Experiments

Design and Planning

The experimental section of our project focused on developing a genetic system capable of selecting and optimizing RNA–RNA interactions between modified single guide RNAs (sgRNAs) bound to catalytically inactive Cas proteins (dCas). The system was designed to identify cooperative sgRNA pairs that could form stable RNA dimers through secondary structural elements known as kissing loops. Such dimeric RNA interactions would enable two dCas proteins to form a DNA loop when bound to adjacent target sites, thereby increasing targeting precision and stability without introducing double-strand breaks.

To achieve this goal, the work was divided into two major stages:

Evaluation of Known RNA Dimerization Domains

At the first stage, we focused on testing the functionality of previously characterized RNA structures capable of forming hairpin loops and mediating RNA–RNA dimerization. We selected the dimerization domain described by Hiromi Mutsuro-Aoki et al. (2022), “Acquisition of Dual Ribozyme-Functions in Nonfunctional Short Hairpin RNAs through Kissing-Loop Interactions,” Life, 12(10):1561. This domain was inserted into the 3′ end of a standard sgRNA scaffold via a short linker to assess its ability to enhance RNA–RNA dimerization and the formation of paired dCas complexes on model DNA fragments. The modified sgRNAs were tested alongside unmodified controls using complementary methods, including electrophoretic mobility shift assay (EMSA), bio-layer interferometry (BLI), and in vivo fluorescence-based assays in bacterial systems. The obtained data demonstrated that such modification partially enhanced complex stability but was not sufficient to achieve consistent or reproducible results. Consequently, the team decided to design an improved RNA domain with stronger dimerization potential to increase the stability of paired dCas complexes.

Development of a New RNA Dimerization Domain and Selection System

Based on the results of the initial experiments, the second stage focused on developing a new RNA–RNA interaction motif and constructing a three-plasmid genetic screening system to identify and quantify the most stable sgRNA–sgRNA interactions. This selection system was built around the Ptet/TetR regulatory circuit, which allowed the formation of stable RNA dimers between sgRNAs to be translated into a measurable fluorescent signal.

The system consisted of three functional plasmids:

• Reporter plasmid — expressing GFP under the control of the Ptet promoter.
• Sensor plasmid — constitutively expressing TetR and containing dCas protein binding sites.
• Expression plasmid — encoding the dCas protein and pairs of modified sgRNAs.

In this system, strong RNA–RNA interactions between sgRNAs promote DNA looping between two dCas complexes, reducing TetR expression and, consequently, activating GFP fluorescence. This feedback provides a quantitative measure of RNA–RNA interaction stability inside living cells.

At this stage, we also introduced randomized sequences (10–20 nucleotides) at the 3′ end of the sgRNA scaffolds to identify novel dimerization sequences beyond known kissing-loop motifs. The use of chimeric sgRNAs with double scaffolds was intentionally avoided to minimize RNA size, preserve activity, and reduce potential off-target effects. The resulting sgRNA constructs were subsequently validated in vitro using minicircle-derived recombined cccDNA (mcccDNA) — a non-infectious model closely mimicking the structure and behavior of natural HBV cccDNA.

Design Objectives

At the planning stage, the following key objectives were defined:

• Construct and validate plasmids expressing sgRNAs and dCas proteins under different promoter systems.
• Establish a controllable in vivo screening system capable of detecting RNA–RNA cooperativity through GFP activation.
• Quantitatively assess RNA–RNA binding strength using EMSA and BLI biophysical assays.
• Determine how RNA dimerization influences dCas–DNA complex formation and transcriptional repression efficiency in bacterial models.
• Develop an expandable platform adaptable to different Cas protein variants and regulatory circuits.

Stage I — Dual dCas via RNA–RNA (kissing-loop) to modulate gene expression

Choice of dimerization domain and sgRNA modification

We selected the dimerization domain reported in Life 12(10):1561 (10.3390/life12101561). A standard sgRNA was modified by appending this domain to the 3′ end through a short linker. We assessed the adaptability of this modification versus unmodified sgRNA using EMSA, BLI, and cell-based assays.

Creation of literature-based structures (kissing loop) and primer design

After an extensive literature review, we selected novel RNA sequences for primer design to obtain kissing-loop structures capable of stabilizing double dCas complexes. We then compared the binding and dissociation kinetics of these structures with conventional sgRNA in dCas–RNA–DNA complexes.

