Effects of intravitreal injection of siRNA against caspase-2 on retinal and optic nerve degeneration in air blast induced ocular trauma

Ocular repeated air blast injuries occur from low overpressure blast wave exposure, which are often repeated and in quick succession. We have shown that caspase-2 caused the death of retinal ganglion cells (RGC) after blunt ocular trauma. Here, we investigated if caspase-2 also mediates RGC apoptosis in a mouse model of air blast induced indirect traumatic optic neuropathy (b-ITON). C57BL/6 mice were exposed to repeated blasts of overpressure air (3 × 2 × 15 psi) and intravitreal injections of siRNA against caspase-2 (siCASP2) or against a control enhanced green fluorescent protein (siEGFP) at either 5 h after the first 2 × 15 psi (“post-blast”) or 48 h before the first blast exposure (“pre-blast”) and repeated every 7 days. RGC counts were unaffected by the b-ITON or intravitreal injections, despite increased degenerating ON axons, even in siCASP2 “post-blast” injection groups. Degenerating ON axons remained at sham levels after b-ITON and intravitreal siCASP2 “pre-blast” injections, but with less degenerating axons in siCASP2 compared to siEGFP-treated eyes. Intravitreal injections “post-blast” caused greater vitreous inflammation, potentiated by siCASP2, with less in “pre-blast” injected eyes, which was abrogated by siCASP2. We conclude that intravitreal injection timing after ocular trauma induced variable retinal and ON pathology, undermining our candidate neuroprotective therapy, siCASP2.


Materials and methods
Experimental design. This study investigates whether caspase-2 promotes RGC death and ON axonal pathology in a b-ITON mouse model. We also considered the importance of intravitreal (invit) injection timing into an injured eye. Caspase-2 was knocked down by intravitreal injections of siCASP2 (2 μl of 1 μg/μl solution in sterile PBS) or equal concentration of siRNA against enhanced green fluorescent protein (siEGFP) as a control. In our first experiment, siCASP2 and siEGFP were intravitreally injected 5 h after the initial 2 × 15 psi blast and followed by two further blast waves, this is referred to as the "post-blast" injection study (Fig. 1A,B). Mice in this group were euthanized at 28 days post injury (dpi). We also performed injections 48 h before b-ITON, this is referred to as the "pre-blast injection" study ( Fig. 1C,D). Mice in this group were euthanized at 14 dpi. Optical coherence tomography (OCT) imaging was performed bilaterally at baseline and immediately before mice were euthanized (Fig. 1E). Mice in the "post-blast" study were euthanized at 28 dpi, eyes processed for immunohistochemistry (IHC), and RGC positive for RNA-binding protein with multiple splicing (RBPMS), a specific cytoplasmic RGC marker 51,52 , were quantified on retinal whole mounts (Fig. 1E). Mice in the "pre-blast" injection study were euthanized at 14 dpi and eyes processed for immunohistochemistry (IHC) and RBPMS + RGC quantified in retinal cryosections, as previously described 48 . In both groups, the far proximal ON tissue was processed for resin semi-thin cross-sections and PPD-stained intact and degenerating ON axons were quantified. The remaining ON were processed as longitudinal cryosections for IHC analysis (Fig. 1E).
Animal care and procedures. 12-week-old male C57BL/6 mice purchased from Jackson Laboratory (Bar Harbor, Maine, USA) were used in this study. Animal procedures were approved by the Institutional Animal Care and Use Committee of Vanderbilt University and conformed to the Association for Assessment and Accreditation of Laboratory Animal Care guidelines and conducted in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. The study was carried out in compliance with the ARRIVE guidelines. Animals were randomly assigned, and the experimenters masked to the treatment and procedural conditions. All procedures and investigations were performed between 07:00 and 12:00.
Air blast-induced indirect optic neuropathy (b-ITON) injury mouse model. Animals were anaesthetised using 3% isoflurane in 1.8 l/min of O 2 and male 12-week-old C57BL/6 mice were exposed to a blast overpressure wave produced by a device modified to produce an air blast wave, as previously described [40][41][42]47,48 . Repeated air blast wave injury was chosen to more closely approximate real-world blast injury from linked mines or to mimic the multiple blast wave exposures and blunt-force injuries that occur during a single large explosive blast event. We used a 15 psi air blast, which alone does not cause pathology 42 . The average of 8 air blast tests shows that the system induces peak pressure at the location of the eye, when the mouse is in the system, at 4 ms post-trigger, stays elevated for 1 ms, and returns to baseline by 9 ms. This repeated paradigm causes greater RGC axonal degeneration compared to a single 26 psi air blast wave exposure. The left eye of mice was exposed to two 15 psi air blast waves with an interblast interval of ~ 0.5 s, repeated for 3 consecutive days for a total of 6  42 . The mouse eye was positioned 162 mm from the end of the device. Separate mice were exposed to equivalent procedures excluding the air blast wave which was blocked and verified to deliver a pressure of < 2 psi, and did not receive intravitreal injections. A pressure transducer recorded the air blast overpressure wave, which was viewed using LabVIEW software (National Instrument Austin, TX, USA). GenTeal Tears (Alcon, Novartis, Bilateral siCASP2 or siEGFP control were intravitreally injection 5 h after the initial 2 × 15 psi blast wave, and injections repeated every 7 days until perfusion and tissue collection at 28 dpi. (B) Experimental groups for "post-blast" injection study. (C) Timeline for "pre-blast" study. Bilateral siCASP2 or siEGFP controls were intravitreally injected 48 h before the initial 2 × 15 psi blast exposure and injections repeated every 7 d until perfusion and tissue collection at 14 dpi. (D) Experimental groups for "pre-blast" study. (E) Measured endpoints and eyes analysed. www.nature.com/scientificreports/ Fort Worth, Texas, USA). Eye drops were applied after the air blast waves to prevent corneal dehydration from anaesthetic exposure and the mice were allowed to recover fully. Intravitreal injections were performed using a 31-gauge   needle with a bevelled tip attached to a 10 μl Gastight Syringe (Hamilton, Reno, NV, USA) under inhalational  anaesthetic of 2-3% isoflurane at a 45° angle 1 mm peripheral to the limbus and the lens was avoided. Unilateral  b-ITON was performed and 2 µl of 1 μg/μl siCASP2 or siEGFP (provided by Quark Pharmaceuticals Inc. under a Material Transfer Agreement) as control administered by bilateral intravitreal injection. Full details of siCASP2 and all modifications to the sequence are detailed as described previously 16,26 . Briefly, siCASP2 is a naked RNA duplex with chemical modifications to prevent degradation by vitreal and serum nucleases. In the first "postblast" injection study, siRNA injections were performed 5 h after the initial blast wave and repeated every 7 days until killing and tissue collection at 28 dpi (n = 10 per group) (Fig. 1B). In the second "pre-blast" injection study, injections were performed 48 h before the b-ITON (n = 5 per group) and repeated every 7 days until euthanasia and tissue collection at 14 dpi (Fig. 1D). Animals were perfused under terminal anaesthesia as described below.

