Effect of conditional deletion of cytoplasmic dynein heavy chain DYNC1H1 on postnatal photoreceptors

Cytoplasmic dynein (dynein 1), a major retrograde motor of eukaryotic cells, is a 1.4 MDa protein complex consisting of a pair of heavy chains (DYNC1H1) and a set of heterodimeric noncatalytic accessory components termed intermediate, light intermediate and light chains. DYNC1H1 (4644 amino acids) is the dynein backbone encoded by a gene consisting of 77 exons. We generated a floxed Dync1h1 allele that excises exons 24 and 25 and truncates DYNC1H1 during Six3Cre-induced homologous recombination. Truncation results in loss of the motor and microtubule-binding domain. Dync1h1F/F;Six3Cre photoreceptors degenerated rapidly within two postnatal weeks. In the postnatal day 6 (P6) Dync1h1F/F;Six3Cre central retina, outer and inner nuclear layers were severely disorganized and lacked a recognizable outer plexiform layer (OPL). Although the gene was effectively silenced by P6, DYNC1H1 remnants persisted and aggregated together with rhodopsin, PDE6 and centrin-2-positive centrosomes in the outer nuclear layer. As photoreceptor degeneration is delayed in the Dync1h1F/F;Six3Cre retina periphery, retinal lamination and outer segment elongation are in part preserved. DYNC1H1 strongly persisted in the inner plexiform layer (IPL) beyond P16 suggesting lack of clearance of the DYNC1H1 polypeptide. This persistence of DYNC1H1 allows horizontal, rod bipolar, amacrine and ganglion cells to survive past P12. The results show that cytoplasmic dynein is essential for retina lamination, nuclear positioning, vesicular trafficking of photoreceptor membrane proteins and inner/outer segment elaboration.


Introduction
Mouse photoreceptor rod outer segments (OS) are continuously renewed every ten days [1,2]. This high degree of turnover requires reliable delivery of large amounts of transmembrane regulatory components of the dynein complex and docking sites for cargoes, including adaptor proteins [21] (Fig 1A). DYNC1H1 folds into a motor domain comprising six ATPase domains (AAA1-6) and a microtubule binding stalk region at its C-terminal region. The mouse gene encoding DYNC1H1 consists of 77 exons ( Fig 1B) spanning 65 kb of genomic DNA. To enable conditional knockouts, we acquired a cell line [22] in which a gene trap flanked by loxP and FRT sites was placed in intron 23 and a third loxP site was placed in intron 25 ( Fig 1C). A floxed Dync1h1 allele (Dync1h1 F ) ( Fig 1D) is generated following FRT-FLP recombination with Flp-recombinase. Deletion of DYNC1H1 in retina was achieved by mating with Six3Cre transgenic mice [23] expressing Cre recombinase at embryonic day 9 (E9) to yield Dync1h1 F/F ; Six3Cre knockouts ( Fig 1E); deletion of exons 24 and 25 truncates Dync1h1 after exon 23 as exon 26 is out-of-frame. The truncation point is located prior to the motor domain ( Fig 1A,  arrow), and the truncated protein is likely nonfunctional since the motor domain and MT binding stalk are lost. Genotyping (Fig 1F-1I) confirmed loss of exons 24 and 25 at P6 (F), presence of loxP (G), (Dync1h1 F/F vs. Dync1h1 F/+ ), presence of Six3Cre (H), and presence of EGFP-CETN2 (I) (for details see Methods). DYNC1H1 immunolocalizes prominently in the P21 photoreceptor inner segment (IS) (Fig 1J, right panel). At P6, DYNC1H1 is undetectable in the knockout IS ( Fig 1J, middle panel). Breeding to generate Dync1h1 F/F ;Six3Cre knockouts occurred normally and with average litter sizes. Phenotypically we observed that Dync1h1 F/F ; Six3Cre mice have smaller eyes but the extent of microphthalmia varies, even between two eyes of one animal.

