Subretinal Delivery of Ultrathin Rigid-Elastic Cell Carriers Using a Metallic Shooter Instrument and Biodegradable Hydrogel Encapsulation

Abstract

Purpose.: To develop a surgical technique for the subretinal implantation of cell carriers suitable for the transplantation of cultured retinal pigment epithelium (RPE) in a preclinical animal model.

Methods.: Cell carriers were porous 10-μm-thick polyester membranes. A custom-made shooter instrument consisted of a 20-gauge metallic nozzle with a nonstick plunger. Fetal human RPE cultures were used for vitality assessment during instrument handling. Transvitreal subretinal implantation of carriers without RPE was performed in 31 rabbits after vitrectomy. Fourteen of 31 implants were encapsulated in gelatin. Fluid turbulence over the implantation site was minimized using a novel infusion cannula. Six rabbits had intravitreal plasmin injections before surgery. SD-OCT in vivo images were obtained after 3, 7, and 14 days, followed by perfusion-fixed histology.

Results.: Gelatin encapsulation of RPE/polyester implants made cell loss during handling reproducible, compared with 40% of controls showing random, large damage zones. Gelatin implants were ejected smoothly in 12 of 14 surgeries (86%), whereas “naked” implants frequently became trapped with the instrument, which reduced success to 9 of 17 cases (53%). Vitreous remnants after vitrectomy alone complicated subretinal placement of encapsulated and naked implants in 7 of 25 cases (28%). Plasmin-assisted vitrectomy resulted in implant ejection unperturbed by vitreous adhesions in six experiments. SD-OCT and histology demonstrated atraumatic subretinal implant delivery after uncomplicated surgery.

Conclusions.: A novel shooter instrument design allows for safe and atraumatic transvitreal delivery of hydrogel-encapsulated, ultrathin, rigid-elastic carriers into the subretinal space. The procedure may be used in the future to deliver cultured RPE.

Tissue-engineering strategies of the retinal pigment epithelium (RPE) could be beneficial to patients with retinal degeneration, including age-related macular degeneration. ^1,2 Current clinical protocols using an autologous patch of RPE/choroid have a number of limitations, among the foremost of which are senescence of the RPE/Bruch's membrane (BM) complex and surgical complications. Engineering the transplant in the laboratory could make use of stem cell–derived tissue. ³ A supportive cell carrier could enable easy surgical handling and long-term function. By contrast, cell suspensions alone may fail to adhere, survive, or function on an aged submacular Bruch's membrane. ^4,5 Allogeneic suspension transplants also carry a higher risk for immunologic rejection. ^6,7  

Various artificial carrier substrates for RPE replacement have been studied to date in vitro (see Refs. 1, 8 for reviews). However, in vivo validation, particularly in larger animals, is limited to three studies. ^{9 –11} Of those, Thumann et al. ¹¹ developed specialized instrumentation for an artificial biological cell carrier (collagen). 

One crucial aspect to ensure implant function is a safe, easy, and reproducible surgical delivery technique. Assuming a transvitreal approach will become the clinical method of choice, the following variables are among those that may need consideration in preclinical animal models: choice of cell carrier, instrument design, implant handling, cell viability, vitreous body as a harbinger of complications, iatrogenic retinal trauma, surgical learning curve for the implantation technique, potential for industrial scale-up, and fast yet accurate methods for postimplantation quality control. 

We chose a rigid-elastic carrier substrate because it affords safe handling and maintains its shape. We used commercially available ultrathin porous polyester as a prototype. Because RPE transplantation is an established treatment paradigm for RPE-related disease, we deliberately chose a healthy animal model (rabbit) to assess specific retinal changes incurred by surgical iatrogenic trauma. Given the size of a rabbit eye, we hypothesized that the information obtained would be relevant to eventual human procedures. Custom-designed instrumentation and ancillary techniques were developed for implantation of the cell carrier to the subretinal space, with subsequent analyses including video recordings, (in vivo) spectral domain optical coherence tomography (SD-OCT), and histology. In addition, the viability of fetal human RPE cultured on the polyester carrier was studied during instrument maneuvering with the implants. 

