system for plant expression. The system consists of two entry vectors, pWHITE and pBLUE, for cloning genes of interest and alternate sub-cloning in the binary destination (expression) vector pMIDAS. The first transcriptional unit is constructed in pWHITE and transferred into pMIDAS using the type IIS restriction enzyme BsaI. A second transcriptional unit in pBLUE can subsequently be transferred into pMIDAS using BsmBI. Further TUs can be added by alternating transfer from pWHITE and pBLUE. The inclusion of lacZα in pMIDAS and pBLUE allows blue/white screening at each stage. The destination vector pMIDAS also has right and left T-DNA borders for Agrobacterium-mediated plant transformation. GOI, gene of interest; P, promoter; pA, terminator and polyA signals; SAR, scaffold attachment region; UTR, untranslated region PINNEH ET AL. | 1663 (CaMV) 35S promoter with duplicated enhancer, the 5ʹ untranslated region (UTR) of tobacco etch virus (TEV), gene of interest (GOI) cloning sites (NcoI/XbaI) and a CaMV 35S polyadenylation site/terminator (Figure 1). pBLUE has an additional lacZ gene for blue/white selection during the cloning process. The pWHITE and pBLUE vectors further differ in their type II restriction sites used for transfer of the TU into the multigene cassette assembly in the destination vector. In pWHITE, the TU is flanked by BsaI; in pBLUE, the TU is flanked by BsmBI.
The GOI is cloned in pWHITE or pBLUE based simply on the order in which the GOIs are assembled into the destination vector -pWHITE is used for genes going into odd-numbered positions in the final multi-gene cassette, while pBLUE is used for evennumber positions (Figure 1).
The assembly of the multiple genes in the destination vector pMIDAS crucially depends on the configuration of BsaI and BsmBI type IIS restriction sites in pWHITE and pBLUE, respectively. TUs assembled in pWHITE can be used for cloning into destination vectors using a BsaI-mediated one-pot Golden Gate assembly reaction, which introduces BsmBI sites that are used for addition of the next TU. In turn, cloning from pBLUE introduces new BsaI sites, allowing cloning again from pWHITE. This cycle of cloning can be repeated indefinitely, and each plasmid generated by cloning a TU into the multigene construct becomes the destination vector for the next cycle of TU addition. Following each cloning cycle, positive clones can be identified by blue or white colony screening.
The plasmids in the Agrobacterium were all intact before transformation into the plants, and there were no detectable recombinationmediated rearrangements or deletions ( Figure S1).
At 6 dpi, there were no significant differences in DsRed expression level for 1-and 2 TU constructs (Figure 2b). There was a slight drop in the mean expression level when DsRed was expressed as a 3 TU construct but this was not statistically significant. When expressed as a 4 TU construct, DsRed expression dropped to barely detectable levels. Surprisingly, when expressed as a 5 TU construct, DsRed expression was unexpectedly restored to levels comparable to when DsRed was expressed as a 3 TU construct or when co- F I G U R E 3 Positional effects with TU permutations on protein expression of MIDAS-P constructs. Representative western blots of extracts from leaves infiltrated with different constructs harboring permutations of 3 TUs (a) and of 4-5 TUs (b) at 6 dpi. HC-LC-GRFT-J3His is shown as a representative example of 4 TU constructs. Leaves infiltrated with pMIDAS were used as negative controls (−ve control). Positive controls (+ve control) were leaves infiltrated with either a 2 TU construct containing VRC01 HC + LC, or a 1 TU construct containing GRFT or 6x histidine tagged J3-VHH (J3His). (c) DsRed expression levels in 4 or 5 TU constructs. Leaves infiltrated with pMIDAS only were used as negative controls (Control). Expression levels were quantified using a DsRed standard and box plots represent the mean, minimum and maximum of six biological repeats. Data were analyzed using Brown-Forsythe and Welch ANOVA tests with Tamhane T2 multiple comparison test (***p < 0.001 and ****p < 0.0001). ANOVA, analysis of variance; TUs, transcription units PINNEH ET AL.

| 1665
The presence of constructs harboring 4 or 5 TUs in the plant leaves 6 days after Agrobacterium infiltration was confirmed, even though protein expression was not detected. DNA was isolated from infiltrated leaves and the target sequences were amplified with gene-specific primers (Table S1). Leaves infiltrated with vector only and purified plasmids containing the single GOI were used as negative and positive controls, respectively. All genes of interest in the multi-gene constructs were detected in DNA isolated from infiltrated leaves ( Figure S3). However, we were not able to distinguish between T-DNA that were transferred into plant cells and the plasmids still contained in the Agrobacterium using this method.