Disrupted ONL/INL lamination of Dync1h1 F/F ;Six3Cre retina
Six3 (sine oculis-related homeobox 3), a transcription factor expressed in retina, RPE and brain during embryonic development [23], is required for maintenance of multipotent retinal progenitors [24]. The effect of Six3Cre expression in Dync1h1 F/+ ;Six3Cre and Dync1h1 F/F ; Six3Cre retinas was examined in transverse full-retina plastic sections at P6 and P8 (Fig 2A-2D). In Dync1h1 F/+ ;Six3Cre control sections, the retina nuclear layers are well established at each time point (Fig 2A and 2B). In Dync1h1 F/F ;Six3Cre sections next to the optical nerve head (ONH), the outer plexiform layer (OPL) and IS can be distinguished in part at P6, but both are absent at P8 (Fig 2C and 2D). The RPE appears heavily pigmented, and the merged ONL/INL is reduced in thickness while the IPL appears minimally affected. We investigated whether connecting cilia (CC), equivalent to transition zones of primary cilia, are formed at P6 when ciliogenesis begins. Transmission Electron Microscopy (TEM) images reveal rare examples where basal bodies docked to the IS membrane generating a CC (Fig 2E and 2F). At P8, control photoreceptors begin to form an outer segment ( Fig 2G) whereas Dync1h1 F/F ;Six3Cre photoreceptors degenerate and discontinue CC and OS formation ( Fig 2H).

Absence of outer segments and functional OPL in DYNC1H1-depleted photoreceptors
DYNC1H1 is expressed prominently in the photoreceptor IS of control retinas with traces of DYNC1H1 in ONL somata, OPL and IPL (Fig 3A). In Dync1h1 F/F ;Six3Cre cryosections (Fig 3,  row B), OS are unrecognizable and DYNC1H1 is absent at locations where IS should have formed. DYNC1H1 apparently persists in the IPL (Fig 3, row B) although exons 24 and 25 of Dync1h1 were excised by P6 (Fig 1F) suggesting that DYNC1H1 could not be cleared by the ubiquitin proteasome or autophagy-lysosome pathway [25].
In the P6 Dync1h1 F/F ;Six3Cre central retina, the OPL is fragmented, and a border between ONL and INL is often unrecognizable, suggesting that bipolar cell dendrites do not develop

PLOS ONE
Deletion of dynein heavy chain in retina  and knockout sections (row D) probed with anti-CtBP2/Ribeye at P6-P12. CtBP2/Ribeye localizes specifically to rod ribbon synapses and cone pedicles of control sections. CtBP2/Ribeye accumulates at nuclear layer borders and becomes progressively undetectable in synaptic termini of knockout sections. E, quantitative analysis of combined (ONL and INL) nuclear layer thickness from P6 to P12. Student's t-test, � indicates p<0.05. well (Fig 3, rows B, D). The combined (ONL and INL) nuclear layer thickness is reduced by 20% at P6 (Fig 3E) suggesting reduced neurogenesis or premature cell death. Centrosomes, identified by EGFP-CETN2, fail to extend connecting cilia and aggregate with traces of DYNC1H1 in foci (Fig 3, row B', P6 and P8). The merged ONL/INL progressively shrinks from P6 to P12 and reveals 2-5 nuclear rows by P12 (Fig 3E). We probed for presence of CtBP2/Ribeye, a rod and cone synaptic marker delineating the OPL in controls (Fig 3, row C) [26], and observed that mutant photoreceptor synaptic termini are devoid of CtBP2/Ribeye (Fig 3, row D) suggesting absence of functional ribbon synapses. The results support a role for dynein in postnatal nuclear migration, lamination and ONL/INL architecture.

