Isolation of photoreceptors from mature, developing, and regenerated zebrafish retinas, and of microglia/macrophages from regenerating zebrafish retinas

Abstract

This paper describes experimental procedures for the dissociation of retinal cells of the zebrafish (Danio rerio) for subsequent fluorescence-activated cell sorting (FACS) and gene expression studies. Methods for dissociation of zebrafish retinas followed by FACS, and RNA isolation, were optimized. This methodology was applied to isolate pure sorted samples of rods, long wavelength-sensitive (LWS) cones, medium wavelength-sensitive (MWS; RH2–2) cones, short wavelength-sensitive (SWS2) cones, and UV-sensitive (SWS1) cones from retinas obtained at selective life-history stages of the zebrafish, and for some of these photoreceptors, following retinal regeneration. We also successfully separated lws1-expressing and lws2-expressing LWS cones from fish of a transgenic line in which lws1 is reported with green fluorescence protein (GFP) and lws2 is reported with red fluorescence protein (RFP). Microglia/macrophages were successfully sorted from regenerating retinas (7 days after a cytotoxic lesion) of a transgenic line in which these immune cells express GFP. Electropherograms verified downstream isolation of high-quality RNA from sorted samples. Examples of post-sorting analysis, as well as results of qRT-PCR studies, validated the purity of sorted populations. For example, qRT-PCR samples derived from isolated Rh2–2 cones contained detectable rh2–2 (opn1mw2) opsin transcripts, but lws opsin transcripts (lws1/opn1lw1, lws2/opn1lw2) were not detected, suggesting that the procedure likely separated double cone pairs. Through this method, pure, sorted cell samples can provide RNA that is reliable for downstream gene expression analyses, such as qRT-PCR and RNA-seq, which may reveal molecular signatures of photoreceptors and microglia for comparative transcriptomics studies.

1. Introduction

Zebrafish (Danio rerio) is a useful vertebrate model system for studies of development, disease, and regeneration. Presently this animal is also the major vertebrate model system that primarily relies on vision for predator avoidance and foraging, and the genome is sequenced and genetically modifiable, making the zebrafish an excellent model for vision science (Bilotta and Saszik, 2001; Link and Collery, 2015). Within the retinas of zebrafish, photoreceptor diversity includes one type of rod, and four morphological types of cones. The morphological photoreceptor subtypes correspond to spectral sensitivity subtypes: two members of double cones include those that are long wavelength-sensitive and medium wavelength-sensitive (LWS and MWS cones, respectively), long-single cones that are short wavelength-sensitive (SWS2 cones), short-single cones that are ultraviolet-sensitive (UV cones, a.k.a. SWS1 cones), and rods (Raymond et al., 1993). With the four cone photoreceptor subtypes in zebrafish, this animal’s vision is considered tetrachromatic (Fleisch and Neuhauss, 2006). The spatial arrangement of these photoreceptors presents a regular and stereotyped neuronal mosaic (Allison et al., 2010; Fadool, 2003). LWS/MWS double cones form alternating rows with SWS2 and UV cones (Allison et al., 2010), and rods typically form a regular array in association with the UV cones (Fadool, 2003). There is growing interest in understanding the intrinsic factors governing the biological processes within each photoreceptor subtype, and therefore a need to isolate photoreceptors of specific subtypes for such analyses.

The wavelength sensitivity of a photoreceptor subtype is determined by the photopigment that is responsible for the light transduction (Allison et al., 2004) and this pigment is composed of an opsin protein and a chromophore. Zebrafish express 10 different photoreceptor opsins, each encoded by a separate gene (Chinen et al., 2003; Morrow et al., 2011). There are two, tandemly-duplicated lws opsin (lws1 and lws2) and four, tandemly-quadruplicated mws opsin genes (rh2–1, rh2–2, rh2–3, and rh2–4). The UV (Sws1; short wavelength-sensitive) and Sws opsins are each encoded by a single gene (sws1 and sws2, respectively) (Chinen et al., 2003). The opsins of rods are encoded by the rh1 (rhodopsin) gene and a second gene, rh1–2 or rhol (rhodopsin-like)(Morrow et al., 2011). The expression of the replicated cone opsins at given developmental stages show distinct spatiotemporal patterns, such that specific opsins are expressed at specific locations of the retina (Chinen et al., 2003; Morrow et al., 2017; Sun et al., 2018; Takechi and Kawamura, 2005). Distinctive features of the transcriptomes of developing, mature, and regenerated zebrafish rod and cone photoreceptors have not been fully unveiled, and especially the identification of photoreceptor subtype-specific transcripts is yet to be realized (Hennig et al., 2008; Nelson et al., 2008; Qian et al., 2005).

During the development of the zebrafish retina, electrophysiological responses of rods are not evident until 15 to 21 days post-fertilization (dpf). In contrast, UV cone function becomes distinct at 5 dpf, SWS2 cone function at 7 to 8 dpf, and the double cones are functional at 10 to12 dpf. By the end of the first two weeks post-fertilization, zebrafish larvae have fully functional cones (Neuhauss, 2003; Saszik et al., 1999). At 30 dpf, all photoreceptor subtypes are functional (Bernardos et al., 2007; Suliman and Novales Flamarique, 2014). Our prior analysis of expression of a subset of transcripts in rods from 30 dpf zebrafish retinas suggests molecular signatures similar to those of rods from adult retinas (Sun et al., 2018). It remains untested if developing photoreceptors of other subtypes can be experimentally isolated and show molecular signatures similar to their counterparts in adult retina.

Zebrafish have the capacity to regenerate damaged retinal neurons. Following damage, Müller glia cells re-enter the cell cycle, and generate multipotent neuronal progenitors that regenerate the missing retinal cell types (Bernardos et al., 2007; Fimbel et al., 2007; Kassen et al., 2009; Nagashima et al., 2013; Powell et al., 2016; Raymond et al., 2006). Some of these new neurons have been shown to possess normal morphologies and connectivities (McGinn et al., 2018). Furthermore, there is behavioral and electrophysiological evidence of functional recovery of vision (McGinn et al., 2018; Sherpa et al., 2008; Sherpa et al., 2014). Beyond these measures, however, the structural and functional characteristics of regenerated retinal neurons, particularly photoreceptors, are not well-documented. Our previous report of the rod transcriptome of the zebrafish suggested at least some similarity in gene expression in regenerated vs. native rods (Sun et al., 2018). Isolation of regenerated photoreceptors will make the subsequent transcriptomic analysis possible for establishing their molecular signatures.

Microglia are a specialized subpopulation of macrophages resident in brain and retina, which respond to neuronal damage and to infection (McCarthy et al., 2013). Microglia have been investigated in the models of retinal degenerative diseases (Couturier et al., 2014; Li et al., 2015; Ma et al., 2012; McCarthy et al., 2013). However, the functions of microglia in the context of a retinal regenerative response in adult zebrafish remain largely unknown, although these immune cells mount a robust response to retinal damage (Mitchell et al., 2018). The ability to isolate microglia and any other immune cells that may respond to retinal damage, and are present during regeneration, and determine their transcriptomes, will help us to understand how these cells may participate in, and possibly support, regeneration.

Studies of retinal development, retinal disease, and retinal regeneration in the zebrafish therefore would be considerably advanced by the development of methods for the isolation of specific photoreceptor subtypes, and of microglia, for cell-selective analysis of gene expression. Presently a reliable and efficient practice is to use a flow cytometry cell sorting methodology to purify photoreceptors (Lakowski et al., 2011; Sharma et al., 2012). As of this writing, adult mouse photoreceptors (Feodorova et al., 2015), mouse embryonic retinal precursors (Cheng et al., 2006; Muranishi et al., 2010), retinal progenitor cells derived from human induced-pluripotent stem cells (Lamba et al., 2010), embryonic chick rod and cone photoreceptors (Enright et al., 2015), adult zebrafish rod photoreceptors (Sun et al., 2018), and adult zebrafish cone photoreceptors (all cone subtypes together) (Glaviano et al., 2016b) have been successfully isolated by flow cytometry. Photoreceptor layers from undamaged, damaged, and regenerating zebrafish retinas have also been isolated using laser capture microdissection methods (Craig et al., 2008). Zebrafish Müller glia have been successfully FACS-isolated from damaged retinas for gene expression analysis by microarray (Qin et al., 2009) and RNA-seq (Sifuentes et al., 2016). In contrast, a systematic method for the isolation of each zebrafish photoreceptor type, at different developmental stages and in regenerating retina, has not been fully established. In addition, a method for the isolation of microglia during retinal regeneration in zebrafish has not been developed, to our knowledge.

In this paper, we present a technique that allows the isolation of photoreceptors of any given subtype from the subtype-specific transgenic reporter lines at adult and developmental stages, as well as following retinal regeneration. Another, similar technique is also described for the isolation of microglia/macrophages from regenerating retinas from mpeg1:eGFP fish. The goals of these techniques are to: (1) rapidly dissociate the retinal tissue for flow cytometry; (2) establish a workflow to purify specific cone subtypes, rods, and microglia/macrophages based on fluorescent transgene expression; (3) extract high-quality RNA for specific downstream applications including RNA-Seq; and (4) improve dissociation and flow cytometry efficiencies to allow photoreceptor isolation from different biological conditions, such as larval and regenerated retinas.