Intravitreal injection of siCASP2 and siEGFP.
Optical coherence tomography (OCT) imaging, retinal thickness and vitreous haze analysis. OCT scans were performed under anaesthesia (3% isoflurane in O 2 ) at 27 dpi in the post-blast group and 13 days in the pre-blast group, to construct a high resolution cross-sectional retinal image using a Bioptigen ultra-high-resolution SD-OCT system with a mouse retinal bore (Bioptigen, North Carolina, USA). Pupils were dilated using 1% tropicamide and GenTeal™ lubricant gel was used to maintain corneal clarity and prevent drying. All images were acquired with the same level of A-scan averaging (100 averages per A scan) and with the retinal position central to the image. A total of 2 B-scans were analysed per eye either side of the ON head. Whole retinal thickness and ganglion cell complex (GCC) thickness (ganglion cell layer, GCL, and inner plexiform layer, IPL) were measured in OCT images in line with the optic nerve head (ONH). Image J was used to manually segment the layers and measure the area, which was divided by the length of the retinal segment measured to calculate the layer thickness. Analysis to quantify vitreous inflammation was performed using Image J (http:// rsbweb. nih. gov/ ij), based on the method previously described 53 : two images either side of the ONH were analysed per eye and the pixel intensity in five regions of interest in the vitreous were measured and then displayed as a percentage of the average of retinal pigment epithelium (RPE) intensity. Results are displayed for "pre-blast" and "post-blast" studies and we have also grouped siCASP2 and siEGFP injections from the "preblast" and "post-blast" intravitreal injection groups and compared to sham and no invit eyes to determine if intravitreal injections cause changes in vitreal inflammation.
Tissue preparation for IHC. Animals were euthanized by overdose of anaesthetic (Isofluorane) and intracardially perfused with 4% EM-grade paraformaldehyde (PFA; Electron Microscopy Sciences, Hatfield, Pennsylvania, USA) dissolved in phosphate buffered saline (PBS). Eyes were then cryoprotected in ascending concentrations of sucrose (10%, 20%, 30%) and embedded in optimal cutting temperature compound. Sections were cut at 15 µm-thick using a cryostat (Bright Instruments, Huntingdon, UK), collected onto SuperFrost (Fisher Scientific, Loughborough, UK) coated glass slides and stored at − 20 °C until required for IHC. Whole retinae were dissected out of the eyes and IHC staining performed in 24-well plates, as described below.