Dynein organizes membrane protein trafficking and photoreceptor polarization
Developmental processes defining photoreceptor polarity, i.e., synaptogenesis versus OS elaboration, are key to maturation of sensory neurons. We examined the fates of two prominent OS proteins, rhodopsin and cGMP phosphodiesterase (PDE6), in the dynein knockout environment relative to controls. In P6 control retina, rhodopsin is located in the distal IS near the basal body; rhodopsin localizes to the ROS exclusively at P8 and later (Fig 4, row A). In the Dync1h1 F/F ;Six3Cre central retina at P6 and P8 (Fig 4, row B), ROS fail to form and rhodopsin is deposited together with centrioles in disorganized pockets within the nuclear layer, likely incorporated into ER membranes (Fig 4, row B'). At P10 and P12, rhodopsin accumulates around nuclei and in sloughed membrane at the sclerad edge of the ONL. At P16, the nuclear layer consists of one row of nuclei presumably consisting of surviving mutant cones (results not shown).
Rod and cone PDE6, key enzymes of the phototransduction cascade, are heterotetrameric soluble proteins that are membrane-anchored by isoprenylation. Both are synthesized in the cytosol, dock to the ER for posttranslational processing (isoprenylation) and traffic to the OS in a complex with PDE6D by diffusion, independent of dynein [27]. Using anti-PDE6 antibody (MOE) that recognizes both rod and cone PDE6 [28], we show that PDE6 localizes exclusively to OS of control cryosections (Fig 4, row C). Iin P6 mutant sections, PDE6 accumulates together with centrosomes in a cytosolic space or in foci at the sclerad edge of the ONL (Fig 4, row B). At P8 and later, PDE6 accumulates in immunoreactive structures distal to the basal body (see Fig 4, row D' at P12, white arrows). We interpret the MOE-positive structures to be most likely diminutive remnants of rod or cone OS (lacking visual pigment) as observed in germline rhodopsin knockouts [29,30].

Inner retina cells survive longer than rods
Persistence of DYNC1H1 in the mutant IPL (Fig 3, row B) prompted us to identify surviving cells of the inner retina (horizontal, rod bipolar, amacrine, ganglion cells and Müller glia). Calbindin, a soluble Ca 2+ -binding protein, serves as a marker for horizontal (HCs), amacrine and ganglion cells (GCs) [31]. HCs are located at the sclerad INL of control retina (Fig 5, row A). In the P8 mutant retina, calbindin-positive horizontal and amacrine neurons appeared fewer in number. At P12 when photoreceptors have degenerated, several calbindin-positive neurons survived (Fig 5, row B). We tested for presence of rod bipolar cells (BPs) employing anti-PKCα, a protein kinase abundantly expressed in rod BP dendrites, cell bodies and axons ( Fig  5, row C) [32]. Rod synapses contact BP dendrites to transmit the scotopic light impulse to ganglion cells. At P6 and later, BP nuclei are located at the sclerad edge of the INL; by contrast, Dync1h1 F/F ;Six3Cre BP nuclei are scattered throughout the INL (Fig 5, row D).
Müller glia cells (MGCs) are the last-born and last postmitotic cell type of the mammalian retina [33]. To identify the fate of MGCs in the Dync1h1 F/F ;Six3Cre retina, we employed a glutamine synthetase antibody in P6-P12 cryosections (S1 Fig). The results show that while MGCs in control sections are undetectable at P6, MGC endfoot structures form around P8. MGCs extend processes into the ONL by P10 and microvilli surrounding the IS can be

PLOS ONE
Deletion of dynein heavy chain in retina detected by P14 (S1 Fig, row A). Nuclei are precisely centered in the INL by P10. In Dync1h1 F/F ;Six3Cre retina, endfoot structures, microvilli and processes are absent (S1 Fig,  row B). We conclude that functional MGCs never form the Dync1h1 F/F ;Six3Cre retina, and that persistent DYNC1H1 expression in the mutant IPL mostly originates from remaining amacrine, BP and ganglion cells.

Cone survival in Dync1h1 F/F ;Six3Cre central retina
To investigate the fate of cones in the absence of dynein, we employed anti-S-opsin, anti-MLopsin, and anti-cone arrestin as cone markers. Mouse cones express two visual pigments in most cones, M-opsin and S-opsin, and both localize exclusively to the cone OS. Cone arrestin (ARR3) is a soluble protein that freely diffuses through the cell, thereby visualizing synaptic pedicles, cone nuclei and cone dendrite/axons. Cone OS arise as early as P4 and cone pedicles around P7 [34]. In control sections, cone OS are weakly detectable at P6 (not shown) and rise