2. Materials and Supplies

2.1. Animals

Zebrafish were maintained in recirculating, filtered, and monitored system water at 28.5°C, on a 14:10 light/dark cycle, according to (Westerfield, 2007). All procedures involving animals were approved by the University of Idaho Animal Care and Use Committee. Rods were isolated from the transgenic xops:eGFP line, in which rods express green fluorescent protein (GFP) under control of a Xenopus rod opsin promoter (Fadool, 2003). LWS cones were isolated from trβ2:tdTomato transgenic zebrafish in which LWS cones express tdTomato fluorescent protein under control of a thyroid receptor beta 2 (trβ2) cis-regulatory region (Suzuki et al., 2013). A subset of MWS (RH2) cones were isolated from rh2–2:GFP transgenic zebrafish, in which RH2 subtype RH2–2 cones express GFP under control of regulatory regions upstream of rh2–2 (Tsujimura et al., 2015). UV cones were isolated from sws1:GFP transgenic zebrafish, in which UV cones express GFP under control of the UV opsin promoter (Takechi et al., 2003). SWS2 cones were isolated from sws2:mcherry transgenic zebrafish, in which SWS2 cones express mCherry fluorescent protein under control of the sws2 promoter (Salbreux et al., 2012). Lws:PAC(H) transgenic zebrafish harbor a P1-artificial chromosome (PAC) clone that encompasses the lws locus. The PAC clone was modified such that GFP-polyA sequence is inserted after the lws1 promoter, reporting expression of lws1, and RFP-polyA sequence is inserted after the lws2 promoter, reporting lws2 (Tsujimura et al., 2010). Expression of lws1, and correspondingly GFP, is most abundant in the ventral-nasal region of the adult retina, whereas expression of lws2, and hence RFP, is confined to the central and dorsal-temporal region of the adult retina (Tsujimura et al., 2010). Microglia/macrophages in regenerating retina were isolated from mpeg1:eGFP (a.k.a gl22 Tg), in which the macrophage expressed gene 1 (mpeg1) promoter drives expression of eGFP (Ellett et al., 2011). The transgenic lines were the kind gifts of Jim Fadool (xops:eGFP), Rachel Wong (trβ2:tdTomato), Shoji Kawamura and the RIKEN international resource facility (rh2–2:eGFP, sws1:eGFP, and lws:PAC(H)), and Pamela Raymond (sws2:mCherry). The gl22:eGFP transgenic line was obtained from the Zebrafish International Resource Center (ZIRC).

2.2. Chemicals

Chemicals in the cell dissociation buffer included trypsin (0.05% trypsin) (TrypLE™ Express Enzyme (1X), no phenol red, Gibco; product # 12604013), neutral protease (Dispase) (Lyophilized, Worthington Biochemical; product # LS02100), papain (Lyophilized, Worthington Biochemical; product # LS003118), catalase (Worthington Biochemical; product # LS001896), and superoxide dismutase (Worthington Biochemical; product # LS003540). HI-FBS (HeatInactivated Fetal Bovine Serum, Gibco; product # 16140063) was used as the dissociation reaction quencher.

Fish were anaesthetized with MS-222 (a.k.a. tricaine methanesulfonate; Sigma-Aldrich; product # A5040) before sacrifice and dissection. Chemical lesions to retina were induced by intravitreal injection of ouabain (ouabain octahydrate, Sigma-Aldrich; product # O3125) (Sherpa et al., 2008; Sherpa et al., 2014).

We used rDNase set (Macherey-Nagel; product # 740963) for DNAseI reaction. For RNA extractions, we used NucleoSpin® RNA kit (Macherey-Nagel; product # 740955) and TRIzol LS Reagent (Ambion; product # 10296010). We made RNAse-free PBS in the lab with molecular grade Milli-Q water and adjusted pH to the values indicated.

2.3. Equipment

Dissection tools were purchased from Fine Science Tools. The flow cytometer used in this study was SONY Cell Sorter SH800. RNA quality was assessed by 2100 Bioanalyzer (Agilent Genomics) or Fragment Analyzer (Advanced Analytical). Quantitative real-time PCR (qRT-PCR) was performed with a 7900HT Fast Real-Time PCR System and SYBR-Green PCRMaster Mix (Applied Biosystems, Inc.).

3. Detailed Methods

3.1. Retinal damage

The retinas of adult zebrafish of transgenic lines xops:eGFP and trβ2:tdTomato (1.5 yrs) were subjected to chemical lesion by intravitreal injection of 10 μM ouabain (ouabain octahydrate, Sigma-Aldrich), a lesion which destroys all retinal neurons but spares Müller glia (Sherpa et al., 2008; Sherpa et al., 2014). The retinas of adult zebrafish of transgenic line mpeg1:eGFP (8 mos) were subjected to chemical lesion by intraocular injection of 2.0 μM ouabain, which destroys inner retinal neurons but spares photoreceptors and Müller glia (Fimbel et al., 2007; McGinn et al., 2017; Mitchell et al., 2018; Nagashima et al., 2013; Sherpa et al., 2014). The working stocks of ouabain were prepared in 0.65% sterile saline solution. Fish were anaesthetized by MS-222, and an incision was made across the cornea with a sapphire blade. 0.4–0.6 μL of ouabain solution containing 200 μM (for extensive lesions in xops:eGFP and trβ2:tdTomato fish; final intraocular concentration estimated at 10 μM) or 40 μM (for inner retinaselective lesions of mpeg1:eGFP fish; final intraocular concentration estimated at 2 μM) was injected into the vitreal chamber of the right eye using a Hamilton syringe. Left eyes were not damaged. Loss of fluorescence-positive photoreceptors (in transgenic lines xops:eGFP and trβ2:tdTomato) was verified by viewing retinas of anaesthetized fish with epifluorescence stereomicroscopy (Leica M165 FC) at three days post-injury (3 dpi). The presence of activated microglia/macrophages was verified by viewing retinas of mpeg1:eGFP fish at 6 dpi. Lesioned zebrafish were allowed to recover, and regenerate their neurons until 14 dpi or 30 dpi for xops:eGFP and trβ2:tdTomato fish, these times correspond to the re-establishment of retinal layers, while optic nerve head growth and neuronal production continue (Sherpa et al., 2008; Sherpa et al., 2014). Retinas of mpeg1:eGFP fish were collected at 7 dpi, a time of near-peak proliferation and when markers of neural progenitors appear (Fimbel et al., 2007).

3.2. Retinal tissue dissociation

Prior to retina collection, larvae, juveniles, and adult fish were dark-adapted for at least 12 hours, then anaesthetized with MS-222. Corneas and lenses were removed with fine forceps and scissors. The retinal tissues were dissected from either the larvae (14 dpf), juveniles (30 dpf), or adult fish, and retinal pigment epithelium (RPE) was removed from the retina cup with fine forceps. Retinas were collected into microcentrifuge tubes each containing 100μL of cooled (4°C) RNAse-free phosphate-buffered saline (PBS) (pH 7.4). Numbers of fish used in each study are listed in Table 1.

Table 1.

Summary of fish samples used in each FACS study.

Transgenic lines	Age	Retinas/ Sample	Examples of RNA yield/sample
sws1:GFP	10mo	2	21ng for GFP+	32ng for GFP−
sws2:mCherry	10mo	4	16ng for mCherry+	30ng for mCherry−
rh2–2:GFP	10mo	4	15ng for GFP+	44ng for GFP−
trβ2:tdTomato	1yo	2	26ng for tdTomato+	48ng for tdTomato−
lws:Pac(H)	10mo	8	19ng for GFP+	21ng for RFP+
XOPS:eGFP	1yo	2	24ng for eGFP+	45ng for eRFP−
14dpf XOPS:eGFP	14dpf	30–40	17ng for eGFP+	33ng for eGFP−
30dpf XOPS:eGFP	30dpf	20	27ng for eGFP+	49ng for eGFP−
14dpi XOPS:eGFP	1yo	3	25ng for eGFP+	39ng for eGFP−
30dpi XOPS:eGFP	1yo	2	19ng for eGFP+	33ng for eGFP−
14dpf trβ2:tdTomato	14dpf	30–40	18ng for tdTomato+	43ng for tdTomato−
30dpf trβ2:tdTomato	30dpf	20	22ng for tdTomato+	50ng for tdTomato−
14dpi trβ2:tdTomato	1yo	3	27ng for tdTomato+	57ng for tdTomato−
30dpi trβ2:tdTomato	1yo	2	20ng for tdTomato+	45ng for tdTomato−
7dpi mpeg1:eGFP	8mo	8	14ng for eGFP+	29ng for eGFP−

The following steps were used for enzymatic dissociation of retinal cells:

• Retinas were subjected to the tissue dissociation within 2 hours of isolation.

