Targeting cells with single vectors using multiple-feature Boolean logic

Abstract

Precisely defining the roles of specific cell types is an intriguing frontier in the study of intact biological systems and has stimulated the rapid development of genetically encoded tools for observation and control. However, targeting these tools with adequate specificity remains challenging: most cell types are best defined by the intersection of two or more features such as active promoter elements, location and connectivity. Here we have combined engineered introns with specific recombinases to achieve expression of genetically encoded tools that is conditional upon multiple cell-type features, using Boolean logical operations all governed by a single versatile vector. We used this approach to target intersectionally specified populations of inhibitory interneurons in mammalian hippocampus and neurons of the ventral tegmental area defined by both genetic and wiring properties. This flexible and modular approach may expand the application of genetically encoded interventional and observational tools for intact-systems biology.

Figure 1. Diversifying recombinase targeting strategies

a) Recombinase recognition site sequences used in Double Inverted Open reading frame (DIO) constructs. b) Schematic of recombinase-mediated expression of a DIO construct. The ORF is initially inverted relative to the promoter and flanked by two sets of incompatible recombinase recognition sites. The recombinase first inverts the gene of interest into the forward orientation then either returns the gene to the starting position or excises two recognition sites and locks the gene in the forward orientation. c) HEK293 cells transfected with recombinase/DIO combinations. Arrowheads indicate Cre/Dre cross-reactivity. d) Flow cytometry analysis of cells from c. Dot plot (top) and histograms (bottom) include biological replicates and controls. DIOs in isolation were comparable to negative control (92%-98% of samples fit by single curve) and were robustly expressed when paired with correct recombinase (99%-100% of samples fit by multiple curves). Cross-reaction was noted between Cre and Dre (multiple populations fit in 78% and 92% for cDIO + Dre, 100% for dDIO + Cre; arrowheads), but not Flp (96%-100% of samples fit with single curve). e) Viruses used in f-h. Each DIO was injected into mouse hippocampus alone or with a recombinase. f,g) Recombinase/DIO expression in vivo at low (f) or high (g) magnification. No leak was visible in absence of recombinase, and correct recombinase/DIO pairs were expressed well. Sparse expression was seen with dDIO + Cre (arrowheads; Supplementary Fig. 1; imaging individually optimized). h) In vivo optrode recordings in animals co-injected with proper recombinase/DIO pairings.

Figure 2. Intron engineering and validation for INTRSECT

a) Schematic of constructs and nomenclature. ‘empty’ denotes native introns, ‘lox’ indicates addition of lox2722 site (see methods). b) cDNA was prepared from mRNA of HEK293 cells transfected with indicated constructs. Source DNA and cDNA were amplified by PCR with primers (top) in either exons 1 and 2 (‘F1/R1’) or exons 2 and 3 (‘F2/R2’) and separated by gel electrophoresis (bottom). Note that non-spliced DNA bands containing introns vary in size, while cDNA bands from spliced mRNA are uniform. c) Sequence of DNA isolated from bands in b, with uniform removal of intron sequence with these primer sets. d) Images of eYFP expression in primary neuronal cultures transfected with indicated constructs. e) Whole cell photocurrents from primary neurons transfected with intron constructs (ChR2-eYFP: 1422.1±91.7 pA n=19, intron 1: 1590.5±186.7 pA n=12, intron 1 + lox: 1721.7±97.8 pA n=5, intron 2: 1861.6±159.1 pA n=12, intron 2 + lox: 2279.1±246.9 pA n=5, intron 1 + intron 2: 1539.5±188.8 pA n=5, p<0.01 Dunnett's multiple comparison tests ChR2-eYFP vs. intron 2 + lox). All error bars s.e.m. f) Representative images of eYFP expression in medial prefrontal cortex of mouse brain at 2 weeks post-infection timepoint with lentivirus made from indicated constructs (see Supplementary Figure 6 for full dataset). g) Whole cell photocurrents illustrating increased peak photocurrent activity after intron 1 optimization (ChR2-eYFP: 1316±120 pA n=8, intron 1 + lox2722: 1603±236 pA n=5, intron donor: 2219±258 pA n=6, p<0.05 Dunnett's multiple comparison tests, ChR2-eYFP vs. intron donor; see Supplementary Figure 4).