PLOS ONE
to near full length by P12 (Fig 6, rows A, C). In knockout sections, S-opsin and M-opsin aggregate near the edge of the nuclear layer and OS are not formed (Fig 6, rows B, D).
Cone nuclei are scattered throughout the ONL at P6 but move toward the OLM by P12, presumably pulled by dynein [35,36]. In our control sections, cone nuclei had mostly reached their final position by P12 (Fig 6, row E). In contrast, mutant nuclei are positioned mostly near the OPL edge (Fig 6, row F, P12). Synaptic enlargements representing developing cone pedicles can be observed as early as P6 in control sections (Fig 6, row E), clearly marking the OPL. In Dync1h1 F/F ;Six3Cre retina, developing mutant cone pedicles are formed allowing delineation of the mutant OPL (Fig 6, row F). Mutant pedicles, however, do not contain CtBP2/ Ribeye, suggesting dynein may be necessary for its correct positioning.
Preserved retina lamination and function at the peripheral retina. Mitosis ceases in the central retina by P6, and in the peripheral retina by P11 [37]. Further, Six3Cre drives Cre recombinase expression more thoroughly in the central than in the peripheral retina [24]. Consequently, lamination of the nuclear layer and morphogenesis of inner and outer segments can be observed at the peripheral edge of the Dync1h1 F/F ;Six3Cre retina while the central retina is degenerated. This is demonstrated by comparing a Dync1h1 F/F control cross section retina with a knockout at P16 (Fig 7). The control retina shows consistent expression of DYNC1H1 in the IPL, OPL, and IS of the center and periphery (Fig 7A). By contrast, the Dync1h1 F/F ; Six3Cre central retina shows low and patchy presence of DYNC1H1 within the merged ONL/ INL or at the sclerad edge of the ONL; the peripheral retina is less affected. Presence of DYNC1H1 in the P16 IPL persists in both peripheral and central retina (Fig 7B).
Extended survival of P16 peripheral retina is further demonstrated by IHC analysis of control and Dync1h1 F/F ;Six3Cre retina sections. ONL thickness of control central and superior retina is similar at P16 (Fig 8A-8D). The central Dync1h1 F/F ;Six3Cre retina is nearly completely degenerated at this age (Fig 8E and 8G), but the Dync1h1 F/F ;Six3Cre superior retina still forms ONL, INL and OPL layers (Fig 8F and 8H). DYNC1H1 is much reduced in the superior Dync1h1 F/F ;Six3Cre IS ( Fig 8F), and short OS are present as evidenced by presence of rhodopsin ( Fig 8H). Formation of OS in the retina periphery encouraged analysis of function by panretina ERG. Scotopic a-wave amplitudes are almost extinguished at P16 and later (Fig 8I). Flashes at 1.4 log cd s�m -2 and higher under photopic conditions elicit a b-wave response (20% of normal) (Fig 8J) suggesting cone survival at the periphery.

Discussion
In this communication, we truncated the dynein heavy chain after exon 23 removing the motor and the microtubule-binding domain. Truncation of the heavy chain is expected to eliminate the function of dynein, with drastic consequences on membrane protein trafficking. The fate of the N-terminal fragment encompassing exons 1-23 has no motor function and was not investigated for lack of N-terminal antibodies. However, since heterozygous animals are phenotypically normal on both ERG and immunohistochemistry, we assume the presence or absence of truncated protein does not affect the knockout phenotype. In control animals, DYNC1H1 is prominently expressed in photoreceptor IS, OPL and IPL, and to a lesser extent within the ONL and INL (Fig 3, row A). In conditional knockouts, DYNC1H1 could not be detected in the IS area but trace amounts persisted at the distal edge or within the nuclear layer (Fig 3, row B). These pockets also contained traces of rhodopsin, cone pigments, PDE6, EGFP-CETN2 labeled centrosomes and presumably other photoreceptor proteins (Fig 4). Photoreceptors do not elaborate IS, OS or functional synaptic terminals and degenerate in the first two postnatal weeks. Cones apparently form axons connecting to pedicles but are nonfunctional (Fig 6).
We studied the fate of two rod phototransduction proteins, rhodopsin and PDE6, in the Dync1h1 F/F ;Six3Cre retina. Rhodopsin is a G-protein-coupled receptor [38,39] that is synthesized at the rough ER surrounding the rod nuclei and known to depend on dynein for transport to the Golgi [5,40] following the conventional secretory pathway [3,41]. Transport of vesicles terminates near the basal body at the periciliary membrane [4] where cargo is assembled for transport through the connecting cilium. In P6 control retina, rhodopsin is located in the distal IS and is found in budding OS by P8 (Fig 4, row A). In the Dync1h1 F/F ;Six3Cre