Materials and Methods

Implant Materials

Polyester membranes (PET, 10-μm thick; Transwell) were obtained from Corning Inc. (Corning, NY). The substrate is rigid elastic with 0.4-μm pores and was successfully used for RPE culture. ^12,13  

Acellular subretinal implants were generated from the PET materials with a custom trephine (Fig. 1B). Some membranes were encapsulated in sterile-filtered 15% porcine gelatin (Fig. 1D; Bloom Index 100; Sigma-Aldrich, Taufkirchen, Germany). Gelatin implants with and without penicillin/streptomycin (n = 20 each) were cultured on microbiologic culture plates and tested sterile according to German transplant law. 

Figure 1.

 

RPE cell culture and implant manufacturing exemplified here with commercially available, ultrathin polyester membranes. (A) Mature fetal human RPE after 6 weeks in culture on a porous polyester membrane. Inset: histologic (semithin) section of mature fhRPE on PET. Black arrows: artifacts (tears in resin from cutting). Scale bars: micrograph, 50 μm; inset, 25 μm. (B) Light micrograph of the naked bullet-shaped implant. Inset: scanning electron micrograph taken at 5000× magnification of randomly, widely spaced 0.4-μm pores and the smooth surface of the polyester membrane. Scale bar, 5 μm. (C) Overview on freshly trephined, RPE-seeded implant. Note only minimal cell loss at the margins. (D) Trephined, RPE-seeded implant, encapsulated in gelatin. (B–D) Implants are shown at the same magnification. (D) Scale bar, 200 μm.

Custom-Made Implant Shooter Instrument

The subretinal shooter instrument has been developed in two prototype variants (Figs. 2 3–4). The hand piece for both was taken from a model used clinically in vitreoretinal instruments. It is connected to a newly developed nozzle with a nonstick (Polytetrafluorethylen; MACEPLAST GmbH, Jüchen, Germany) plunger. The latter will push implants distally by a single-handed squeeze of the actuator on the hand piece. The 20-gauge nozzle is an oval, flattened, 0.1-mm-thick metal tube with drills on the upper or lower surface, or both (Figs. 2 3–4). 

Figure 2.

 

Common elements and measures of the shooter instrument. (A) Instrument length, 159.5 mm; maximum width at hand piece (plunger actuator), up to 17 mm. (B) Nozzle measurements: diameter, 1.0 mm; wall thickness, 0.1 mm; angle, 25°. (C) Total nozzle length, 38 mm; nozzle length from carriage to angle at nozzle orifice, 31 mm; carriage length, 8.2 mm. (D) Plunger actuator.

Figure 3.

 

Variant 1: shooter design with single opening. (A) Top view of oval, angled nozzle orifice with plunger in primary position (retracted). Outer width, 1.32 mm; inner width,: 1.1 mm. (B) Top view of nozzle orifice with plunger in maximal position. Diameter large drill (center), 0.5 mm; diameter small drill (distal end), 0.25 mm; plunger width, 0.8 mm. (C) Side view on angled, oval, flattened nozzle orifice. Angle, 25°; nozzle flattening from proximal 1.0 mm to distal 0.5 mm. Wall thickness, 0.1 mm. (D) Front view on nozzle orifice. Diameter inner height, 0.3 mm. (E) Bottom view with plunger in primary position. Drill diameters, 0.25 mm.

Figure 4.

 

Variant 2: shooter design separate loading port. (A) Top view of oval, flattened nozzle orifice with loading port and plunger in maximal position. Length of loading port, 2.45 mm; length from proximal end of loading port to lower (distal) end of nozzle orifice, 5.7 mm. (B) Side view on angled, oval, flattened nozzle orifice with plunger in maximal position. Plunger feed 0.15 mm beyond edge of nozzle orifice.

The first variant contains a single loading and ejecting orifice (Fig. 3). The second variant has a separate loading port on the side of the nozzle (Fig. 4). Through it, the implant is loaded and then pushed with the plunger under the bridge to enable transit until it is ejected subretinally through the orifice by further advancing the plunger (Fig. 4). 

RPE Cultures

One pair of fetal eyes at 22 weeks' gestation was obtained from Advanced Bioscience Resources (Alameda, CA). RPE cultures were established according to Hu and Bok. ¹⁴ Cells (2 × 10⁵/cm²) were seeded onto PET inserts and grown for at least 6 weeks. 

Cell-Seeded Implant Manufacturing

Retinal pigment epithelial implants were made from pigmented cultures when transepithelial electrical resistance, as measured by a volt ohm meter (EVOM; World Precision Instruments, Sarasota, FL), stabilized above 500 Ωcm² (Fig. 1A). Implants were trephined (Fig. 1C) and transferred onto sterile microscopic glass slides covered with a thin layer of 37°C gelatin. They were then chilled for 20 minutes to achieve hardening of the gelatin. Excess gelatin was removed with a prewarmed razor blade (Fig. 1D), and implants were loaded into the precooled shooter instrument and used immediately. 

