Cell specific photoswitchable agonist for reversible control of endogenous dopamine receptors

Dopamine controls diverse behaviors and their dysregulation contributes to many disorders. Our ability to understand and manipulate the function of dopamine is limited by the heterogenous nature of dopaminergic projections, the diversity of neurons that are regulated by dopamine, the varying distribution of the five dopamine receptors (DARs), and the complex dynamics of dopamine release. In order to improve our ability to specifically modulate distinct DARs, here we develop a photo-pharmacological strategy using a Membrane anchored Photoswitchable orthogonal remotely tethered agonist for the Dopamine receptor (MP-D). Our design selectively targets D1R/D5R receptor subtypes, most potently D1R (MP-D1ago), as shown in HEK293T cells. In vivo, we targeted dorsal striatal medium spiny neurons where the photo-activation of MP-D1ago increased movement initiation, although further work is required to assess the effects of MP-D1ago on neuronal function. Our method combines ligand and cell type-specificity with temporally precise and reversible activation of D1R to control specific aspects of movement. Our results provide a template for analyzing dopamine receptors.

ChR2 has been used to control DA neurons in the SNc 3 and VTA 4 , each of which innervates diverse brain areas 5,6 . Several approaches have been implemented to increase target specificity (B-D): (B) ChR2 can be activated in presynaptic axon terminals in a selected brain area, as has been done in specific subregions of the striatum 7 . (C) ChR2 can be activated in upstream neurons that innervate specific DA neurons, as has been done with laterodorsal tegmentum (LDT) neurons (yellow), which innervate a subpopulation of VTA neurons that project to the ventral striatum 8

. (D)
ChR2 can be activated in neurons that form axo-axonal synapses onto the terminals of other neurons in a brain area, as do (i) striatal cholinergic interneurons (ChIs; orange), which release acetylcholine and activate nicotinic acetylcholine receptors on DA terminals to enhance DA release 9 or (ii) cortical neurons, which release glutamate and activate Group I mGluRs on DA axons to suppress DA release 10 . However, ChIs and cortical neurons also connect to other striatal neurons and excitation of axons could lead to antidromic action potentials that evoke release in axon collaterals that project elsewhere. (E) When ChR2 activation evokes DA release, multiple DA receptor subtypes can be activated simultaneously, via direct synapses 11 and volume transmission 12 , including sometimes a Gs/olf-coupled D1-like receptor (D1, D5) and an opposing Gi/o/z-coupled D2-like receptor (D2, D3, D4) in the same cell. DA activates receptors in each of the cell types of the striatum, in presynaptic D2 autoreceptors that inhibit release from DA neuron terminals 13 , and D1 and D2 heteroreceptors in the terminals of glutamatergic inputs to the striatum 14,15 . DA neurons co-release glutamate (Glu) 7 and/or GABA 16 , which activate postsynaptic ligand-gated ion channels (LGICs) and G protein-coupled receptors (GPCRs) (not shown). *D3R is expressed in the ventral striatum but not the dorsal striatum 17 . dMSN = direct-pathway medium spiny neurons, iMSN = indirect-pathway medium spiny neurons, ChI = cholinergic interneurons, FS = fast-spiking interneurons, LTS = low-threshold spiking interneurons, CR = calretinin interneurons.
* Receptor specificity is determined by the inherent selectivity of the pharmacophore. # opto-XRs and PTL-or PORTL-gated receptors are variants of wildtype endogenous GPCRs. However, they are not completely identical and thus may not accurately reflect the activity of endogenous receptors.

Supplementary Figure 2. Comparison of existing methods for targeting neuronal receptors
in vivo. Ligand-gated ion channels (LGICs) and G protein-coupled receptors (GPCRs) can be controlled through a variety of means. Conventional pharmacology (agonists, antagonists, and allosteric modulators) is used to increase or decrease endogenous receptor activity. However, these ligands are freely diffusible and thus not cell type selective and difficult to constrain to the site of infusion. Furthermore, infusion kinetics in vivo are slow (minutes to longer) relative to the millisecond to seconds timescale of physiological receptor activation [18][19][20] . Faster kinetics can be achieved with photopharmacology, where free caged or photoswitchable ligands can be converted to the active state in milliseconds 21 . However, free photopharmacologic ligands cannot target specific cell types. Genetic modifications (knockout, knockdown, overexpression) can target a specific brain area and cell type, but the effect is chronic and could result in compensation.