Table 1. Primer sequences for kissing-loop sgRNA construction

Primer Name	Sequence (5'→3')
sg1 AF1-F	TAATACGACTCACTATGTTTCGGTTGAGCAAATAACGGTTTTAGAGCTAGAAATAGCAAGTT
sg1 AF2-F	GCTAGAAATAGCAAGTTAAAATAAGGCTAGTCCGTTATCAACTTGAAAAAGTGGCACCGAGTC
sg1 AF3-F	AAAAGTGGCACCGAGTCGGTGCTGTGATGCTGAAGTGCACACGGCA
sg2 AF1-F	TAATACGACTCACTATAGTTTCGGTTGAGCAAATAACGGTTTTAGAGCTAGAAATAGCAAG
sg2 AF2-F	GAGCTAGAAATAGCAAGTTAAAATAAGGCTAGTCCGTTATCAACTTGAAAAAGTGGCACCGAG
sg2 AF3-F	GAAAAAGTGGCACCGAGTCGGTGCTGTGATGCTGAAGTGCACACGGCAGTGGCACCGAGTCGGTGCT
sg1 Tracr-R	TGTTGCCGTGTGCACTTCAGCATC

Overlap PCR groups and conditions

We created six Overlap PCR groups as follows.

Table 2. Six primer groups for Overlap PCR

Reagent	G1	G2	G3	G4	G5	G6	Volume
T7 FW10	+	+	+	+	+	+	1 μL
Sg1AF1	+	+	+	+	+	+	2 μL
Sg1AF2	+	+	+	+	+	+	2 μL
Sg1AF3	+	+				+	2 μL
Sg3AF3			+	+			2 μL
Sg4AF3					+		2 μL
Sg1TracrR		+					1 μL
Sg2TracrR		+					1 μL
Sg3TracrR			+	+			1 μL
Sg3TracrR-T				+			1 μL
Sg4TracrR					+		1 μL
Sg4TracrR-T						+	1 μL

Common mix for all six groups: dNTP 2 μL; polymerase 1 μL; buffer 10 μL; PCR-grade water 69 μL.

After Overlap PCR, products were analyzed on a 2% agarose gel (1 g agarose in 50 mL 1× TAE).

Figure 1. Results of electrophoresis on a 2% agarose gel for the six primer sets.

In vitro transcription checks and PAGE analysis

We verified kissing-loop RNA products using Taq and a master mix. To exclude kit-related issues, six transcription reactions were set up: three with an old kit and three with a new kit.

Ultrapure Water	6 µl
DNA template	1 µl
2X sgRNA Reaction Mix	10 µl
RNase Inhibitor (40 U/µl)	1 µl
Enzyme Mix	2 µl

Table 3. Reaction (Groups 1 & 4)

Ultrapure Water	2 µl
DNA template	1 µl
2X sgRNA Reaction Mix	10 µl
RNase Inhibitor (40U/µl)	1 µl
Enzyme Mix	2 µl
NTP Mix	4 µl

Table 4. Reaction (Groups 2 & 5)

Ultrapure Water	2.8 µl
DNA template	1 µl
2X sgRNA Reaction Mix	10 µl
Ribolock	0.2 µl
Enzyme Mix	2 µl
NTP Mix	4 µl

Table 5. Reaction (Groups 3 & 6)

Incubation: 4 h at 37 °C. Products checked on 10% urea PAGE.

Figure 2. 10% TBE gel results with different enzymes (SyBr-stained). EMSA image placeholder.

Preparation of targeted DNA molecules (for EMSA/BLI)

To compare binding kinetics between kissing-loop RNA and normal sgRNA, we prepared 2PAM, 1PAM, 0PAM DNA, dCas protein, kissing-loop RNA, and normal sgRNA for EMSA and BLI. DNA results for the three PAM variants are shown below.

Figure 3. 10% TBE gel results of 2PAM, 0PAM, 1PAM DNA (SyBr-stained). EMSA image placeholder.

Verification of RNA–RNA dimer formation

We examined the kissing-loop structure <A>, <Hairpin-U> , their complexity, and the binding of K1+U1 and K1+A1 (K1 is the kissing-loop RNA from the first experiment).

Figure 4. 10% TBE gel results of RNA structures (SyBr-stained). EMSA image placeholder.