Viability Testing of Cell-Seeded Implants

Observation of cell morphology and Trypan blue staining were performed on implants to assess cell loss during manufacturing and instrument handling. Implants were studied after gelatin encapsulation (Fig. 1D) and after ejection from the instrument with and without passage through a sclerotomy in a vitrectomized and balanced salt solution–perfused vitreous cavity of an anesthetized rabbit (intraocular pressure was set to 30 mm Hg). They were reassessed again after 24-hour storage in the CO₂ incubator (see Figs. 6, 7). Each variable was tested with 10 implants. 

Rabbits

Female chinchilla bastard rabbits weighing 2 to 2.5 kg were purchased from Charles River Laboratories (Sulzfeld, Germany). All procedures were approved by the state regulatory authorities of North Rhine-Westphalia (LANUV 8.87-50.1035.09.076 and 84-02.04.2011.A130) and in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Subretinal Implantation Procedure

Rabbits had their pupils dilated and were anesthetized by intramuscular injection of 0.65 mg/kg ketamine and 5 mg/kg xylazine. BVS (n = 20) and NE (n = 11) performed surgery on the rabbits (Table 1). Standardized, bare PET implants (n = 17) and PET encapsulated in gelatin (PETG; n = 14) were implanted without cultured RPE. One hour before implantation, six rabbits received an intravitreal injection with 1 U homologous rabbit plasmin in balanced salt solution. ¹⁵ Two-port, 23-gauge core vitrectomy (Megatron, Geuder, Germany) was performed through a noncontact, wide-angle system (BIOM II; Oculus, Wetzlar, Germany) with a 27-gauge chandelier illumination system (D.O.R.C., Berlin, Germany). A custom-made sutured 23-gauge infusion with two internal side ports and an enlarged floor plate ensured undisturbed implantation (Fig. 5). The retina was detached with a 41-gauge subretinal cannula (D.O.R.C.) with balanced salt solution, the retinotomy was enlarged with scissors and an implant was inserted subretinally using the shooter instrument. The sclerotomies and conjunctiva were sutured with 7-0 vicryl. An antibiotic/steroid eye ointment (dexamethasone+neomycin+polymyxin B; Alcon Pharma, Freiburg, Germany) was given twice daily for 1 week. 

Table 1.

 

Outcomes of Subretinal Carrier Substrate Implantation without Cultured RPE

Animal	Surgeon	Shooter Variant	Implantation
			Success*	Trapped†	Vitreous Incarceration‡	Plasmin
PET Alone
1	NE	2	Yes	No	No	No
2	NE	2	No	Yes	No	No
3	NE	2	No	Yes	Yes	No
4	NE	2	Yes	Yes	No	No
5	NE	2	Yes	Yes	Yes	No
6	BVS	2	No	Yes	No	No
7	BVS	1	Yes	No	No	No
8	BVS	1	Yes	No	No	No
9	BVS	1	Yes	Yes	No	No
10	NE	1	Yes	No	Yes	No
11	BVS	1	Yes	Yes	No	No
12	NE	1	No	Yes	No	No
13	BVS	1	Yes	No	No	No
14	BVS	1	Yes	No	No	No
15	NE	1	No	Yes	Yes	No
16	BVS	2	Yes	No	No	Yes
17	BVS	2	Yes	No	No	Yes
	BVS, 9	Variant 1 = 9
	NE, 8	Variant 2 = 8	12 of 17	9 of 17	4 of 17	2 of 17
PET + Gelatin
1	BVS	1	Yes	No	No	No
2	BVS	1	Yes	No	No	No
3	BVS	1	Yes	No	Yes	No
4	BVS	1	Yes	No	No	No
5	BVS	1	Yes	No	No	No
6	BVS	1	No	Yes	No	No
7	BVS	1	Yes	No	Yes	No
8	BVS	1	No	No	Yes	No
9	BVS	1	Yes	No	No	No
10	NE	1	Yes	No	No	No
11	BVS	2	Yes	No	No	Yes
12	BVS	2	No	Yes	No	Yes
13	NE	2	Yes	No	No	Yes
14	NE	2	Yes	No	No	Yes
	BVS, 11	Variant 1 = 10
	NE, 3	Variant 2 = 4	11 of 14	2 of 14	3 of 14	4 of 14

* Defined as successful eventual subretinal insertion on first attempt with shooter instrument, irrespective of intraoperative course or postoperative follow-up (see Supplementary Video S4).