| The silencing suppressor P19 can rescue protein expression
We showed that mRNA levels of the genes of interest are substantially reduced when 3 or more TUs are assembled in the multi-gene vector regardless of TU permutation. However, this experiment did not show whether reduced mRNA levels were due to the mRNA not being transcribed or due to post-transcription gene silencing. Therefore, the RNA silencing suppressor P19 (Voinnet et al., 2003) under control of the CaMV 35S promotor in pMIDAS was co-infiltrated with the multigene construct harboring 3 TUs for the expression of GRFT, VRC01 LC, and HC.
p19 mRNA was detected when the pMIDAS-P19 construct was expressed on its own, and co-expressed with VRC01 in separate or the same constructs ( Figure S4). Co-infiltration of p19 with the 3 TU construct as separate vectors caused a significant increase in (p < 0.0001) mRNA levels of HC, LC, and GRFT (Figure 5a). This translated to a 2-3-fold expression level increase for both VRC01 and GRFT (Figure 5b). When p19 was expressed as the first TU in the same vector as VRC01 (HC-LC; Figure S5a), there were no significant differences in VRC01 expression level between P19-HC-LC and HC-LC-GRFT. However, the leaves were comparatively necrotic suggesting a hypersensitive response (HR). When P19 was expressed as the third TU, there were significant differences in VRC01 expression levels between HC-LC-P19 and HC-LC-GRFT, so much so it approached levels of the 2TU VRC01 construct ( Figure S5a). On the other hand, when p19 was expressed as the fourth TU together with VRC01 and GRFT, there was only a slight increase in expression level (9 ± 6 mg/kg compared with no expression in VRC01-GRFT-J3; Figure S5b).
We also investigated whether P19 co-expression could improve the yield of VRC01, GRFT and J3-VHH expressed as 5 TU constructs, which were at levels undetectable by Western blot (Figure 3). P19 did increase the yield of all the components of the 5 TU construct

| DISCUSSION
In this study, we have created MIDAS-P, a DNA assembly system with blue and white bacterial colony screening, for quick and easy cloning of multiple, tandemly arranged genes for expression in plants.  Table S2. Data represent the mean of n = 3 biological repeats done in triplicates ± SD. Data were analyzed using Brown-Forsythe and Welch ANOVA tests followed by Tamhane T2 multiple comparison test (***p < 0.001 and ****p < 0.0001). (b) Protein expression levels of VRC01 and GRFT, 6x His-tagged J3-VHH (J3His) and DsRed expressed using 3 TU and (c) 5 TU constructs with or without P19 co-expression. VRC01, GRFT, J3-VHH, and DsRed expression levels were quantified by ELISA using a human IgGk standard (Sigma-Aldrich), GRFT standard (gift from Evangelia Vamvaka and Paul Christou), plant-made J3-VHH and DsRed standard (gift from Johannes Buyel) respectively. Box plot for expression levels represent the median, minimum, and maximum of three biological repeats. ANOVA, analysis of variance; ELISA, enzyme-linked immunosorbent assay; mRNA, messenger RNA; TUs, transcription units; SD, standard deviation example of a disease target for which low-cost cocktails of biologics (e.g., for therapy, prophylaxis, or vaccination) are likely to be needed, due to the likelihood of viral escape. We deliberately selected anti-HIV biologics of different classes (an antibody, a lectin, and a nanobody) that were unlikely to interact with each other negatively in planta. Vamvaka et al. (2018) have shown that extracts of transgenic rice expressing three anti-HIV products (2G12 antibody, GRFT, and another lectin Cyanovirin-N) using separate constructs had synergistic HIV-1 neutralization capabilities.

Expression
When testing the capabilities of the MIDAS-P system, we found a tendency for reduced protein expression levels for individual proteins as the number of TUs increased. When VRC01 HC, LC, and GRFT were expressed as 3 TU constructs, mRNA levels and the corresponding protein expression decreased, although protein expression levels of up to approximately 2.5 g/kg for GRFT and approximately 0.4 g/kg for VRC01 were still achieved. Although position-dependent differences were observed in mRNA expression, this did not impact VRC01 or GRFT expression levels. Diamos et al. (2020) reported that in a Bean Yellow Dwarf Virus (BeYDV)-based replicating vector system, no difference in fluorescent protein accumulation levels was observed, even though there was reduced expression of the larger replicon compared to the smaller ones "split" by viral genetic elements. This is most probably due to the complex relationship between mRNA and protein expression levels (reviewed in Liu et al., 2016;McManus et al., 2015). Protein expression levels do not rely on mRNA concentration alone, and factors that can have an impact include regulation of translation rate by small RNAs (discussed below), mRNA competition for free ribosomes (Chu et al., 2011), as well as the regulation of protein concentration independent of transcript concentrations by the ubiquitinproteasome pathway or autophagy (Balchin et al., 2016). During state transition (e.g., cell differentiation), correlation between mRNA and protein expression levels can also be affected due to delayed synthesis between mRNA and protein (Jovanovic et al., 2015;Lee et al., 2011).