PLOS ONE
Deletion of dynein heavy chain in retina retina, rhodopsin aggregated near centrosomes marked by EGFP-CETN2 at the sclerad edge of the ONL (Fig 4, rows B and B'). Unexpectedly, rhodopsin did not accumulate at the rod perinuclear rough ER at P6 suggesting suppression of opsin synthesis at a time when OS should form.
Rod and cone PDE6 are peripheral membrane proteins attached to membranes by C-terminal isoprenylation [42][43][44] and traffic to the OS by diffusion independent of dynein. Diffusion likely occurs via a complex with the prenyl binding protein PDE6D (PDEδ) [27,45]. In the mutant retina at P6, PDE6 is deposited at the ONL sclerad edge, presumably membrane bound (Fig 4, row D). At P8-P12, PDE6-positive cells are collapsing into foci rich in CETN2, centrosomes and presumably other soluble proteins. Interestingly, a few mother centrioles (basal bodies) appear to be able to engage in ciliogenesis and form diminutive membrane OSlike structures filled with PDE6. In zebrafish cnb and mok mutants, basal body docking was unaffected supporting this result [46]. As MOE antibody recognizes both rod and cone PDE6 it is not possible to distinguish rod and cone OS, but based on structural integrity of cones ( Fig  6) we interpret these structures to originate from cones. Dync1h1 F/F ;Six3Cre central rods are degenerated at P6 (Fig 4), Dync1h1 F/F ;Six3Cre cones still formed pedicles connected to nuclei by axons through P12 (Fig 6, row B).
Another major observation in the knockout model is the nuclear layer disorganization beginning at P6, before OS are formed, most likely caused by nuclear mispositioning. Active nuclear positioning in the retina takes place in the first two postnatal weeks and is managed by two motors, dynein and kinesin-2, walking towards the minus end and plus end of microtubules respectively [47]. In P4 WT retina, the neuroblastic layer is uniform and does not show an OPL with synaptic connections [48]. At P6, the Dync1h1 F/F ;Six3Cre OPL could not easily be distinguished (Figs 2-5). The nuclear layer was often interrupted suggesting that nuclei were not positioned correctly, a phenotype consistent with polarity defects [49][50][51]. Similar disorganization is observed when Syne2/Nesprin2 and Sun2, members of the LINC complex mediating nuclear migration during retina development, are deleted [36,52].
All retinal cell types arise from a pool of multipotent progenitors [53]. Expression of Cre beginning at E9 affects the entire developing retina and RPE by E12.5 [23]. In mouse, central retina neurogenesis completes by P5 and ciliogenesis begins at P6. We observed that control and Dync1h1 F/F ;Six3Cre basal bodies indeed formed a CC at P6, a rare event at this early stage, but at P8 ciliogenesis discontinued in Dync1h1 F/F ;Six3Cre retina (Fig 2E-2H). The 25% reduction in the neuronal layer at P6 (Fig 3E) suggests reduced neurogenesis and/or increased cell death during development in the absence of dynein. This reduction is consistent with the role of dynein in mitotic cell division such as spindle orientation and assembly, nuclear migration, chromosome segregation and cytokinesis (see review by [54,55]) as well as in neuronal survival [56]. Surprisingly, DYNC1H1 deletion did not affect cells in the retina uniformly. While the Dync1h1 gene is completely knocked out at P6 (Fig 1G), the mutant IPL continued to harbor DYNC1H1 polypeptides (Fig 3B and 3B') through P16 (Fig 7B) suggesting that inner retina may be much less effective than photoreceptors in eliminating "old" heavy chains synthesized before Dync1h1 gene knockout. We believe this accounts for the survival of inner retina neurons through P16 (Fig 5, row B, D).