† Defined as implant “trapped” within instrument lumen and/or compromised implant dislodging (see Supplementary Video S2).

‡ Defined as implant bouncing through the retinotomy and concomitant retinal folding (see Supplementary Video S3).

Figure 5.

 

Infusion cannula to minimize fluid turbulence over implantation site. (A) Overview on infusion cannula with intraocular side port (arrow) and enlarged floor plate. (B) With the infusion on, fluid will be directed sideways (arrow), thereby avoiding direct jet stream onto the posterior pole.

Surgical videos were taken with an analog camera connected to a digital video recorder (DVD-HR775A; Samsung, Seoul, South Korea) and were subsequently processed using Handbrake (open-source freeware, http://handbrake.fr/downloads.php) and commercial software (QuickTime and iMovie; Apple Inc., Cupertino, CA). Video analysis of the implantation success, defined as the eventual subretinal placement of the membrane with either shooter instrument variant, irrespective of intraoperative course or postoperative outcome, was carried out by a grader (RB) blinded to the time point, surgeon, implant type, and plasmin use. If the surgeon achieved eventual subretinal implantation with more than one attempt (e.g., if two instrument loadings were required), only the first was used for grading. 

Postoperative In Vivo Follow-up

Rabbits underwent SD-OCT with a commercially available device (Spectralis; Heidelberg Engineering, Heidelberg, Germany) at 3, 7 and 14 days after surgery. The OCT was recalibrated for rabbit eye optics according to manufacturer's instructions in reference to the standardized implant. 

The rabbits pupils were dilated, anesthetized intramuscularly, and held by an assistant. The cornea was moisturized with artificial tears (Optive; Pharm Allergan, Ettlingen, Germany). Subretinal implants were imaged at a 30° field-of-view in the IR-OCT mode. 

Histologic Processing

All animals were euthanized at 14 days after surgery by intracardial injection of 5 mL T61 in deep intramuscular anesthesia and were processed for histology by perfusion with 2% glutaraldehyde (GA). Eyes were then enucleated and placed in the same fixative for 24 hours at 4°C. Long-term fetal human RPE cultures were immersed in Karnovsky fixative (1:1) for 1 hour. Tissue samples were washed five times in 0.1 M Sørensen buffer, dehydrated in ethanol, embedded in Epon 812, and cured for 48 hours at 58°C. Semithin sections were cut at 1 to 2 μm and stained with toluidine blue. 

Results

In Vitro Viability Testing of RPE/PET Implants

Bullet-shaped RPE implants generated with a custom-made trephine showed continuous RPE monolayers, with cell densities comparable to those of original cultures (Fig. 1). These observations remained unchanged 24 hours after excision (Figs. 6, 7). Few scattered trypan blue–positive cell nuclei were observed throughout the monolayer immediately after punching (Fig. 7). 

Figure 6.

 

Viability of RPE/PET implants during preoperative handling: morphologic results. (A) RPE/PET implant stored for 24 hours at 37°C in a CO₂ incubator. (B) RPE/PET implant, after in vitro ejection, stored for 24 hours at 37°C in a CO₂ incubator. Note the large damage zone despite the undisturbed monolayer in other regions. (C) Appearance of RPE/PET implant after it was loaded into the shooter and passed through a sclerotomy in a vitrectomized and balanced salt solution–perfused vitreous cavity of an anesthetized (live) rabbit. (D) RPE/PET implant with transient gelatin encapsulation, thereafter stored in culture media for 24 hours at 37°C in a CO₂ incubator. (E) RPE/PET Implant with transient gelatin encapsulation and in vitro ejection, thereafter stored in culture media for 24 hours at 37°C in a CO₂ incubator. (F) Appearance of RPE/PET implant with transient gelatin encapsulation but otherwise treated identically to the implant in C. After 24 hours, peripheral zones of all implant condition variants show, to varying degrees, decreased cell densities with polygonal morphologies (compare, for example, insets in B and E); however, the central parts were largely unaffected. All insets show intact monolayers from respective conditions. All micrographs are shown at the same magnification. (F) Scale bars, 100 μm; 15 μm (insets).

Figure 7.