This shows successful formation of a kissing-loop structure from <A> + <Hairpin-U> , outperforming K1.

Verification of double dCas complexes (EMSA)

We next evaluated formation of double Cas complexes. The gel below shows results for varying concentrations of sgRNA–Cas9 RNP mixed with 2PAM or 0PAM DNA.

Figure 5. 10% TBE gel of RNP–DNA complexes (2PAM vs 0PAM; SyBr-stained). EMSA image placeholder.

EMSA indicated no binding to 0PAM DNA , while 2PAM DNA showed binding at 100 nM and 250 nM. Further quantification was performed by BLI.

BLI kinetics vs standard scaffolds

We used BLI to measure binding kinetics.

Figure 6. Normal sgRNA—dual Cas protein—DNA (Blue: 2PAM; Red: 0PAM).

Figure 7. Kissing-loop RNA—dual Cas protein—DNA (Blue: 2PAM; Red: 0PAM).

BLI revealed enhanced binding to 2PAM vs 0PAM. Despite design specificity, RNPs showed some non-specific binding to 0PAM. Kinetic parameters were analyzed and compared in GraphPad Prism.

Figure 8. Four groups of BLI data analyzed in GraphPad Prism.

Because 0PAM binding is nonspecific/unstable, we compared 2PAM data only:

Table 6. Average Kon, Koff, Kd for 2PAM DNA groups

Group	Kon (min⁻¹·s⁻¹)	Koff (s⁻¹)	Kd (nM)
Normal sgRNA	93,800	0.002576	27.46
Kissing-loop RNA	410,211	0.002422	5.90

Kissing-loop RNA increased Kon and reduced Koff/Kd, indicating faster and tighter targeting by dCas.

Comparative CRISPR–dCas9 activity: dual kissing-loop vs. dual simple sgRNA

The sgRNAs in this study were designed to target promoter regions upstream of the GFP gene. When guided by these sgRNAs, dCas9 binds promoter DNA and blocks transcription, thereby reducing GFP expression. Consequently, cellular GFP fluorescence provides an indirect but robust readout of CRISPR–dCas9 repression: lower median GFP intensity ⇒ stronger transcriptional repression.

To compare alternative dual-sgRNA designs (simple vs. kissing-loop scaffolds) fairly, we first attempted to restrict analysis to only those cells that actually expressed dCas9. The workflow was: fixation/permeabilization → primary antibodies (Recombinant Anti-Flag Tag (mouse mAb)) → secondary antibody (Alexa Fluor 647 AffiniPure Goat Anti-Mouse IgG) → flow cytometry (CytoFLEX). The plan was to gate Alexa 647–positive events (dCas9-expressing cells) and then quantify GFP exclusively within this population, thereby minimizing bias from imperfect transfection efficiency.

Despite multiple titrations (including 1:100), nonspecific primary binding produced background Alexa 647 signals sufficient to compromise confident separation of dCas9-positive from negative cells. Given this persistent ambiguity, we abandoned antibody gating to avoid inflating variance and instead adopted a simpler, more direct fluorescence readout.

Cells were washed three times with PBS to remove residual transfection reagents and unbound molecules, then immediately analyzed on CytoFLEX. We used median GFP intensity as the primary metric because it is robust to outliers and better reflects central tendency in potentially heterogeneous populations (half the cells have lower GFP, half higher). Representative density plots and the singlet gating strategy (to exclude aggregates) are shown in Figure 9, confirming proper sample preparation and that the collected distributions faithfully capture construct-dependent differences in GFP.

In both tested lines (HEK293T, K562), expressing dual sgRNA CRISPR–dCas9 systems reduced GFP fluorescence relative to the dCas9-only control, indicating successful promoter repression. Importantly, the dual kissing-loop sgRNA scaffold consistently yielded lower median GFP than the dual simple sgRNA scaffold, i.e., stronger repression at the same targeting sites. This pattern reproduced across multiple biological repeats.

Below is a representative dataset (three independent K562 samples) reporting median GFP (FITC-A) and normalization to the dCas9-only control (=100%). In replicate 1, the median GFP decreased from 85,558.5 (control) to 79,569 with dual simple sgRNA (−7.0%) and further to 76,756.3 with dual kissing-loop sgRNA (−10.3%). Similar trends persisted in replicates 2–3, with kissing-loop constructs outperforming simple sgRNA in each case.