Animals
All procedures were approved by the University of Utah Institutional Animal Care and Use Committee (Protocol 18-11005). Mice were sacrificed by cervical dislocation. For ERG studies, mice were anesthetized with intraperitoneal injection of 1% ketamine/0.1% xylazine at 10 μl/g body weight.

Generation of the Dync1h1 conditional allele
A cell line (Dync1h1 tm1a(KOMP) ) in which a gene trap was placed in intron 23, and a loxP site in intron 25 (Fig 1), was obtained from the UC-Davis KOMP repository. Chimeric mice were produced by the University of Utah Transgenic Gene Targeting Mouse Facility by blastocyst injection. Germline transmission of the Dync1h1 GT allele was verified by PCR using primers upstream and downstream of the loxP site in intron 25 (see genotyping paragraph). Founder mice were bred with flp-recombinase transgenic mice to create the Dync1h1 F allele. Mice carrying the Dync1h1 F allele were outbred to C57BL/6J mice to remove the rd8 mutation [57]. Dync1h1 F/F were crossed with Six3Cre transgenic mice (Stock No: 019755, The Jackson Laboratory) which were Egfp-Cetn2 positive (JAX 008234-CB6-Tg(CAG-EGFP/CETN2)3-4Jgg/J) [58] to generate Dync1h1 F/+ ;Six3Cre;Egfp-Cetn2 mice. Mice were then crossed back to Dync1h1 F/F to generate experimental animals. Expression of Egfp-Cetn2 allows centriole identification without use of a specific antibody. All transgenic mice (Six3Cre, flp, Egfp-Cetn2) were outbred routinely to C57BL/6J mice.

Electroretinography
Scotopic and photopic ERG was performed using P16 and P28 mice. Prior to ERG the mice were dark-adapted overnight and anesthetized with intraperitoneal injection of 1% ketamine/ 0.1% xylazine at 100 μl/g body weight. The mice were kept warm during ERG by using a temperature-controlled stage. Scotopic and photopic responses were recorded as described [59] using a UTAS BigShot Ganzfeld system (LKC Technologies, Gaithersburg, MD). Scotopic single-flash responses were recorded at stimulus intensities of -4.5 log cd s�m -2 (log candela seconds per square meter) to 2.4 log cd s�m -2 ). Mice were light-adapted under a background light of 1.48 log cd s�m -2 for 5-10 min prior to measuring photopic responses. Single-flash responses of control and knockout were recorded at stimulus intensities of -0.1 log cd s�m -2 to 1.9 log cd s�m -2 .

Eye collection methods
Mice were sacrificed by cervical dislocation. Eyes were enucleated and fixed by immersion in 4% paraformaldehyde in 0.1M phosphate buffer, pH 7.4, for 1 hour. Anterior segments were removed after 10 minutes of fixation. Eyecups were equilibrated in 15% sucrose in PBS, then equilibrated overnight in 30% sucrose in PBS, embedded in OCT, frozen on dry ice and stored at -80˚C. Blocks containing eyecups were equilibrated to -20˚C prior to sectioning. Transverse sections, including or adjacent to the optic nerve, were cut at 14μm using a Leica cryostat and transferred to charged slides (Thermo-Fisher). Slides were stored at -80˚C. Control retinas came from Dync1h1 F/F , Dync1h1 F/+ , or Dync1h1 F/+ ;Six3Cre mice.

Statistical analysis
We performed an unbalanced two-factor ANOVA to compare experimental and control animals for their quantified A-and B-wave ERG response across multiple ages. Post-hoc multiple comparison was performed using Tukey's honestly significant difference criterion. Statistical significance was determined using an alpha value of p < 0.05. All electroretinography statistics were computed using MATLAB's statistical toolbox "anovan" and "multcompare" functions.
We performed a Student's t-Test assuming unequal variances to compare control and KO central retinal thicknesses, measured just superior and inferior of the optic nerve on fixed retinal slices. Microsoft Excel function "T.TEST assuming unequal variances" was used to calculate the p-value, with statistical significance determined using an alpha value of p < 0.05.