Figure 3. INTRSECT: diverse recombinases with engineered introns enable intersectional targeting in vitro

a,b) Schematics representing (a) cells targeted by nesting Cre-dependent directional control of a central exon and Flp-dependent directional control of all three exons (b top), such that both recombinases are necessary for all exons of the construct (C_on/F_on-ChR2-eYFP) to be in the sense direction (b bottom). c) Intermediate configurations in the activation of C_on/F_on-ChR2-eYFP. d) cDNA prepared from HEK293 cells transfected with indicated constructs and recombinases and source DNA were amplified using primers in exons 1 and 3 and separated by gel electrophoresis. Note size difference in bands from intron-containing DNA and uniformity of cDNA bands, suggesting successful splicing. e) Sequences of bands from d. Yellow denotes difference from ChR2-eYFP map sequence, representing intron sequence. f) eYFP is only expressed in the presence of both Cre and Flp in cultured neurons transfected with C_on/F_on-ChR2-eYFP (green), Cre (blue), and Flp (red) (see Supplementary Fig. 8). g) Flow cytometry dot plot (top) and biological replicate histograms (bottom) of HEK293 cells transfected with indicated recombinase/construct combinations. YFP expression is only seen with C_on/F_on-ChR2-eYFP and both Cre and Flp (arrowheads; 100% of norecombinase, Cre-alone, and Flp-alone transfections fit by a single curve, 100% of Cre/Flp co-transfections fit by multiple curves). h,i) Whole cell photocurrents (h) and action potentials (i) recorded from primary neurons transfected with ChR2-eYFP or C_on/F_on-ChR2-eYFP and Cre/Flp indicate robust ChR2 function comparable to native ChR2 (ChR2-eYFP: 1120±187 pA, n=6, C_on/F_on-ChR2-YFP: 1621±157 pA, n=7 p=0.3542: two-sided unpaired t-test); all error bars s.e.m.

Figure 4. INTRSECT specificity and functionality in vivo

a) Hippocampal slices prepared four weeks after SOM-IRES-Flp (left), PV-2a-Cre (middle), or PV-2a-Cre;SOM-IRES-Flp (right) animals were injected with AAV-DJ-C_on/F_on-ChR2-eYFP. Expression was only seen in double transgenic animals, with eYFP+ cell bodies limited to the superficial hippocampus (so: stratum oriens; sp: stratum pyramidale; sr: stratum radiatum; slm: stratum lacunosum-moleculare). b-d) eYFP+ cells in PV-2a-Cre;SOM-IRES-Flp animals expressing C_on/F_on-ChR2-eYFP were patched with biocytin filled pipettes (b), and assayed for peak photocurrent (c) and action potentials (d). e-f) After patching, slices were stained for biocytin (grey), PV (blue), and SOM (red) for identification (e,f left), followed by morphological reconstruction (e,f right). All error bars s.e.m. All identified neurons were either O-LM (e) or bi-stratified (f) and stained SOM(+), but apparently PV(lo/-) and were therefore identified largely based on their morphology. g-j) To characterize more thoroughly the cells targeted in PV-2a-Cre;SOM-IRES-Flp animals by C_on/F_on-ChR2-eYFP, slices from wild type, PV-2a-Cre, SOM-IRES-Flp, and PV-2a-Cre;SOM-IRES-Flp animals injected with virus driven by either hSyn or nEF were stained for SOM and PV. YFP+ cells driven by both nEF (g,h) and hSyn (i,j) stained SOM(+), but PV(-) (g,i; arrowheads). Imaging settings optimized for sensitivity detected negligible inappropriate expression in all but a single of 11 control animals, while double transgenic animals had high levels of expression (h,j). k,l) To examine the contribution of transgenic lines on virus expression, nEF-C_on/F_on-ChR2-eYFP was injected into dorsal CA1 of PV-IRES-Cre;SOM-IRES-Flp mice; despite relatively weak expression (k), YFP was almost completely restricted to cells co-expressing PV and SOM (k,l; arrowhead).