 

Viability testing of RPE/PET implants during preoperative handling: trypan blue. (A) Implant without gelatin, without shooter, stored at 37°C in CO₂ incubator for 24 hours. (B) Implant without gelatin, ejected from shooter, stored at 37°C in CO₂ incubator for 24 hours. (C) Implant without gelatin, after shooter and sclerotomy-pass, analogous to Figure 6C. (D) Implant after gelatin, without shooter, stored at 37°C in CO₂ incubator for 24 hours. (E) Implant after gelatin, ejected from shooter, stored at 37°C in CO₂ incubator for 24 hours. (F) Implant after gelatin, after shooter and sclerotomy-pass, analogous to Figure 6F. All micrographs are shown at the same magnification. (F) Scale bars, 100 μm; 15 μm (insets).

Cells encapsulated temporarily in gelatin (hydrogel) maintained their morphologic appearance compared with freshly trephined implants (Figs. 6A, 6D). Hydrogel coating reduced RPE viability compared with controls, as determined immediately after gelatin melting by higher amounts of trypan blue–positive cells in the implant periphery. These results correlated with lesser cell packing per area (density); that is, the cell bodies were larger and slightly elongated compared with controls after 24-hour storage in the CO₂ incubator (Fig. 6). 

Loading and ejection of RPE/PET implants without gelatin encapsulation resulted in random large damage zones to the cell monolayer (Fig. 6) caused by either the implant side of the RPE coming in contact with the inner wall of the instrument or the plunger sliding over the implant. The large damage zones were defined as >10% implant surface devoid of an RPE monolayer. Roughly 40% of all implant conditions without gelatin showed such damage zones, compared with only approximately 10% of hydrogel-encapsulated implants. Gelatin-embedded implants loaded and ejected with the shooter instrument showed a reproducible minimal cell loss both by morphology and by trypan blue viability assay (Figs. 6, 7). 

Thus, a temporary gelatin encapsulation of fetal RPE monolayers cultured on a rigid-elastic carrier substrate (polyester) appeared to induce some, yet acceptable, cell loss during implant handling compared with nonencapsulated controls with frequent random, larger damage zones. With the abovementioned combined, a temporary gelatin encapsulation resulted in more controlled implant manipulation. 

Subretinal Implantation of PET Implants in Rabbits

Subretinal implantation of PET without cultured RPE was performed in 31 rabbits to establish an atraumatic surgical technique and to demonstrate the safety of subretinal implantation (Table 1). 

Implantation success on the first attempt was achieved overall in 23 cases (74.2%). When stratified by implant type, PET membranes were successfully implanted in 12 of 17 cases (70.6%), whereas the PETGs could be placed subretinally in 11 of 14 procedures (78.6%). In a subseries of six animals with intravitreal plasmin pretreatment, 5 of 6 membranes were inserted successfully into the subretinal space. 

Using a conventional straight intraocular infusion cannula resulted in collapsing of the bleb retinal detachment as a consequence of the directed jet stream (Supplementary Video S1). A custom-made 23-gauge infusion with two intraocular side ports eliminated the problem (Fig. 5). A large suture-on floor plate proved advantageous for the thin rabbit sclera. 

Thorough core vitrectomy over the implantation site was necessary to ensure smooth subretinal implantation. Even so, mechanical removal of the cortical vitreous appeared inconsistent because PET implant ejection from the instrument was compromised in 10 of 15 cases with vitrectomy alone (PET without plasmin; Table 1). 

In some instances, the membranes would slide underneath the plunger while they were still partially in the instrument lumen. It was then challenging to eject the implant entirely (Table 1). In other instances, when vitreous adhesions were negligible, we observed the implant to remain adherent to the instrument tip when ejected (Supplementary Video S2). Dry laboratory experiments suggested electrostatic adhesions between the metallic instrument and the ultrathin polyester material to be involved in this phenomenon. Both phenomena were classified as “trapped with the instrument.” To completely dislodge (place) the membrane, additional maneuvering was necessary in 9 of 15 instances with PET, in stark contrast to 1 of 10 with PETG implants (Table 1). In the former, this resulted in damage to the neural retina. 

One problem remaining even with gelatin coating was that just subretinally placed membranes would “jump” through the retinotomy again, suggesting persistent vitreous adhesions (Supplementary Video S3). Based on surgical notes and subsequent video analysis this occurred in 7 of 25 cases (PET and PETG) (Table 1). 