Table 7. Quantitative comparison of median GFP fluorescence intensity in K562 cells transfected with different CRISPR–dCas9 systems
(Median GFP intensity measured by flow cytometry; values normalized to dCas9-only control = 100%)

	K562 cell sample	Median GFP FITC-A
1	dCas9	85558.5
	dCas9 + dual simple sgRNA	79569
	dCas9 + dual kissing-loop sgRNA	76756.3
2	dCas9	79644
	dCas9 + dual simple sgRNA	79086
	dCas9 + dual kissing-loop sgRNA	76174
3	dCas9	99292
	dCas9 + dual simple sgRNA	94264
	dCas9 + dual kissing-loop sgRNA	92708

Each row represents an independent sample set. Median GFP intensity was measured by flow cytometry, and values were normalized to the dCas9-only control (set as 100%). Both dual sgRNA systems reduced GFP expression, with the kissing-loop sgRNA system showing consistently higher repression efficiency compared to the simple sgRNA system.

Figure 9. CytoFLEX density plots and singlet-gating table illustrating cell density, singlet fraction, and GFP intensity distributions used for median calculations.

Figure 10. Bar graph of median GFP intensity in K562 for dCas9 only, dual simple sgRNA, and dual kissing-loop sgRNA across three independent experiments.

Concluding note on Stage I progress

Across Stage I we established that sgRNA scaffolds engineered with kissing-loop dimerization motifs improve biophysical and cellular performance relative to simple dual-sgRNA designs. On the biophysical side, Overlap PCR and transcription workflows (Tables 1–5; Figures 1–3) produced the intended RNA species and PAM-variant DNA substrates; EMSA verified RNA–RNA dimer formation and supported assembly of dual dCas–DNA complexes with 2PAM, but not 0PAM, targets (Figures 4–5). BLI kinetics demonstrated a higher association rate and lower apparent dissociation constant for kissing-loop sgRNAs versus unmodified scaffolds on 2PAM DNA (Table 6; Figures 6–8), indicating faster target engagement and tighter complex stability. In cells, median GFP measurements on CytoFLEX confirmed functional repression by both dual-sgRNA formats, with a reproducible additional decrease in GFP for the kissing-loop design across biological replicates in HEK293T and K562 (Table 7; Figures 9–10). Collectively, these data support a mechanistic model in which engineered RNA–RNA interactions stabilize paired dCas occupancy and promoter occlusion, thereby enhancing transcriptional repression without double-strand breaks. The residual non-specific interactions observed on 0PAM in BLI, along with modest but consistent effect sizes in populations, motivated Stage II:

1. building a robust parts pipeline (restriction/ligation/assembly) to iterate kissing-loop variants at scale,
2. implementing selectable/reportable circuits (e.g., TetR/Ptet) to enrich higher-affinity RNA dimers in vivo,
3. and preparing validated target architectures (including mcccDNA-based constructs) for downstream screening and functional evaluation.

Stage II — Constructing advanced kissing-loop systems (genetic parts and assembly)

Plasmid sourcing, culture, and purification

At the second stage, we created and tested the genetic constructs necessary to implement advanced kissing-loop systems. A series of experiments was conducted to clone and assemble genetic parts. For several designs, we used ready-made components from the iGEM collection.

Initially, we used two plasmids from the iGEM BOX (the complete list of plasmids is shown in Table 8). We transformed E. coli Top10 to obtain the target plasmids. After propagation, plasmids were purified using an Omega midi kit.

Figure 11. Results of gel electrophoresis of the obtained parts.

Table 8. Plasmids containing the target genes used in the experiment.

Plasmid species	Part ID	Well	Antibiotic	Plasmid size (bp)	Target size (bp)
Constitutive promoter	BBa_J23114	I5	Chloramphenicol	2,424	39
Strong constitutive promoter	BBa_J23110	G5	Chloramphenicol	2,424	39
Weak constitutive promoter	BBa_J23116	K5	Chloramphenicol	2,424	39
RBS	BBa_J428032	O5	Chloramphenicol	2,413	28
TetR	BBa_J428028	M4	Chloramphenicol	2,670	625
Terminator 1	BBa_J428092	C1	Chloramphenicol	2,520	135
Terminator 2	BBa_J428091	A1	Chloramphenicol	2,432	47

Additional plasmid preparation and quality check

On the following day, we obtained the remaining plasmids listed in Table 8 using the same Top10 culture and Omega midi kit purification workflow; however, the concentrations were not very high. Electrophoresis results of these additional plasmids are shown below.