• Enzymatic dissociation was carried out by adding 1mL dissociation buffer (Recipes for specific conditions are listed below) to retina samples placed in a well of 12 well plate (Falcon™).

• The sample mixtures were incubated at 37°C for 10 mins, and were occasionally triturated (aspirated and expelled through a Pasteur pipet) to prevent aggregation of the tissues during the incubation process.

• 150μL HI-FBS (Heat-Inactivated Fetal Bovine Serum, Gibco) was added to the mixtures to quench the enzymatic reaction.

• Visible pieces of cell clumps from non-dissociated tissues were removed with fine forceps before and after incubating the sample mixtures on ice for 3 mins, and discarded.

• Samples were centrifuged with Eppendorf Centrifuge 5427R at 4000rpm for 3 mins, followed by removal of supernatant from cell pellets.

• Cells were resuspended in DNAseI solution and incubated at room temperature for 10 mins without agitation. DNAseI digests any leaked DNA in the cell mixture. Timing for incubation was longer for samples of regenerating retinas, as more retinae were processed in each sample (Details are described in Potential Pitfalls and Troubleshooting section).

• Samples were centrifuged (Eppendorf Centrifuge 5427R) at 4000rpm for 3 mins, and this was followed by removal of supernatant from cell pellets.

• 150–200μL RNAse-free PBS (pH=6.5) was added to the tube after removal of supernatant, and cell were resuspended immediately before FACS.

• The resulting suspension was used for cell sorting with a SONY Cell Sorter SH800 at the Center for Reproductive Biology (CRB) FACS Core, Washington State University. The cell samples were maintained on ice prior to sorting for up to four hours of wait time.

• The samples were free of visible air bubbles which might also interfere with analytic processes.

• The entire dissociation procedure should be completed within 45 mins.

To prepare 10mL dissociation buffer (for 10 reactions):

1. Dissolve 5mg glucose and 3.3mg papain (5 U/mL) into 5mL PBS (pH=7.4).

2. 2Incubate at 37°C for 20 mins, and cool the mixture to room temperature.

3. Trypsin concentrations are specific for each preparation of retina tissue.

   • •
     Adult retinas: add 5mL 0.05% trypsin.
   • •
     Regenerating or regenerated retinas: add 3mL 0.05% trypsin and an additional
   • •
     0.66mg papain (5 U/mL) in 2mL PBS (pH=7.4).
   • •
     Retinal tissues of 30 dpf fish: add 3mL 0.05% trypsin and 2mL PBS (pH=7.4).
   • •
     Retinal tissues of 14 dpf fish: add 2mL 0.05% trypsin and 3mL PBS (pH=7.4).

4. Add 1.2mg neutral protease (0.5 U/mL).

5. Add 166uL of 6mg/mL catalase (0.1mg/mL).

6. Add 30 units of superoxide dismutase.

7. Filter with a PVDF syringe filter (Thermo Scientific) to sterilize the mixture.

8. Keep the buffer refrigerated at 4°C for up for 1 month.

3.3. Fluorescence-activated cell sorting (FACS)

The primary gating strategy for FACS was based on the fluorescence intensity of cells within a sample. For the identification of target cells from tissue, the scatter characteristics (forward scatter, FSC; and side scatter, SSC) of the retinal cells were also used for gating. Sorting was performed using a 100-micron nozzle. Lasers used in these experiments were 488nm and 561nm, and photomultiplier tubes (PMT) were 525/50 for GFP sorting and 600/60 for RFP, mCherry, and tdTomato sorting. In each FACS run, an electronic threshold value was applied to forward scatter for each cell subtype so that only events with an intensity greater than that threshold value were acquired and processed.

The dissociated cellular samples that were fed into the FACS machine and subjected to fluorescence-based sorting contained fluorescent photoreceptors or microglia/macrophages, and all other non-fluorescent retinal cells. Following sorting, some of the sorted samples were analyzed by microscopy, with or without DAPI staining, to identify which gated populations included any debris or aggregates (Sun et al., 2018). Cell debris was usually smaller in size than actual singlet cells, and could be identified and eliminated from collection by gating based on side scatter parameters, and collecting two adjacent gated populations (within the fluorescent or the non-fluorescent population) showing higher scatter characteristics (SSC or FSC) than the debris. Large aggregated events were removed by using a gating selection referenced to the maximum forward scatter of singlet cells. The same strategy was practiced to set the gating boundary between non-fluorescent and fluorescent cell populations. Although the scatter characteristics of developing, regenerated and mature photoreceptors of a given photoreceptor type were distinct in some cases, fluorescence intensity of a particular cell type roughly remained at the same magnitude. Once the two cell populations of interest (e.g. fluorescent microglia/macrophages vs. other retinal cells) were identified, a gating strategy was adopted in order to ensure that fluorescent cells, and ultimately highly pure populations, were clearly separated and sorted from the non-fluorescent populations. User-defined gating can be adjusted as most suitable for the desired downstream applications.

The SH800 instrument employs a compensation system that is used to calculate the levels of different detected signals in a sample, and subsequently subtract the unwanted components from each channel. We used compensation analysis in these FACS experiments to adjust the mean value of fluorescence intensity of a non-transgenically-labeled population of cells from a transgenic fish, with a predefined, non-fluorescent population of a wild-type fish or a confirmed non-fluorescent cell population. When compensated, the fluorescent-negative populations were effectively aligned into the non-fluorescent region and clearly distinguished from fluorescence-positive populations. Sorted samples with successful compensation settings were occasionally checked with epifluorescence microscopy to verify dissociation efficiency and purity. Once the gating regions for a given cell type were determined and set on the instrument panel, cells were directly sorted into 15mL Falcon Conical Centrifuge tubes (Fisher Scientific) containing TRIzol LS Reagent (Ambion), or lysis buffer (from Macherey-Nagel RNA kit) to reduce the risk of RNA degradation. Volumes of reagent (TRIzol LS or lysis buffer from the kit) were selected based upon the sample:reagent ratio recommended by the manufacturer’s RNA extraction method. For example, TRIzol LS ratios within the range of 1:3 and 1:5 were achieved by using 300μL of TRIzol LS to accommodate 100μL of sorted cells.

We wish to highlight the methods used to sort lws1:GFP vs. lws2:RFP cells from the LWS:PAC(H) transgenic fish, as the sorting involved two fluorescent reporters at the same time. Once the GFP signal was compensated, the lws1:GFP population was separated from the non-GFP population. Similarly, the lws2:RFP population was separated from the non-RFP population, with RFP signal compensation. The non-GFP and non-RFP populations then contained an overlapping population of non-fluorescent cells, which was compensated with the non-trβ2:tdTomato population (all LWS cones), obtained from a separate sorting experiment.

Once the characteristics of the non-fluorescent cell population were determined, further alterations in compensation setting did not cause a change in scattering characteristics of the fluorescent populations (LWS1, GFP+ and LWS2, RFP+ cones) which helped to define the gates for the two populations. In addition, since LWS1 and LWS2 cones were virtually identical in cellular size and structure, side- and forward-scattering characteristics of the populations were congruent. With the sorting characteristic of LWS1+ and LWS2+ cells set up in the system, two distinctive populations can be identified on the bi-fluorescence plot.

We opted not to use a centrifugation step for concentrating the sorted cells. In a preliminary test, we observed that centrifugation caused a major material loss, likely due to the small mass of individual cells, and residual charges carried on the cells after the fluorescence activation during FACS. For specific downstream applications that require concentrated (pelleted) cells, this problem (material loss during a centrifugation step) may be minimized by adding 10μL 3% BSA to every 300μL solution of sorted photoreceptors, and the sample mixture can subsequently be centrifuged up to 6000rpm (Eppendorf Centrifuge 5427R) for 5 mins.

3.4. RNA extraction and quantitative real-time PCR (qPCR)

RNA was extracted from sorted samples using the NucleoSpin® RNA kit (MachereyNagel) following the manufacturer’s protocol, and then quantified and quality-checked on a 2100 Bioanalyzer (Agilent Genomics), or a Fragment Analyzer Automated CE System (Advanced Analytical) at the IBEST Genomics Resources Core, University of Idaho. cDNA was synthesized using the SuperScript® kit (Invitrogen) using random hexamer primers. Gene-specific primers for qRT-PCR experiments (Table 2) were designed using AlleleID7/84 (Premier Biosoft) and the GRCz10 version of the zebrafish genome, with the exception of sws1, which was designed using NCBI Primer Blast and GRCz11. All primers were verified by NCBI Primer Blast vs. GRCz11. qPCR amplification reactions were performed using SYBR-Green PCRMaster Mix (Applied Biosystems, Inc.) with a model 7900HT Fast Real-Time PCR System (Applied Biosystems, Inc.). 18S was chosen as the reference transcript (Sherpa et al., 2014). qRT-PCR reactions were performed with 40 amplification cycles.