Chemogenetic
LGICs (PSAMs) 22 and GPCRs (DREADDs/RASSLs) 23 combine genetic targeting with timed activation by orthogonal synthetic ligands. They are not available for most neuronal receptors and are used to override native signaling. They activate and deactivate with the slow kinetics of conventional pharmacology. Light-sensitive GPCRs made by fusion of a transmittergated GPCR with rhodopsin (opto-XRs) are rapidly activated, but deactivate slowly, recover incompletely 24 , and do not recapitulate all of the functions of the native receptor.
Two methods of photo-pharmacology control native receptors in a way that combines genetic targeting and tight spatio-temporal control of native signaling proteins: a) Photoswitchable Tethered Ligands (PTLs) that attach covalently to a receptor via an engineered cysteine and b) Photoswitchable Orthogonal Remotely Tethered Ligands (PORTLs) that attach covalently to an orthogonal anchoring domain (e.g., SNAP-tag) that is fused to the full-length wildtype receptor [25][26][27] . These engineered receptors must be overexpressed or knocked in. Membrane-anchored ligands selectively bind a target receptor as a result of a random encounter at the cell surface in combination with their inherent binding affinity 28 . They can be incorporated with peptidic ligands (t-toxins) 28 or chemical ligands (DARTs) 29 . However, once applied, they cannot be turned off until removed by the cell, a process that can at best take days 29 . The light-sensitive domain LOV was recently incorporated into t-toxins, making them acutely activatable (LumiToxins) but takes minutes to turn off and has poor efficacy 30 . Lumitoxins cannot be incorporated with chemical ligands like those that bind many physiologically and clinically relevant neuronal receptors.
The membrane-anchored PORTL (MP) approach used in this study combines the advantages of DARTs with the spatio-temporal precision of photo-pharmacology to control unmodified native receptors.

Supplementary
D3R was coexpressed with GαoA, and M1R and mGluR1 were coexpressed with the chimeric G protein Gαiq5, allowing the receptors to couple to GIRK channels in HEK293T cells. MP-D1ago also has no effect on ionotropic receptors that are widely expressed in the brain: the AMPA receptor that delivers two wavelengths of light from two fibers. The light from a UV laser (375 nm) and a blue laser (450 nm) are combined with a wavelength combiner, and then divided by a wavelength splitter into two fibers, one for each side of the brain. The splitter rotates to prevent twining of the fibers, which are attached to the head of a mouse via the custom head implant described above.
(D) MP-D1ago requires two wavelengths of light: UV light to turn it off (left) and blue light to turn it on (right). If these wavelengths are not calibrated, D1R will be heterogeneously activated and deactivated with MP-D1ago across the brain. The efficiency of light propagation through brain tissue depends on wavelength 74  Schematic of mouse brain (coronal section) highlighting the location of the nucleus accumbens (NAc). An AAV encoding mVenus or the membrane-anchor MEAAAK:ERE and mVenus was injected into the NAc of D1-Cre mice (red dot). (C) Expression of MEAAAK:ERE in the NAc. grey bar = 1 mm. blue = DAPI staining, red = HA-tag (MEAAAK:ERE) staining. Representative of brains from n = 6 mice.

(D)
The speed of mice with MP-D1ago in NAc-dMSNs increased in response to a brief flash of blue light (450 nm, ~6 mW, 1 s) and returned to baseline after a brief flash of UV light (375 nm, ~9 mW, 1 s). There was no effect under any condition. (E) The speed of each mouse was averaged over the following two-minute periods: (i) just before exposure to blue light (UV pre), (ii) three minutes after exposure to blue light (blue), (iii) and one minute after exposure to the second flash of UV light (UV post). There was no significant difference between any condition. RM one-way ANOVA, F-values from left to right: 1.9, 0.6, 0.4, 0.8, Bonferroni. n = 6 mice for each condition.