Figure 12. Electrophoresis of the remaining plasmids (concentration check).

Restriction digestion and target-fragment isolation

To obtain target DNA fragments from the purified plasmids, we performed restriction digestion using Rapid Cut II BsaI (Biosharp), a type IIS endonuclease recognizing 5′-GGTCTC-3′ and cutting outside the site to generate sticky ends compatible with subsequent ligations.

Reaction (10 μL total): 8.8 μL purified plasmid (50 ng/μL), 1 μL 10× Rapid Cut Buffer, 0.2 μL BsaI (10 U/μL). Incubation 37 °C for 60 min, then heat inactivation 80 °C for 20 min. Digestions were verified by gel electrophoresis.

Figure 13. Electrophoresis results.The red-marked parts indicate lanes with uncut plasmids (control groups), showing single bands at the original plasmid size. The blue-marked parts indicate lanes with digested plasmids, showing two bands: one corresponding to the target DNA fragment and the other corresponding to the plasmid backbone . The presence of two distinct bands confirms successful digestion by BsaI. Note: The small target fragments may appear faint or migrate off the gel due to their small size; longer fragments are more clearly visible.

Gel extraction (recovery), ligation into backbone, and transformation

To isolate target fragments (e.g., 39 bp promoter, 625 bp TetR) from the digested mixtures, we performed DNA gel recovery using the Beyotime kit. Steps: excise bands under UV (minimize exposure), weigh gel, dissolve in Binding Buffer, load onto spin column, centrifuge to bind DNA, wash with Wash Buffer, elute with Elution Buffer. Concentration and purity were measured on a Nanodrop (results below).

Table 9. Concentration and purity of recovered target DNA fragments.

Sample ID (Target Fragment)	ng/µL	A260/A280	A260/A230	Purity assessment
1 (BBa_J23114)	0.5	1.09	0.01	Low
2 (BBa_J23110)	0.7	1.43	0.09	Low
3 (BBa_J23116)	2.6	3.25	0.00	High
4 (BBa_J428032)	1.6	0.95	0.19	Low
5 (BBa_J428032)	2.2	2.40	0.00	High
6 (BBa_J23116)	7.1	1.58	0.44	Low
7 (BBa_J428028)	2.2	1.20	0.17	Low
8 (BBa_J428326)	22.1	1.51	0.36	Moderate

Recovered fragments (promoter, RBS, TetR, terminators) were ligated into a linearized vector backbone (e.g., pSB1C3, BsaI-opened and gel-purified).

To verify transformation efficiency of Top10 competent cells, we transformed serial dilutions of a control plasmid (final DNA: 10, 1, 0.1, 0.01 ng/µL) into 100 µL aliquots, plated 100 µL on kanamycin LB plates, and incubated overnight. Colony counts were used to calculate efficiency (colonies/µg DNA). The results indicate high efficiency, adequate for routine cloning. Colony counts are shown below.

Figure 14. Colony counts from the transformation-efficiency test.

After overnight incubation, all colonies on kanamycin LB exhibited green fluorescence under 365 nm UV, indicating expression of the GFP reporter (linked to the assembled construct). The absence of non-fluorescent colonies suggests that ligation efficiency for intended non-GFP assemblies was low (i.e., background GFP-positive constructs predominated).

Figure 15. Green fluorescent colonies on LB-kanamycin plates (UV transillumination).

Design and synthesis of target-DNA–containing structures

When selecting target sequences, we relied on Permanent Inactivation of HBV Genomes by CRISPR/Cas9-Mediated Non-cleavage Base Editing (Mol Ther Nucleic Acids. 2020 Jun 5;20:480-490), which evaluated Cas protein activity across various regions of the hepatitis B virus (HBV) genome. Based on this analysis, we selected targets judged optimal and effective by the authors and used them in our work.

In subsequent experiments, we plan to use minicircle-derived recombined cccDNA (mcccDNA) as a model DNA molecule. After analyzing the paper, we identified a set of targets matching our criteria (see Supplement for the full list). For our experiments, we used sequences gP3 (GGGAACAAGATCTACAGCAT) and gP7 (TGCTCCAGCTCCTACCTTGT) as target sites for Cas proteins.