Table 2.

Primers used for qPCR studies.

Gene	Sense Primer 5’ -> 3’	Anti-sense Primer 5’-> 3’
Lplastin	GCAGTGGGTGAACGAAACAC	TCGAGATCGCATACTTGGCG
mpeg1.1	CATGTCGTGGCTGGAACAGA	ATGGTTACGGACTTGAACCCG
nr2e3	CTTGCTCAACATATTCAC	GGAAGGAGAAGTAATAGTC
opn1lw1	CCCACACTGCATCTCGACAA	AAGGTATTCCCCATCACTCCAA
opn1lw2	AGAGGGAAGAACTGGACTTTCAGA	TTCAGAGGAGTTTTGCCTACATATGT
opn1sw1	ATGGTCCTTGGCTGTTCTGG	CCTCGGGAATGTATCTGCTCC
opn1sw2	ATCTGGGTGGTTTCCAACCG	ACAGGAGCGGAACTGTTTGTT
opn1mw1	CAGCCCAGCACAAGAAACTC	AGAGCAACCTGACCTCCAAGT
opn1mw2	TTTTTGGCTGGTCCCGATACA	CAGGAACGCAGAAATGACAGC
opn1mw3	TGCTTTCGCTGGGATTGGATT	CCCTCTGGAATATACCTTGACCA
opn1mw4	CACGCTTTCGCAGGATGC	CGGAATATACCTGGGCCAAC
rho	ACTTCCGTTTCGGGGAGAAC	GAAGGACTCGTTGTTGACAC
18S	GAACGCCACTTGTCCCTCTA	GTTGGTGGAGCGATTTGTCT

4. Results

4.1. FACS analysis

In FACS, it is desirable to use cell suspensions with low numbers of cell clumps, to result in a better yield of sorted cells. The FACS instrument provides a read-out, found on the instrument control panel, which can be used to indirectly assess this quality of the cell suspension. This readout is a percentage defined as singlet events/all events X 100 (where an event is any object in the FACS flow stream that is detected, and a singlet is any event that is considered to be a single cell based on size/scatter properties and falls into a designated gate). In the present study, the percentages of singlet events for all samples were maintained above 92%.

The total number of instrument-detected events derived from 2 retinas of a single adult zebrafish subjected to FACS yielded approximately ~3×10⁵ events. However, since some of the material passing through the FACS instrument was not in a gated and collected population (e.g. cell debris, which instead entered the waste stream), fewer events (2–2.5×10⁵) were usually collected.

The settings for sorting specific, fluorescently-labeled cell types were determined empirically as described in the Detailed Methods section, and reproduced in each subsequent experiment for that cell type from dissociated retinas of the corresponding transgenic zebrafish.

4.1.1. SWS1 (UV) cones (sws1:eGFP+).

FACS sort results derived from a single adult sws1:eGFP fish are presented in Fig. 1. Events (Fig. 1A) that fell within the pre-defined region indicated by the ellipse in Fig. 1A were subjected to further gating discrimination by cell counts versus fluorescence (Fig. 1B) and FSC-H versus fluorescence (Fig. 1C). Two peaks were separated based on fluorescence levels, and with compensation, the magnitude of the difference in fluorescence intensities was usually about 10² (GFP+ versus GFP-) (Fig. 1B). Gates were then set to sort GFP- and GFP+ cells (Fig. 1C, blue boxes). A typical sorting of GFP+ events accounted for up to 10–15% of total sorted events (the right gate in Fig. 1C), while GFP- events accounted for 80–85%. The two peaks in Fig. 1B corresponded to the events collected in the two gates presented in Fig. 1C, with the right gate in Fig. 1C matching the right peak in Fig. 1B and the left gate in Fig. 1C matching the left peak in Fig. 1B. Numbers of sorted GFP+ events ranged from 13,000–16,000, while sorted GFP- populations consisted of 105,000–118,000 events per fish (Fig. 1D, three biological replicates). Ratios of GFP+ (SWS1+) versus GFP- (SWS1-) events were consistent with the percentages defined by gating shown in Fig. 1C.

Figure 1. Sorting reports and numbers of events from adult sws1:eGFP (A-D) and sws2:mCherry (E-H) fish retinal tissues.

A-C. Representative (100,000 sorted events) sorting reports of adult SWS1:eGFP fish retinal tissue. A. Forward scatter (FSC-H) vs. (FSC-A) plot demonstrates effective dissociation of retinal cells. B. Events from within the purple oval of (A), appear in two distinctive peaks, the eGFP- cells and the eGFP+ cells. C. Gating strategy, with sorting percentage of the eGFP+ cells indicated. D. Numbers of sorted events collected in eGFP- (SWS1-) and eGFP+ (SWS1+) cell populations for the three biological replicates. E-G. Representative (100,000 sorted events) sorting reports of adult SWS2:mCherry fish retinal tissue. E. Forward scatter (FSC-H) vs. (FSC-A) plot demonstrates effective dissociation of retinal cells. F. The mCherry- cells and the mCherry+ cells appear in two distinctive peaks. G. Gating strategy, with sorting percentage of the mCherry+ cells indicated. H. Numbers of sorted events collected in mCherry- (SWS2-) and mCherry+ (SWS2+) cell populations for the three biological replicates.

4.1.2. SWS2 cones (sws2:mCherry+).

Fluorescence peaks for gated events derived from two sws2:mCherry+ zebrafish (4 retinas) (Fig. 1E) following compensation (Fig. 1F) matched the gates defined for sorting (Fig. 1G). Gated mCherry+ events contributed to 7–9% of total sorted events, while mCherry- events accounted for 68–79% (Fig. 1G). mCherry+ (SWS2+) cells showed similar FSC-H characteristics as compared with SWS1+ cells (compare Fig. 1G to Fig. 1C), possibly because these cone types are morphologically similar. The mCherry+ (SWS2+) gate collected 11,000–14,000 cells from 2 fish, while there were 107,000–132,000 sorted mCherry- (SWS2-) events (Fig. 1H).

4.1.3. RH2–2 cones (rh2–2:eGFP+).

For rh2–2:eGFP+ retinas (Fig. 2A-D), a bi-fluorescence plot was used to evaluate the compensation of the eGFP+ population (Fig. 2B). This method is useful when the target (fluorescence+) cells represent low percentages of the overall cell suspension. This was anticipated because adult RH2 cones can express one of four rh2 cone opsin genes (Chinen et al., 2003), and so the rh2–2:eGFP+ cones represent a fraction of the total number of RH2 cones. When appropriately compensated, the eGFP+ cells became distinguishable and distinct from the eGFP- population. With the reference to the large fluorescence-negative cell population (the left peak in Fig. 2B), the purity of the sorted eGFP+ events was ensured by judging the ‘tail’ of the cell population gated at the right peak in Fig. 2B, which ideally would have as few fluorescence-negative events as possible. Collected eGFP+ events usually made up 5–8% of total sorted events, and eGFP- events were about 86%−81% of the total (Fig. 2C). The numbers of sorted events in eGFP+ (RH2–2+) and eGFP- (RH2–2-) populations were approximately 10,000 and 150,000 respectively (Fig. 2D) from 2 fish.

Figure 2. Sorting reports and numbers of events from adult rh2–2:eGFP (A-D) and trβ2:tdTomato (E-H) fish retinal tissues.

A-C. Representative (100,000 sorted events) sorting reports of adult rh2–2:eGFP fish retinal tissue. A. Forward scatter (FSC-H) vs. (FSC-A) plot demonstrates effective dissociation of retinal cells. B. Bi-fluorescence plot shows the eGFP+ cells as a distinctive population. C. Gating strategy, with sorting percentage of the eGFP+ cells indicated. D. Numbers of sorted events collected in eGFP- (RH2–2-) and eGFP+ (RH2–2+) cell populations for the three biological replicates. E-G. Representative (100,000 sorted events) sorting reports of adult trβ2:tdTomato fish retinal tissue. E. Forward scatter (FSC-H) vs. (FSC-A) plot demonstrates effective dissociation of retinal cells. F. The tdTomato- and tdTomato+ cells appear in two distinctive peaks. G. Gating strategy, with sorting percentage of the tdTomato+ cells indicated. H. Numbers of sorted events collected in tdTomato- (trβ2-) and tdTomato+ (trβ2+) cell populations for the three biological replicates.

4.1.4. LWS cones (trβ2:tdTomato+).

For trβ2:tdTomato+ retinas (Fig. 2E-H), two distinctive peaks, corresponding to tdTomato – and +, were clearly separated (Fig. 2F). To ensure purity of the tdTomato+ population, the gating was set with the reference to the FSC-A domain of the tdTomato- cells (shown as ~400 × 1,000 FSC-A in Fig. 2G). Note that Fig. 2G plots FSC-A versus fluorescence, rather than FSC-H versus fluorescence (as in Figs. 1C, 1G, 2C), as FSC-A in this case revealed a clearer difference between the two sorted populations. Object scatter parameters of tdTomato+ events in the plot of FSC-H versus fluorescence were similar to those collected within fluorescence gates for SWS1+, SWS2+, and RH2–2+ populations (not shown). Numbers of sorted events were 23,000–28,000 collected in the tdTomato+ (TRβ2+) population and 140,000–150,000 for the tdTomato- (TRβ2-) population (Fig. 2H) per fish.