Figure 5. Exclusion logic using INTRSECT

a) Schematics representing target populations of cells expressing Cre and Flp (left) by nesting Cre-dependent directional control of a central exon and Flp-dependent directional control of all three exons (right). b) Intermediate states driven by Cre and Flp on constructs dependent on the exclusive presence of a single recombinase for functional expression (dashed box). c) cDNA prepared from mRNA of HEK293 cells transfected with indicated constructs and recombinases and source DNA were amplified by PCR with primers in exons 1 and 3 and separated by gel electrophoresis. Note variation in the size of bands from intron-containing DNA template and uniformity of cDNA bands, suggesting successful splicing. d) Sequences of bands from b. Yellow denotes differences from ChR2-eYFP map sequence, representing intron sequence. e,f) eYFP is only expressed in cultured neurons transfected with combinations of Cre (blue), Flp (red), and C_on/F_off (e; green) or C_off/F_on (f; green) in the exclusive presence of Cre or Flp. g) Dot exemplar (top) and replicate histograms (bottom) of flow cytometry analysis of HEK293 cells transfected with indicated recombinase/construct combinations indicates expression in the presence of a single recombinase (arrowheads) that is extinguished when both recombinases are present, which was confirmed by Gaussian mixture model analysis (multiple populations fit in 85%-96% of matched single-recombinase transfections, single population fit in 94%-99% of double-recombinase transfections). h,i) Whole cell photocurrents (h) and action potentials (i) recorded from primary neurons transfected with ChR2-eYFP or C_on/F_on, C_on/F_off, C_off/F_on-ChR2-eYFP and Cre and Flp as indicated. All error bars s.e.m.

Figure 6. Combinatorial targeting with multiplexed recombinases: beyond genetic properties

a) Comparison of short-term (ST-HSV) and optimized long-term (LT-HSV) backbones. b) Projection-based targeting of the NAc-projecting VTA cells in a wild-type mouse. Retrograde LT-HSV-mCherry-IRES-Cre was injected into the NAc, while AAV-cDIO-ChR2-eYFP was injected into the VTA. c,d) In vivo expression pattern observed with projection-based targeting. Robust eYFP expression was observed within VTA, co-localized with TH in only 48% (69/145) of cells, indicating expression in both dopaminergic and non-dopaminergic VTA-NAc projection neurons. eYFP+ fibers were seen within the NAc. e) Schematic illustrating strategy for combination projection/genetic targeting of the NAc-projecting TH+ VTA cells in a TH-IRES-Cre mouse. Cre-dependent Flp was packaged into a retrograde LT-HSV and injected into the NAc, while AAV-fDIO-ChR2-eYFP was injected into the VTA. f,g) In vivo expression pattern observed with combination projection/genetic based targeting. eYFP expression in the VTA was highly co-localized with TH staining (145/156 cells), indicating restriction of expression to dopaminergic VTA-NAc projection neurons. eYFP+ fibers were seen within the NAc. h) Dot exemplar (top) and replicate histograms (bottom) of flow cytometry analysis of HEK293 cells transfected with indicated recombinase-dependent ChR2-eYFP constructs and recombinases indicate that codon-optimized VCre is able to activate vcDIO in mammalian cells (100% of samples fit to multiple curves) and does not cross-activate cDIO or fDIO (93%-98% of off-target DIOs fit with single curve). Moreover vcDIO is not activated by Cre and Flp. Note again cross-reaction between Cre/dDIO and Dre/cDIO. Specificity was quantitatively verified by Gaussian mixture model fits/analysis.