When DsRed was expressed as the fourth TU, there was no detectable DsRed expression. This was also observed when J3-VHH was used instead of DsRed (see 4TU of Figure 3b e.g.), even though the expression vector was still detected in the leaves ( Figure S3). We hypothesized that this effect might be partially caused by RNA-mediated gene silencing (see Guo et al., 2016 for review), in particular, PTGS and/or translational repression of homologous mRNAs (Vaucheret et al., 2001;Vaucheret, 2006;Vazquez et al., 2004).
In plants expressing transgenes, previous studies have reported an over-abundant transcription of aberrant mRNAs that lack a 5ʹ cap (Gazzani et al., 2004) or a poly-A tail (Luo & Chen, 2007) can trigger PTGS. PTGS suppressor P19 can reduce PTGS by removing dsRNAs generated from aberrant mRNA (Silhavy et al., 2002). Here, we have shown that co-infiltration of p19 separately with a 3 or 5 TU construct significantly increased mRNA levels of the genes of interest, accompanied by up to 40-fold higher recombinant protein yields. Although co-infiltrating p19 separately did not restore expression back to the levels observed for 1 or 2 TU constructs, co-expression as the third TU in the same construct restored VRC01 expression to 2 TU levels. This might be due to the close proximity of P19 in the cell as it was delivered by the same vector. When P19 was co-expressed as the first TU, no increase in protein expression was observed. This was accompanied by mild HR most likely triggered by high levels of P19 (Garabagi et al., 2012;Siddiqui et al., 2008). This might contribute to protein loss.
The p19 co-expression experiments confirmed that PTGS was partly contributing to the drop in mRNA levels and protein expression. High transgene expression driven by the strength of the promoter has been previously reported to trigger silencing (Que et al., 1997). In this study, we have used a CAMV 35S promoter containing a duplicated transcription enhancer, with the aim of increasing transcription activity compared to the native CAMV 35S promoter (Kay et al., 1987 reported that trans-inactivation of transgenes can occur due to homology found in the promoter region and 3' region in the stacking of transgenes in a single transgenic line (Fagard & Vaucheret, 2000;Vaucheret, 1993Vaucheret, , 1994. Our data also suggested that gene silencing was able to be induced in-trans (i.e. from separate plasmids)co- To circumvent silencing due to high transgene dosage and/or exogenous promoter direct transgene silencing, the presence of repeated sequences could be reduced by employing endogenous plant promoters such as PD1 (Jiang et al., 2018) and ubiquitin-10 (Grefen et al., 2010) in some of the TUs. Using different genetic elements such as non-competing vectors  or 5ʹ and 3ʹ UTRs (Diamos & Mason, 2018), have also been shown to increase protein expression levels, even when the same genetic elements were used in multiple TUs (Diamos et al., 2020). Furthermore, BeYVD genetic elements (Diamos et al., 2020) can be used to "divide" PINNEH ET AL.
| 1669 transient expression vectors of more than 3 TUs into smaller operons when the vector is processed in the nucleus.
Interestingly, when DsRed was expressed as the first or last TU in a 5 TU construct, expression was restored to levels comparable to when it was expressed as part of a 3 TU construct. This might also be due to the stability of the DsRed transcript. When expressed as the first or last TU in a 5 TU construct, mRNA transcripts were detected at levels lower than when VRC01 HC was expressed as the last TU.
However, DsRed expression was detected while VRC01 HC was not.
This demonstrated that the level of mRNA expression might be sufficient for the expression of detectable levels of DsRed. It would be interesting to quantify the abundance and stability of DsRed mRNA transcripts using a noninvasive mRNA labeling method such as thio-modified uracil (Chan et al., 2018), and correlate it with the expression levels of DsRed protein in plants expressing permutations of the 5 TU construct.
In summary, we have shown that MIDAS-P, with its capacity to iteratively add TUs based on alternating screening of white and blue colonies, is a highly practical and predictable cloning method for the assembly of multiple genes in tandem in one vector. Using the MIDAS-P constructs, we successfully expressed an anti-HIV protein cocktail expressing the broadly neutralizing antibody VRC01 and the lectin GRFT, with expression levels reaching up to approximately 0.8 g/kg for VRC01 and approximately 5 g/kg for GRFT with coexpression of the silencing suppressor P19. This study has highlighted the limitations associated with repeated use of the same sequence elements (promoter, terminator, SAR) and future studies will aim to further extend the repertoire of TU modules optimized for plant expression to improve the number of genes that can be expressed transiently in tandem; and to mitigate the gene silencing effects observed upon coexpression of high numbers of transgenes.