To test our system, we constructed sensor plasmid variants with different arrangements of the selected target sequences.

Figure 16. Schematic of the constructed element showing the positions of the most active targets on the planned mcccDNA molecule for future validation.

Figure 17. Schematics of sensor plasmid variants with different target-sequence arrangements.

Tetracycline resistance analysis and system verification

We confirmed with Sangon Biotech the feasibility to synthesize three DNA sequences, clone them into pCDFDuet-1 using EcoRI and PstI, and introduce random nucleotides at two positions within the third sequence. These three sequences and two restriction endonucleases were ordered.

We plated 10 µL of bacterial suspension containing the target plasmid (from Sangon) on 10 mm kanamycin LB agar.

Figure 18. Colony growth result.

To prevent excessively rapid growth and to standardize comparisons, we performed gradient dilutions (×10 and ×100), plated 100 µL on kanamycin LB, and incubated at 37 °C. Colonies were inoculated into 4 mL LB and grown overnight. We prepared eight plates with different plasmid/antibiotic combinations (see Table 10). We observed no growth on plates 6 and 7, whereas other plates showed growth.

Table 10. Contents of plasmid/antibiotic culture dishes.

No.	Plasmid	Antibiotic
1	P1AP1gP3gP7A + P2sfGFP	Amp + Kan
2	P1AP1gP3gP7B + P2sfGFP	Amp + Kan
3	P1AP1gP3gP7R + P2sfGFP	Amp + Kan
4	P2sfGFP	Amp
5	P1AP1gP3gP7A + P2cmR	Amp + Kan
6	P1AP1gP3gP7A + P2cmR	Amp + Kan + Cm
7	P1AP1gP3gP7A + P2cmR	Cm + Kan
8	P2cmR	Amp

We prepared a tetracycline stock (50 mg/mL) and two sets of six LB-agar plates with tetracycline concentration gradients (Table 11).

Table 11. Different concentration gradients of TetA required.

No.	C_TetA	Volume of stock solution
1	10 ng/mL	0.004 µL
2	100 ng/mL	0.04 µL
3	1 µg/mL	0.4 µL
4	10 µg/mL	4 µL
5	50 µg/mL	20 µL
6	100 µg/mL	40 µL

From the 50 mg/mL stock, we prepared 10 ng/mL, 100 ng/mL, and 1 µg/mL TetA working concentrations. One set of plates contained tetracycline gradients with chloramphenicol + kanamycin; the other set contained ampicillin + chloramphenicol + kanamycin across all six plates.

We took colonies (see Figure 18) from No. 5 (P1AP1gP3gP7A + P2cmR) and No. 2 (P1AP1gP3gP7B + P2sfGFP), prepared suspensions, and spotted 5 µL onto each plate in both sets; incubation 37 °C overnight.

Figure 19. Growth of bacteria on Plates No. 4, 5, 6 (tetracycline gradients).

Cells carrying P1P1gP3gP7B + P2sfGFP (Plate No. 2) and P1AP1gP3gP7A + P2cmR (Plate No. 5) grew at 10 ng/mL and 100 ng/mL tetracycline. Moreover, Plate No. 5 (P1AP1gP3gP7A + P2cmR) showed growth even at 1 µg/mL tetracycline, though colony numbers decreased as tetracycline increased.

Figure 20. Growth of bacteria on Plate No. 5 in tetracycline-containing media.

Figure 21. Growth of bacteria on Plate No. 2 in tetracycline-containing media.

We also prepared six plates with Amp + Kan and the same tetracycline gradients and cultured No. 1 (P1AP1gP3gP7A + P2sfGFP). Then we repeated plate preparation for No. 3 (P1AP1gP3gP7R + P2sfGFP) and performed transformation. Colonies from No. 1, No. 2, and No. 3 plates were inoculated into 4 mL LB and grown overnight at 37 °C.

Concluding note on Stage II progress

In summary, target plasmids were successfully obtained and characterized, workflows for restriction, ligation, and transformation were established, and in vivo functional testing of the assembled constructs was initiated. The antibiotic-selection and tetracycline-gradient results provide a first validation layer for construct integrity and selectable marker function, and they set the stage for integrating these advanced kissing-loop components into the full selection system described in Stage I.