In preliminary experiments (n = 6) with enzyme-assisted vitrectomy, we observed a smooth continuous ejection of the bullet-shaped implants in proper orientation directed along the longitudinal axis of the instrument (Supplementary Video S4). In one plasmin case, the implant was trapped with the instrument, perhaps because of premature melting of the gelatin. 

In most instances the implants were repositioned with another instrument to achieve a location more distant from the retinotomy. All the observations described seemed surgeon independent. 

Postoperative In Vivo Imaging of Acellular Subretinal Implants

After uncomplicated surgery, the neural retina overlying the implants was attached by day 4 in SD-OCT (Fig. 8). The retina surrounding the retinotomy site was initially everted but flattened with further follow-up. Retinal and outer nuclear layer (ONL) thinning on SD-OCT was maximal after 7 days. Gelatin implants showed a diminishing subimplant hyporeflective band on SD-OCT over follow-up times. A hyperreflective band on SD-OCT was discernible above all implants by 14 days. The neural retina adjacent to the implant site showed normal reflexion bands (Fig. 8). Follow-up in complicated implantations resulted in distorted retinal reflexion bands or variable overall retinal thickness, suggesting surgical trauma (data not shown). These results suggest atraumatic delivery and subsequent tolerance of subretinal acellular polyester membranes after uncomplicated surgery. 

Figure 8.

 

Follow-up of subretinal implants without cultured RPE with SD-OCT and histology. Left: PET implants (red arrow) without encapsulation over 14 days. Right: gelatin-encapsulated PET implants. Note the hyporeflective bands in SD-OCT (white arrows, right column) in gelatin implants, which appear to correlate with the large clear vacuoles seen on histologic examination. Black arrows: resin folds; processing artifacts.

Histology of Acellular Subretinal Implants

Likely as a result of surgical manipulation, the region around the retinotomy site showed marked subretinal scarring and ONL atrophy (Figs. 9D, 9H). By contrast, regions adjacent to the distal end of the implant showed normal and adherent neural retinal layering (Figs. 9B, 9F). RPE cells seemed to migrate over the implant edges but failed to produce a continuous epithelial monolayer between the implant and the photoreceptors. Rather, most of the implant was covered with a multilayered fibrocellular scar along with concomitant photoreceptor and ONL atrophy. This resulted in marked thinning of the neural retina overlying the implant. Occasional pigmented granules were observed within both scar and retina. 

Figure 9.

 

Implantation of PET alone after mechanical-only vitrectomy versus gelatin-encapsulated PET after plasmin-assisted vitrectomy. Two representative cases are given at high and low magnification. (A, E, red arrows) Marked subretinal scarring and loss of ONL around the retinotomy site and (D, H) at higher magnification. (A, B, E, F) By contrast, the photoreceptors and ONL thickness are preserved until close proximity to the distal end of the implant. (B, F) Pigmented epithelioid cells (presumably RPE) cover the distal end of the implant, suggesting sub-RPE implantation, RPE wound healing, or both. (C, G) In the middle portions of the implant, the overlying retina typically shows an ONL reduced to 0 to 1 cell layer, with a multilayered-fibrocellular scar filling the space between the PET membrane and the outer limiting membrane. Occasional pigment granules are observed within both the subretinal and the intraretinal scar. Vacuoles are observed underneath the implant and around the retinotomy site and are more prominent in gelatin-encapsulated implants. Differences in toluidine blue staining intensity between (A–D) and (E–H) are the result of varied dye incubation times and are, therefore, insignificant. (red arrowheads) Resin folds/tears that represent processing artifacts restricted to the implant region only. Scale bars, 250 μm (A, E); 50 μm (D, H).

Bruch's membrane appeared contiguous underneath the implants, with reactive changes in the RPE. More vacuoles were observed under PETG than under PET, suggesting correlation with hyporeflective bands seen under PETG on SD-OCT (Figs. 8, 9). In addition, these were observed around the retinotomy site, again more prominent in PETG. Choriocapillaris seemed to remain patent by 14 days after surgery, as judged by occasionally observed erythrocytes in small vessel lumina (of the perfusion-fixed tissue). 

Inner retinal layers were continuous, though slightly thinner, in both implant types in uncomplicated surgeries (Figs. 9C, 9G). However, the innermost retinal layers above the implant displayed increased cystic spaces that were not present in the adjacent normal retina (Fig. 8). 