4.1.5. LWS1 and LWS2 cones (lws:PAC(H)).

In the lws:PAC(H) transgenic line, LWS cones expressing the lws1 opsin gene are reported as GFP+, and those expressing the lws2 opsin gene are reported as RFP+, providing a tool for selective identification and purification of two otherwise morphologically identical cone subtypes (Tsujimura et al., 2010). For separating these cone subtypes (Fig. 3), two gates were set up based on fluorescence (Fig. 3B), in which the upper (red fluorescence) gate (14.15%) collected RFP+ events and the gate at the bottom right (green fluorescence) (5.87%) collected GFP+ events. These gates were compensated based on both RFP and GFP fluorescence intensities, and each gate was compared with the individual plots of FSC-A versus fluorescence (Fig. 3C and 3D). GFP+ (LWS1+) events constituted 7.61% of total sorted events (Fig. 3C), while RFP+ (LWS2+) events constituted 17.31% of total sorted events (Fig. 3D), consistent with the known relative numbers of LWS1 vs. LWS2 cones (Mitchell et al., 2015; Tsujimura et al., 2010). These two percentages were generally consistent with those shown in the gating strategy presented in Fig. 3B. GFP+ (LWS1+) events (Fig. 3C) had scatter characteristics similar to those of the SWS1+ (Fig. 1C), SWS2+ (Fig. 1G), and RH2–2+ (Fig. 2C) events, while RFP+ (LWS2+) events were more abundant than GFP+ (LWS1+) events and shared scatter characteristics with TRβ2+ events (not shown). Scatter characteristics of sorted cones of different subtypes therefore remained overall, generally consistent. The GFP+ (LWS1+) population yielded an average of 9,500 events and the RFP+ (LWS2+) population yielded an average of 26,500 events (Fig. 3E) for 8 retinas.

Figure 3. Sorting reports and numbers of events from adult lws:PAC(H) fish retinal tissue.

A-D. Representative (100,000 sorted events) sorting reports of adult lws:PAC(H) fish retinal tissue. A. Forward scatter (FSC-H) vs. (FSC-A) plot demonstrates effective dissociation of retinal cells. B. Gating strategy, with the GFP+ cells and the RFP+ cells appearing as two distinctive populations. C. FSC-H vs fluorescence plot shows the percentage of the GFP+ cells. D. FSC-A vs fluorescence plot shows the percentage of the RFP+ cells. E. Numbers of sorted events collected in GFP+ (LWS1+) and RFP+ (LWS2+) cell populations for the three biological replicates.

4.1.6. Rods (XOPS:eGFP).

Collected numbers of eGFP+ rods of adult retinas using this methodology were similar to numbers collected using the method described by Sun et al. (Sun et al., 2018) (Fig. 4). With compensation, two peaks were separated based on fluorescence levels and shown in Fig. 4B. The gating strategy presented in Fig. 4C was distinct from that of (Sun et al., 2018), through the use of the compensation system (described in Detailed Methods). The eGFP+ (rod) population yielded an average of 22,000 events (Fig. 4D) per fish.

Figure 4. Sorting reports and numbers of events from adult xops:eGFP fish retinal tissues.

A-D. Representative (100,000 sorted events) sorting reports of adult xops:eGFP fish retinal tissue. A. Forward scatter (FSC-H) vs. (FSC-A) plot demonstrates effective dissociation of retinal cells. B. Events from within the purple oval of (A), appear in two distinctive peaks, the eGFP- cells and the eGFP+ cells. C. Gating strategy, with sorting percentage of the eGFP+ cells indicated. D. Numbers of sorted events collected in eGFP- (rod-) and eGFP+ (rod+) cell populations of adult zebrafish for the three biological replicates.

4.1.7. Rods and LWS cones from larval and juvenile fish retinas and from regenerated retinas.

The dissociation and sorting method described in this paper was also successful at isolating rods (from xops:eGFP retinas), and LWS cones (from trβ2:tdTomato retinas), and from non-fluorescent retinal cells in each case, using as starting material retinas obtained from larval or juvenile fish, or retinas that had regenerated following a chemical lesion. Sorting data related to rod and LWS cone purification from the larval and juvenile retinas (14 dpf, 30 dpf), and regenerated retinas (14 dpi, 30 dpi) are shown in Supplemental Figs. 1–3 and Fig. 5.

Figure 5. Sorting reports and numbers of events from regenerated (14 dpi and 30 dpi) xops:eGFP fish retinal tissues.

A-C. Representative (100,000 sorted events) sorting reports of 14 dpi xops:eGFP fish retinal tissue. A. Forward scatter (FSC-H) vs. (FSC-A) plot demonstrates effective dissociation of retinal cells. B. The eGFP- cells and the eGFP+ cells appear in two distinctive peaks. C. Gating strategy, with sorting percentage of the eGFP+ cells indicated. D. Numbers of sorted events collected in eGFP- (rod-) and eGFP+ (rod+) cell populations of 14 dpi regenerated retinas for the three biological replicates. E-G. Representative (100,000 sorted events) sorting reports of 30 dpi xops:eGFP fish retinal tissue. E. Forward scatter (FSC-H) vs. (FSC-A) plot demonstrates effective dissociation of retinal cells. F. The eGFP- cells and the eGFP+ cells appear in two distinctive peaks. G. Gating strategy, with sorting percentage of the eGFP+ cells indicated. H. Numbers of sorted events collected in eGFP- (rod-) and eGFP+ (rod+) cell populations of 30 dpi regenerated retinas for the three biological replicates.

Gating strategies for larval and juvenile xops:eGFP tissues (Supplemental Fig. 1C, G) were set with the eGFP compensation evaluated in Supplemental Fig. 1B, F respectively. The scatter characteristics of 14 dpf eGFP+ rods (as seen in FSC-H versus fluorescence of Supplemental Fig. 1C) deviated from that of 30 dpf eGFP+ rods (see Supplemental Fig. 1G), which implies that rods at 14 dpf may be distinct from rods of older fish in terms of soma size and internal structure. In both larval and juvenile samples, eGFP+ events contributed to approximately 20% of total sorted events (Supplemental Fig. 1D, H).

Using the same compensation settings as for adult, undamaged xops:eGFP samples (shown in Fig. 4C), distinctive eGFP+ and eGFP- populations were evident in regenerated tissues (using the same sorting parameters, peaks on the right in Fig. 5B and5F shared similar intensity and magnitude). Compared to the case of 14 dpi samples (7.01%; Fig. 5C), the gated eGFP+ population from 30 dpi samples contained more events (10.78%; Fig. 5G), which suggests that rod neurogenesis likely continues during the period of 14 to 30 dpi. Approximately 20,000 eGFP+ rods (3 retinas at 14 dpi, Fig. 5D) and 28,000 eGFP+ rods (2 retinas at 30 dpi, Fig. 5H) were collected from the regenerated tissues.

We practiced the same approach to analyze sorted LWS cones from the 14 dpf larval, 30 dpf juvenile, 14 dpi, and 30 dpi regenerated trβ2:tdTomato retinas (Supplemental Figs. 2–3). As with adult trβ2:tdTomato samples, we used bi-fluorescence plots to set and evaluate gates (Supplemental Fig. 2C,G), so that the populations analyzed in the upper gates of Supplemental Figs. 2B (7.32%) and 2F (9.76%) generally corresponded to populations collected within the right gates of Supplemental Figs. 2C (8.55%) and 2G (10.91%), respectively. These two regions representing groups of similar events can be clearly seen on each of Supplemental Fig. 2C, G plots. Two distinctive peaks can be seen in the plots of events versus fluorescence for sorting of regenerated samples at 14 and 30 dpi trβ2:tdTomato retinas (Supplemental Fig. 3B,F). Similar to the situation with regenerated rods, regenerated tdTomato+ (LWS) cones apparently increase in number from 14 dpi (3 retinas) to 30 dpi (2 retinas) [6.51% at 14 dpi (red box in Supplemental Fig. 3C); Supplemental Fig. 3D, to 12.58% at 30 dpi (red box in Supplemental Fig. 3G); Supplemental Fig. 3H].