Histologic sections at 2 weeks after surgery did not suggest signs of inflammation or overt toxicity to the retina or choroid. Histology results further support the SD-OCT observations of a feasible atraumatic delivery and tolerance of polyester membranes with or without gelatin in the subretinal space. Moreover, SD-OCT and histology images showed good correlation. 

Discussion

With a shooter instrument design, we developed a safe and atraumatic method for transvitreal delivery of ultrathin rigid-elastic cell carrier materials into the subretinal space. Although the method has a relatively short surgical learning curve, its success depends on a clean vitreoretinal interface, an intraocular infusion that avoids fluid turbulence over the implantation site, and a temporary hydrogel encapsulation for cell protection and smooth subretinal implantation. The procedures described herein are limited, however, to the mere subretinal delivery of potential cell carriers. Successful coimplantation of cultured cells on carrier matrices may require additional measures, such as serum-free RPE culture methods, better characterization of RPE viability on cell carriers, removal of the host RPE layer, strategies for implant anchorage, biodegradable carriers, and immunosuppression. 

A number of investigators have presented methods for transvitreal subretinal delivery of RPE cultured on carrier supports. ^{9 –11} Of those, Thumann et al. ¹⁶ described an elegant, custom-made, hollow metallic spatula with a protective sleeve, similar to a concept brought forward by Steinhorst et al. ¹⁷ Interactions of metallic and artificial surfaces can create electrostatic charge, and implant dislodging with the proposed jet stream of Thumann et al. ¹⁶ may cause uncontrolled implant movements. Electrostatic interactions were also observed in our study but were eliminated with gelatin encapsulation. 

In related approaches, instrumentation for subretinal implant delivery has been developed for alginate-encapsulated, uncultured fetal retinal tissue by Aramant et al. ¹⁸ Their patented instrument and method differs in design from ours because it places rather than pushes the implant to its subretinal target site. The instrument nozzle is retracted relative to the mandrel. It requires that the surgeon's hand remain perfectly still during subretinal maneuvering. The same is true for the design of Thumann et al. ¹⁶ Montezuma et al. ¹⁹ described a subretinal inserter instrument for the delivery of subretinal chip implants in pigs but did not disclose any detailed instrument design or handling. Compared to the aforementioned groups, our technique allows implant delivery from epiretinal and, therefore, affords visual control over the instrument and is forgiving of minor movements. This, in our view, enables a faster surgical learning curve and represents a safer technique. On the downside, however, the implant may thereby become entrapped by vitreous. 

Enzymatic vitreolysis has not been studied as yet to aid the delivery of subretinal implants. ^20,21 Incomplete vitreous removal by conventional vitrectomy compared with enzyme-assisted vitrectomy might in part explain the high rates of proliferative vitreoretinopathy (PVR) in patients receiving autologous RPE/choroid patch grafts. ²² Moreover, all transvitreal subretinal implant techniques to date are likely to be compromised by vitreal adhesions, and transscleral approaches are not practical for macular surgery. During intraocular maneuvering with RPE transplants, cells are unavoidably dislodged, and their contact with vitreous induces PVR. ²³ A clean vitreoretinal interface, therefore, not only facilitates subretinal delivery, as shown in this study, but may also reduce long-term complications associated with RPE transplantation techniques. 

The use of an intraocular side-port infusion cannula reduced the jet stream over the implantation region, thereby avoiding collapse of the bleb detachment and uncontrolled tearing of the retinotomy. To the best of our knowledge, this is the first description of such an instrument to facilitate subretinal procedures. A related design was described by Bernd Kirchhof for silicone oil removal (http://www.eyemoviepedia.com/videos/1251861393/104). 

Current cell carrier biomaterials are available in biodegradable or biostable forms. We favored biostable polyester membranes because they have well-established cell growth characteristics and maintain a stable chemical composition once implanted. Both features are critical for long-term transplant survival on an aged Bruch's membrane and unperturbed retinal function. One potential shortcoming of the biostable approach may be pore clogging over time, a phenomenon we observed in our RPE cultures older than 4 to 6 weeks, with the appearance of small RPE detachments. By contrast, derivatives and elutes from biodegradable materials may affect both transplant function and neighboring structures. This may require long-term animal model toxicity screens because some of the currently favored materials (e.g., polycaprolactone/PCL) begin to degrade after years. 

Gelatin for subretinal implants was introduced by Silverman. ²⁴ Tezel ²⁵ used it in clinical trials with homologous adult RPE transplants. Some studies suggested that both concentration and bloom index play a role in inducing subretinal inflammation. ^{26 –29} Accounting for that, we chose a low bloom gelatin with 15% concentration for the encapsulation of our implants. Compared with controls, we did not observe signs of subretinal inflammation on SD-OCT. Histology might be more sensitive to exclude inflammatory cell infiltration; this is the subject of ongoing work. Although in vitro assays had shown the dissolution of gelatin within seconds, OCT and histology data suggested some persistent gelatin underneath the implant and around the retinotomy site. This may indicate that cooling of the shooter instrument before the procedure delayed melting of the gelatin. 

A temporary (gelatin) hydrogel encapsulation proved to be of significant advantage for both loading and unloading of the cell-seeded carrier substrate from the shooter instrument. However, further work is necessary to evaluate the effects of gelatin encapsulation on RPE junction integrity and polarization and on the long-term effects on subretinal cellularized implants. Its animal origin may pose a hurdle for regulatory approval; hence, the development of synthetic hydrogels is desirable. ³⁰  

Localized retinal detachments have been demonstrated by Marmor et al. ^31,32 to resolve spontaneously in rabbits. Although the mechanical ablation of the neural retina was clearly shown to induce RPE and photoreceptor damage in rabbits, ³³ RPE wound healing and, more important, a relatively preserved ONL thickness after reattachment suggest that the RPE/photoreceptor interface can in part compensate such insults. ³⁴ Indirect evidence for this hypothesis may also be derived from subthreshold retinal laser burns in rabbits, which show almost indistinguishable lesion restitution after relatively selective damage to RPE and photoreceptors. ³⁵ Given that the typical area of bleb retinal detachment was significantly larger than the size of the implant, this would explain why outer retinal reflectivity bands on SD-OCT, as well as outer retinal histoarchitecture, appeared fairly normal until they were in close proximity to the implant. 

Although atraumatic delivery of the subretinal implants was feasible, as judged for example by the intact inner retinal layering, we observed almost complete atrophy of the ONL above the implant on OCT and histologic examination. This degenerative response was expected because photoreceptors lost their trophic support from the RPE. ^36,37 Similar atrophic ONL changes were reported after subretinal implantation of artificial retina chips. ^19,38,39 Peachey ⁴⁰ reported the decreased survival of photoreceptors overlying solid implants, prompting several groups to suggest porous implants as more permissive on photoreceptor survival. In a recent work, Julien et al. ³⁸ found gelatin coating of porous polyimide (PI) to improve permeability in diffusion assays, but a relative effect on photoreceptor survival is unclear because uncoated PI implantation data were not presented. Ongoing experiments by our group with another, more permeable carrier material suggest better preserved outer retinal architecture. Taken together, ONL preservation above subretinal implants may depend not only on the presence of a functioning RPE layer but also on adequate permeability of the cell carrier. 

Multilayered scarring around implants was observed on histologic examination and likely corresponded with the hyperreflective layer above the implant on SD-OCT. This encapsulation reaction probably represents subretinal wound healing and has been reported by others. ^19,41,42  

With the derivation of RPE-like cells from several stem cell sources, ^{7,43 –46} RPE replacement strategies are in dire need of a safe technique for delivering tissue-engineered constructs to the subretinal space. The techniques presented here may provide some solutions to this end. Moreover, photoreceptor precursors on rigid carriers, ^47,48 sustained drug-release devices, or Bruch's membrane prosthetics that stimulate RPE wound healing could also be implanted with our method. 

Supplementary Materials

Movie sm1, MOV - Movie sm1, MOV 

Movie sm2, MOV - Movie sm2, MOV 

Movie sm3, MOV - Movie sm3, MOV 

Movie sm4, MOV - Movie sm4, MOV 

Footnotes

 Supported by Rüdiger Foundation grants (BVS), BONFOR/Gerok Scholarship O-137.0015 (BVS), and State Scholarship Fund/China Scholarship Council 2008627116 (ZL).

Footnotes

 Disclosure: B.V. Stanzel, Geuder AG (F), P; Z. Liu, Geuder AG (F), P; R. Brinken, Geuder AG (F), P; N.T Braun, Geuder AG (F, E), P; F.G. Holz, Heidelberg Engineering (C), Geuder AG (F), P; N. Eter, Geuder AG (F), P

The authors thank Claudine Strack for technical assistance with the histology and Ernst Molitor (Department of Medical Microbiology, University of Bonn, Germany) for sterility testing of gelatin-coated implants.