4.1.8. Microglia/macrophages reported by mpeg1:eGFP in regenerating retinas.

The strategies reported in this paper also successfully purified immune cells expressing eGFP driven by the mpeg1 promoter (Ellett et al., 2011) from retinas regenerating after a chemical lesion by ouabain (Fig. 6). Two distinctive peaks can be seen in the fluorescence analysis, with the smaller, more highly fluorescent peak on the right representing the mpeg1+ population (Fig. 6B), which include immune cells expressing mpeg1-driven eGFP. The gating strategy for GFP+ cells (cells within blue box on the right in Fig. 6C), collected 4.83% of the total events. The gate on the left collected non-fluorescent cells as 42.28% of total events subjected to FACS. The collected fractions together represent 40–70% of the total number of events, fewer than those seen from other sorting experiments. It is possible that the events that were not collected represented cellular debris or dying neurons (Fimbel et al., 2007; Mitchell et al., 2018) present in these recently damaged retinas. On average, 12,000 eGFP+ (mpeg1 expressing) cells were sorted from 8 regenerating retinas (Fig. 6D).

Figure 6. Sorting reports and numbers of events from regenerating (7 dpi) mpeg1:eGFP (gl22:eGFP) fish retinal tissues.

A-C. Representative (100,000 sorted events) sorting reports of 7 dpi gl22:eGFP fish retinal tissue. A. Forward scatter (FSC-H) vs. (FSC-A) plot demonstrates effective dissociation of retinal cells. B. Events from within the orange oval of (A), appear in two distinctive peaks, the eGFP- cells and the eGFP+ cells. C. Gating strategy, with sorting percentage of the eGFP+ cells indicated. D. Numbers of sorted events collected in eGFP- and eGFP+ cell populations for the three biological replicates. MiG/MP, microglia/macrophages.

4.2. RNA quantity and quality

Approximately 15–30ng total RNA was extracted from fluorescent samples, and ≥40ng total RNA was yielded from non-fluorescent samples, derived from 2–8 adult or 20–30 larval/juvenile retinas in each case (Table 1). The RNA Integrity Numbers (RIN) (by 2100 Agilent Bioanalyzer) or the RNA Quality Numbers (RQN) (by Fragment Analyzer) of the RNA samples extracted from sorted cells in each case were greater than 8.0. RINs or RQNs of at least 8.0 are recommended by Illumina (www.illumina.com) for downstream transcriptomic analysis using their instrumentation. Representative RIN (Fig. 7A,B) and RQN (Fig. 7C,D) electropherograms are shown for purified eGFP+ (rod) samples from juvenile (Fig. 7A) and regenerated (Fig. 7B) xops:eGFP retinas, and for adult tdTomato+ (LWS) cone samples from adult trβ2:tdTomato retinas (Fig. 7C), and for eGFP+ microglial/macrophage samples from regenerating mpeg1:eGFP retinas (Fig. 7D).

Figure 7. RNA quality of samples obtained following FACS and RNA isolation.

A-D. Representative RIN (A,B) or RQN (C,D) analyses of RNA isolated from larval (14 dpf) rods of xops:eGFP retinas (A), rods of regenerated (14 dpi) xops:eGFP retinas (B), LWS cones of adult trβ2:tdTomato retinas (C), and microglia/macrophages (Mi/MP) of regenerating mpeg1:eGFP retinas (D).

4.3. Quantitative PCR (qPCR)

To determine the purity of the sorted fluorescent+ samples, we performed qPCR analysis of selected transcripts known to be cell-specific, or likely to be absent in the targeted cell populations. For many of the transcripts analyzed, a Cycle Threshold (Ct) value was readily obtained in one population, while Ct values were not obtained (or obtained in ranges indicating near absence of transcript within qPCR sample wells) in the other population. The reference gene, 18s (18s ribosomal RNA), was ubiquitously expressed in all fluorescent and non-fluorescent samples (Tables 3–6). These results indicate highly pure samples from sorted fluorescence+ and fluorescence- populations. Consequentially, for many of the potential sample comparisons, determination of relative levels of expression between samples (using ddCt method) was not justified due to (near) absence of transcript in one sample versus the other (McCall et al., 2014). Therefore, we refrained from performing ddCt calculations based on Ct values obtained throughout our results presented here, and instead provide raw Ct values in Tables to summarize results (Tables 3–6).

Table 3.

qPCR analysis of sorted SWS1 and SWS2 cone photoreceptor and non-cone photoreceptor samples.

Sorted samples		Cycle threshold (Ct) values1
		rho	opn1sw1	opn1sw2	opn1lw1	opn1lw2	nr2e3	18s
Adult sws1:GFP retinas	GFP+samples		23.41	26.60	30.92	30.04		14.20
	GFP−samples		32.35	26.72	31.18	29.21		14.98
Adult sws2:mCherry retinas	mCherry+samples	No transcript detected		24.26	No transcript detected	No transcript detected	25.84	16.89
	mCherry−samples	20.11		No transcript detected	22.03	21.97	23.17	16.41

¹Higher Ct values correspond with the presence of less transcript.

²No transcript detected = cycle threshold was “UND” (undetermined) by the instrument. Gray shading indicates these qPCR analyses were not done.

Table 6.

qPCR analysis of sorted mpeg1:GFP+ and mpeg1:GFP−samples.

Sorted samples		Cycle threshold (Ct) values1
		rho	opn1lw2	L-plastin	mpeg1.1	18s
7dpi (regenerating) mpeg1:eGFP retinas	GFP+ samples	No transcript detected2	No transcript detected	21.56	22.28	14.61
	GFP−samples	20.31	22.24	37.493	Transcript unreliably detected3	13.94

¹Higher Ct values correspond with the presence of less transcript.

²No transcript detected = cycle threshold was “UND” (undetermined) by the instrument.

³Transcript unreliably detected = transcript quantities approach single copy level (Ct = 34–40).

Sws1: GFP+ samples showed expression of opn1sw1 (sws1) (Ct = ~23 cycles), but expression of opn1lw1 (lws1), opn1lw2 (lws2), and opn1sw2 (sws2) were greatly reduced within this GFP+ population (Cts were considerably higher). GFP-negative cells from the same retinal tissues showed very low expression levels of opn1sw1 (Ct = ~32 cycles) (Table 3).

Sws2:mCherry+ samples showed expression of opn1sw2, but expression of rho, opn1lw1, and opn1lw2, was not detected (Ct > 40 cycles, reported as “undetermined” by the instrument). mCherry-negative cells from the same retinal tissues did not show expression of opn1sw2 (Table 3).

Rh2–2:GFP+ samples showed expression of opn1mw2 (rh2–2), but expression of rho, opn1lw1, opn1lw2, opn1mw1 (rh2–1), and opn1mw4 (rh2–4) was not detected, and expression of opn1mw3 (rh2–3) was in the Ct range approaching single copy levels in the sample wells (34–39 cycles). GFP-negative cells from the same retinal tissues did not show expression of opn1mw2 (Table 4).

Table 4.

qPCR analysis of sorted RH2 and LWS cone photoreceptor and non-cone photoreceptor samples.

Sorted samples		Cycle threshold (Ct) values1
		rho	opn1mw1	opn1mw2	opn1mw3	opn1mw4	opn1lw1	opn1lw2	nr2e3	18s
Adult rh2–2:GFP retinas	GFP+ samples	No transcript detected2	No transcript detected	23.84	Transcript unreliably detected3	No transcript detected	No transcript detected	No transcript detected		16.46
	GFP−samples	19.56	22.94	No transcript detected	24.27	26.52	23.74	21.86		16.03
Adult lws:PAC(H) retinas	GFP+(lws1+) samples	No transcript detected	No transcript detected	No transcript detected			26.54	Transcript unreliably detected	29.72	20.72
	RFP+(lws2+) samples	No transcript detected	No transcript detected	No transcript detected			No transcript detected	26.35	28.89	21.18
Adult trβ2:tdTomato retinas	tdTomato+samples	No transcript detected	No transcript detected				21.33	20.27	24.92	15.28
	tdTomato−samples	20.08	24.31				No transcript detected	Transcript unreliably detected	22.36	15.01
14dpf (larval) trβ2:tdTomato retinas	tdTomato+samples	No transcript detected	Transcript unreliably detected				22.89	21.91	27.46	19.04
	tdTomato−samples	21.25	23.72				No transcript detected	37.11	25.23	18.63
30dpf (juvenile) trβ2:tdTomato retinas	tdTomato+samples	No transcript detected	No transcript detected				21.91	21.34	26.16	18.12
	tdTomato−samples	21.08	23.59				No transcript detected	No transcript detected	24.58	17.79
14dpi (regenerated) trβ2:tdTomato retinas	tdTomato+samples	No transcript detected	No transcript detected				20.97	20.81	27.87	18.80
	tdTomato−samples	21.42	23.68				No transcript detected	No transcript detected	26.91	18.03
30dpi (regenerated) trβ2:tdTomato retinas	tdTomato+samples	No transcript detected	No transcript detected				23.66	22.45	27.33	18.24
	tdTomato−samples	21.53	24.50				No transcript detected	Transcript unreliably detected	25.72	17.12

¹Higher Ct values correspond with the presence of less transcript.

²No transcript detected = cycle threshold was “UND” (undetermined) by the instrument.

³Transcript unreliably detected = transcript quantities approach single copy level (Ct 34–40). Gray shading indicates these qPCR analyses were not done.

Lws1:GFP+ samples showed expression of opn1lw1, but expression of rho, opn1mw1, and opn1mw2 was not detected in these samples, and expression of opn1lw2 was in the Ct range approaching single copy levels. RFP+ cells from the same retinal tissues showed expression of opn1lw2, but expression of rho, opn1lw1, opn1mw1, and opn1mw2 was not detected (Table 4). These findings for the LWS and RH2 cone populations are noteworthy because they demonstrate that the lws-expressing and rh2-expressing members of the double cone pair can be successfully separated from each other for downstream analyses.

Trβ2:tdTomato+ samples of all conditions showed expression of opn1lw1 and opn1lw2, but expression of rho, opn1mw1, and opn1mw2 was not detected or was in the singly-copy Ct range. tdTomato-negative cells from the same retinal tissues showed no (or single-copy levels of) expression of opn1lw1 and opn1lw2 (Table 4).

Xops:eGFP+ samples of all conditions examined showed expression of rho, but expression of opn1lw1 and opn1lw2, was not detected. GFP-negative cells from the same retinal tissues did not show expression of rho (Table 5).

Table 5.

qPCR analysis of sorted rod photoreceptor and non-rod photoreceptor samples.

Sorted samples		Cycle threshold (Ct) values1
		rho	opn1lw1	opn1lw2	nr2e3	18s
14dpf (larval) xops:eGFP retinas	GFP+samples	19.86	No transcript detected3	No transcript detected	24.39	18.98
	GFP−samples	No transcript detected	22.77	22.18	26.92	18.42
30dpf (juvenile) xops:eGFP retinas	GFP+samples	20.37	No transcript detected	No transcript detected	25.39	18.19
	GFP−samples	No transcript detected	21.66	20.89	28.76	17.88
14dpi (regenerated) xops:eGFP retinas	GFP+samples	18.65	No transcript detected	No transcript detected	25.39	16.95
	GFP−samples	No transcript detected	20.12	19.74	28.52	16.42
30dpi (regenerated) xops:eGFP retinas	GFP+samples	19.30	No transcript detected	No transcript detected	26.41	17.01
	GFP−samples	No transcript detected	21.35	20.53	28.72	16.56

¹Higher Ct values correspond with the presence of less transcript.

²Transcript unreliably detected = transcript quantities approach single copy level (Ct = 34–40).

³No transcript detected = cycle threshold was “UND” (undetermined) by the instrument.

We also analyzed expression of nr2e3, as a target that would likely be detectable in all samples. Nr2e3 promotes rod identity in mouse (Cheng et al., 2006), but in zebrafish has been detected by in situ hybridization throughout the photoreceptor layer, and in some cells of the inner nuclear layer (Kitambi and Hauptmann, 2007; Nelson et al., 2008). This transcript was indeed detected in all samples that were evaluated for its presence (Tables 3, 4, 5).

Mpeg1:eGFP+ samples showed expression of immune cell and microglial/macrophage-associated transcripts L-plastin and mpeg1.1, while the GFP-negative cells from the same tissues showed Ct values indicated single copy levels (Table 6). Expression of photoreceptor transcripts rho and opn1lw2 were not detected in eGFP+ samples, but were present in eGFPnegative samples (Table 6).

5. Potential Pitfalls and Troubleshooting

5.1. Presence of cell aggregates

The incubation time for tissue dissociation may require adjustment depending on the sample. In general, a dissociation protocol that is cell survival-friendly, is likely to produce more aggregates containing a combination of fluorescent and non-fluorescent cells. Such a protocol then creates additional drawbacks because the presence of aggregates may adversely affect the accuracy of the FACS instrument compensation settings. For the method and sample types described in this paper, the dissociation step required 10 minutes. If tissues weigh more than ~100mg (approximately equivalent to the weight of 10 retinas from adult fish), incubation time for dissociation can be set longer, up to 15 minutes. Pilot experiments should be used to determine optimum incubation time. For example, after 10 to 15 minutes of incubation, suspended cells can be stained with DAPI and viewed with epifluorescence microscopy to evaluate dissociation efficiency (presence or absence of aggregates), and, if desired, cell survival. Our preliminary experiments determined that incubation with the dissociation protocols described in this paper should not exceed 20 minutes, otherwise cell survival was compromised (not shown). Dissociated cells generally survive well during 10 to 15 minutes of incubation. For experiments using 10–100 mg of tissue, we recommend using the indicated incubation time for the dissociation.

5.2. Insufficient materials for downstream applications

The workflow described in this paper provides a sufficient number of sorted events from a rather small number of retina samples after tissue dissociation (Table 1). For the purpose of RNA extraction and analysis of gene expression, the number of sorted events was well-correlated to the RNA yield. For example, this methodology can work for as few as ~9,000 collected cells to yield ~15ng RNA with the suggested protocols for TRIzol® LS Reagent (ThermoFisher) or NucleoSpin® RNA XS kit (Macherey-Nagel), while ~30ng RNA can usually be yielded from sorted samples with ~25,000–30,000 cells. However, we note that we were unable to reliably sort mpeg1:eGFP+ microglia/macrophages from undamaged retinas due to insufficient GFP+ events, even with pooling retinas from multiple fish. A simple solution for insufficient sorted materials for downstream applications such as RNA extraction generally is pooling samples from increased numbers of transgenic fish to increase the sorting yield. If this resource is limited, one alternate, possible troubleshooting method is to improve the dissociation efficiency to reduce the abundance of aggregates. For the proposed dissociation protocol in this paper, an increased concentration of trypsin or neutral protease would boost the dissociation power. However, this adjustment carries risk of the generation of more cell debris and may jeopardize cell survival. Another alternative strategy, which may compromise purity for the sake of increased sorting yields (yet potentially appropriate for certain downstream applications), is to adjust user-defined gating of fluorescent+ versus fluorescent- populations for sorting to be less stringent. For example, in an additional experiment (data not shown), mpeg1:eGFP+ and GFP- populations were sorted using less stringent gating in order to increase cell yield. Following this experiment, qPCR analysis revealed that mpeg1.1 transcripts were present in the GFP+ population, while they were very weakly detected in the GFP- population (average Ct = 34.5, instead of essentially absent following stringent gating; Table 6). Nonetheless, this less stringent gating still produced highly enriched cell populations following FACS.

5.3. Inconsistent sorting of cell populations during a FACS run

An inconsistent sorting of cell populations over the time of a single FACS run is likely to result in decreased purity of the sorted cells. Such inconsistency may result from the fluorescence fading during the sorting process, and/or quenching of signal by cell death, photobleaching, and/or high stickiness of the cell suspension due to leaked DNA, which in turn contributes to aggregate formation.

In addition to using DAPI staining to evaluate cell integrity and survival of collected cells, a drop of DRAQ7 DROP & GO™ (BioStatus) can be added to the cell mixture to label dead cells (not shown), as a supplemental strategy to monitor cell attributes during a sort. DRAQ7 is a fluorescent DNA dye that stains nuclei of permeabilised cells. However, for samples expressing mCherry, RFP, tdTomato, or other red/far red fluorophores, the usage of DRAQ7, which also fluoresces in the red/far red, is not recommended. Photobleaching can be minimized by storing the samples in a dark environment (e.g. keeping on ice in an insulated container) prior to FACS. In the protocols described here, leaked DNA is removed by the DNAseI reaction. However, we noted that samples of regenerating or regenerated retinas generally resulted in stickier solutions than other samples. It is possible that cells in these tissues were more fragile and easily damaged. In such situations, longer incubations with rDNAse may reduce stickiness.

To evaluate sorting consistency using the protocols described here, sorted events were collected in different quantities, and/or at different times during a sorting run, and compared. Fig. 8 shows sorting reports of 10,000 representative sorted events (Fig. 8A-C), for the first 100,000 and second 100,000 events collected (Fig. 8D-G), and for 300,000 sorted events (Fig. 8H,I). These reveal excellent consistency for sorting the given cell types (compare to Fig. 4 for adult rods, and to Fig. 2 for adult LWS cones).

Figure 8. Consistency of sorting reports during FACS runs.

A-C. The first 10,000 sorted events of adult xops:eGFP fish retina tissue. This sorting report is generally consistent (A and B) to that of 100,000 sorted events. Sorted rod photoreceptors usually contribute to 15–25% of total sorted events. C. Gating strategy. D-G. A comparison of reports between the first 100,000 sorted events and the second 100,000 sorted events of the same adult xops:eGFP fish retina tissue sample. The first 100,000 sorted events (D and E) had similar sorting percentages to the second 100,000 sorting events (F and G) using the same gating strategy. H-I. A report of 300,000 sorted events with regenerated 30 dpi trβ2:tdTomato fish retina tissue. Dissociation of retina tissue was effectively performed (H, as FSC-H versus FSC-A). I shows gating strategy and sorting percentages.

5.4. Low dissociation efficiency

A low dissociation efficiency (<90%) may result from uneven distribution of aggregates in a cell suspension. There are three considerations for this problematic situation: (1) The stream of flow cytometry can be run with a lower speed (or air pressure). A typical speed is <500 events/second. (2) Cell samples can be diluted before FACS to reduce the concentrations of aggregates or cell clusters. (3) Post-sort analysis can be performed to evaluate the purity of sorted fluorescent cells. Supplemental Fig. 4 provides an example of such a post-sorting analysis following FACS sorting of eGFP+ rods from regenerated retinas.

5.5. Collection of false-positive artifacts by using fluorescence intensity as the only sorting parameter

Using fluorescence intensity as the only gating method can lead to the increased risk of collecting false-positive artifacts such as aggregates and/or the remains of damaged cells. In the methodology described in this paper, we used a gating method for collection of photoreceptors and microglia based on FSC-H/A and SSC characteristics in addition to straightforward demarcation with fluorescence intensities. In addition, the number of events versus fluorescence intensity should be consistent with known proportions of targeted cell population vs. entire sample (e.g. from histological studies of that tissue).

5.6. Limitations of the instrumentation

A disadvantage of standard, commercially-available flow sorters is that it is almost impossible to separate a population of co-labeled cells (e.g. GFP+; RFP+) from singly-labeled cells which have similar object scatter parameters.

As a final consideration, the determination of various gating strategies requires not only a suitable flow cytometric sorter, but also knowledge of the cell biology of the sorted samples, properly validated protocols, and experience in FACS techniques and troubleshooting.

6. Discussion

We demonstrate and illustrate a multi-step methodology to obtain purified samples of specific cell types of the zebrafish retina, suitable for the preparation of high-quality RNA for downstream gene expression analysis. This workflow was successful for starting material obtained from adult, juvenile, regenerated, or regenerating retina, and was successful for isolating specific photoreceptor subtypes and for isolating microglia/macrophages. These cell types represent distinct target populations in terms of morphology, position within the retina, and associations with surrounding cells. Flow cytometry and cell sorting techniques were optimized to increase dissociation efficiency and reduce sample loss, and various gating strategies were evaluated based upon scatter and fluorescence characteristics of fluorescent versus nonfluorescent populations, as well as with those of the non-fluorescent controls used for compensation.

We are aware of only two previous reports of FACS sorting of zebrafish photoreceptors – isolation of cones (Glaviano et al., 2016a), and our prior study in which rods were isolated (Sun et al., 2018), and none for zebrafish retinal microglia/macrophages, although microglia have been successfully sorted from zebrafish brain (Oosterhof et al., 2017). The protocol described here represents an improvement upon the retina dissociation and sorting methods, due to the requirement for far less starting material, and therefore fewer fish, in comparison to the approach of (Glaviano et al., 2016a). For example, one adult trβ2:tdTomato fish (2 retinas) yielded approximately 25,000 tdTomato+ LWS cones and ~25–30ng RNA in the present study, while ~1,000,000 cones (of all subtypes) were sorted from 30 retinas collected from adult 3.2gnat2:EGFP (of all subtypes), and ~80ng RNA was subsequently extracted, in the Glaviano et al. study. The current protocol is also more versatile than our prior approach (Sun et al., 2018), most importantly due to the use of the SH800 FACS sorter, with additional lasers and the better compensation system to allow efficient sorting of a wider range of fluorescent samples. Both strategies yielded favorable outcomes for the downstream gene expression analysis, such as RNA-seq.

Isolation of specific photoreceptor subtypes and microglia/macrophages with high purity and in sufficient quantities for downstream studies of cell-specific gene expression represents a useful research objective. Knowledge of physiological and pathophysiological features of the transcriptomes of mature, developing, and regenerated rod and cone photoreceptors is currently incomplete, although this information is critical to identify regulatory genes that may be involved in photoreceptor development, degenerations, and regeneration, and in understanding regulation of morphological maturation of these two cell populations. The techniques described in this paper may be practiced for the downstream experiments of next-generation sequencing which advances the characterization of molecular regulators specific to and enriched in photoreceptors and microglia/macrophages. Furthermore, the zebrafish system provides outstanding advantages in identifying molecular signatures of photoreceptors in models of retinal disease, and potentially screening small molecules for therapeutic applications. The methodology described here could also be applied to the assessment of drug treatments affecting a given photoreceptor subtype or microglia/macrophages.

Supplementary Material

1

Supplemental Figure 1. Sorting reports and numbers of events from larval (14 dpf) and juvenile (30 dpf) xops:eGFP fish retinal tissues. A-C. Representative (100,000 sorted events) sorting reports of 14 dpf xops:eGFP fish retinal tissue. A. Forward scatter (FSC-H) vs. (FSC-A) plot demonstrates effective dissociation of retinal cells. B. Bi-fluorescence plot shows the eGFP+ cells as a distinctive population. C. Gating strategy, with sorting percentage of the eGFP+ cells indicated. D. Numbers of sorted events collected in eGFP- (rod-) and eGFP+ (rod+) cell populations of 14 dpf zebrafish for the three biological replicates. E-G. Representative (100,000 sorted events) sorting reports of 30 dpf xops:eGFP fish retinal tissue. E. Forward scatter (FSC-H) vs. (FSC-A) plot demonstrates effective dissociation of retinal cells. F. Bi-fluorescence plot shows the eGFP+ cells as a distinctive population. G. Gating strategy, with sorting percentage of the eGFP+ cells indicated. H. Numbers of sorted events collected in eGFP- (rod-) and eGFP+ (rod+) cell populations of 30 dpf zebrafish for the three biological replicates.

2

Supplemental Figure 1. Sorting reports and numbers of events from larval (14 dpf) and juvenile (30 dpf), trβ2:tdTomato fish retinal tissues. A-C. Representative (100,000 sorted events) sorting reports of 14 dpf trβ2:tdTomato fish retinal tissue. A. Forward scatter (FSC-H) vs. (FSC-A) plot demonstrates effective dissociation of retinal cells. B. Bi-fluorescence plot shows the tdTomato+ cells as a distinctive population. C. Gating strategy, with sorting percentage of the tdTomato+ cells indicated. D. Numbers of sorted events collected in tdTomato- (trβ2-) and tdTomato+ (trβ2+) cell populations of 14 dpf zebrafish for the three biological replicates. E-G. Representative (100,000 sorted events) sorting reports of 30 dpf trβ2:tdTomato fish retinal tissue. E. Forward scatter (FSC-H) vs. (FSC-A) plot demonstrates effective dissociation of retinal cells. F. Bi-fluorescence plot shows the tdTomato+ cells as a distinctive population. G. Gating strategy, with sorting percentage of the tdTomato+ cells indicated. H. Numbers of sorted events collected in tdTomato- (trβ2-) and tdTomato+ (trβ2+) cell populations of 30 dpf zebrafish for the three biological replicates.

3

Supplemental Figure 3. Sorting reports and numbers of events from regenerated (14 dpi and 30 dpi) trβ2:tdTomato fish retinal tissues. A-C. Representative (100,000 sorted events) sorting reports of 14 dpi trβ2:tdTomato fish retinal tissue. A. Forward scatter (FSC-H) vs. (FSCA) plot demonstrates effective dissociation of retinal cells. B. The tdTomato- cells and the tdTomato+ cells appear in two distinctive peaks. C. Gating strategy, with sorting percentage of the tdTomato+ cells indicated. D. Numbers of sorted events collected in tdTomato- (trβ2-) and tdTomato+ (trβ2+) cell populations of 14 dpi regenerated retinas for the three biological replicates. E-G. Representative (100,000 sorted events) sorting reports of 30 dpi trβ2:tdTomato fish retinal tissue. E. Forward scatter (FSC-H) vs. (FSC-A) plot demonstrates effective dissociation of retinal cells. F. The tdTomato- cells and the tdTomato+ cells appear in two distinctive peaks. G. Gating strategy, with sorting percentage of the tdTomato+ cells indicated. H. Numbers of sorted events collected in tdTomato- (trβ2-) and tdTomato+ (trβ2+) cell populations of 30 dpi regenerated retinas for the three biological replicates.

4

Supplemental Figure 4. Post-sorting analysis of sample purity. Post-sorting analysis of 30 dpi, xops:eGFP, GFP+ samples. The re-sort of the sorted sample gives a consistent FSC-H versus FSC-A plot (A) in comparison with the primary sorting, and sample purity of 96.87% (B). Compare to Fig. 5E,G.

Highlights:

• Retinal cell dissociation and FACS procedures are described that successfully isolate individual photoreceptor types from zebrafish retinas.

• Procedures are also described that successfully isolate microglia/macrophages from regenerating retinas, and some photoreceptor types from regenerated retinas and from larval and juvenile retinas.

• Sample purity was confirmed by qRT-PCR of cell-specific transcripts, and RNA sample quality and quantity were assessed by 2100 Bioanalyzer (Agilent Genomics) or Fragment Analyzer (Advanced Analytical).

• These procedures open possibilities for downstream analyses of gene expression in a variety of specific retinal cell types, originating from a wide range of in vivo conditions.