Immunohistochemistry (IHC) in retinal frozen sections.
Frozen cryosections were thawed for 20 min, washed in several changes of PBS, permeabilised and non-specific binding sites blocked by incubation in PBS containing 1% Triton-X-100 (Sigma) and 3% bovine serum albumin (BSA; Sigma). Sections were then incubated overnight at 4 °C with the appropriate primary antibody (Table 1) before washing in several changes of PBS and incubating with Alexa Fluor 488/594 or HRP-labelled secondary antibodies (Table 1) for 1 h at room temperature (RT). Finally, sections were washed in PBS and coverslips mounted in a Vectashield antifade aqueous mounting medium with 4′,6-diamidino-2-phenylindole (DAPI) nuclear stain (Vector Laboratories, Peterborough, UK). Controls, which included omission of primary antibody were included in each run and were used to set the background threshold levels prior to image capture.

IHC in retinal wholemounts.
Retinal wholemounts were stained with appropriate primary antibodies after permeabilization in PBS containing 0.5% Triton X-100 (Sigma). To aid permeabilization, this step was first performed at − 80 °C for 15 min, followed by thawing in 0.5% Triton X-100 for 15 min at RT. Primary antibodies were diluted in PBS containing 2% Triton X-100 and 2% BSA (all from Sigma), added to wholemounts and incubated overnight at 4 °C. Wholemounts were washed in several changes of PBS, incubated with appropriate secondary antibody for 2 h at RT, washed in PBS and coverslips mounted using Vectashield aqueous mounting media containing DAPI. Primary antibodies were omitted in controls and were used to set the background threshold levels. Wholemounts were viewed under an Axioplan 2 fluorescent microscope equipped with an Axi-oCam HRc and running Axiovision Software (All from Zeiss, Hertfordshire, UK). An experimenter masked to the treatment conditions captured representative images from each wholemount for analysis.
Assessment of RGC survival. The number of RBPMS + immunostained RGC were quantified. In the "post-blast injection" study, RBPMS + RGC were counted in the middle portion of the retinal whole mounts ( Supplementary Fig. S1A) and displayed as mean RBPMS + RGC per mm 2 . In the "pre-blast injection" study, RBPMS + RGC in the GCL were counted on retinal cryosections in line with the ONH in pre-determined areas  S1B) and displayed as mean RBPMS + RGC/mm of retina, as previously described 48 . Four images were analysed from two cryosections per animal.

Counting of intact and degenerative ON axonal profiles. The total ON area and the number of ON
axons with intact and degenerative profiles were quantified, as described by us previously 48 . Briefly, the circumference of the ON was measured and the total ON area calculated using Image J (NIH, Maryland, Bethesda, USA). The numbers of intact and degenerating (arrows Fig. 2D) ON axons were counted manually by assessing the morphological characteristics of PPD stained axons. Intact axons had uniform myelin whereas degeneration axons had unravelling or collapsed myelin. ON axons in 9 boxes in the Image J Counting Grid plug in were analysed and the total number of axons in the whole ON extrapolated.

Statistical analysis.
Statistical analysis was carried out using SPSS version 26 (IBM Corp, Armonk, NY, USA) and GraphPad Prism version 7.00 (GraphPad Software, La Jolla, CA, USA). The data were tested for normality using the Shapiro-Wilk test. Normally distributed data without linkage were analysed using one-way ANOVA and post-hoc Tukey tests with P values corrected for multiple comparisons (RBPMS + RGC quantification, ON axon counts, ED1 + cells in ON, retinal thickness in OCT scans). Data including measurements from two eyes of each animal (vitreal intensity in OCT scans) were modelled using generalised estimating equations (GEE; normal distribution with identity link function and independent correlation matrix). To construct a model to fit the data, all available factors were included in the initial model with 2-way interaction terms and then terms with P > 0.05 were serially removed from the model, starting with the least significant and interaction terms. In experiments where multiple comparisons were made, the P values were corrected with Bonferroni correction. Data is reported as mean ± standard error of the mean (SEM). Sample size was based on previous studies demonstrating that n = 5 animals per group detected treatment effects on axon counts 40,48 . No animals were excluded or euthanized due to reaching humane end points before study completion.

Results
The number of degenerative ON axons after b-ITON was not affected by "post-blast" siCASP2 intravitreal injections. We first determined the effects of caspase-2 knockdown in b-ITON on RGC degeneration, ON axonal pathology and vitreal inflammation. To do this, we performed "post-blast" intravitreal injections of our well-characterised siCASP2 16,17,22,23 or a control siEGFP, and then assessed ON axonal morphology, quantified the number of degenerating and intact axons as well as axon density on PPD-stained semi-thin proximal ON cross-sections ( Fig. 2A-D). There was no difference in total axonal numbers between sham and b-ITON without injections at 28 dpi (38,193

Infiltration of ED1 + cells in the ON by 28 dpi. To assess infiltrating inflammatory cells in the ON after
b-ITON, we counted ED1 + cells in longitudinal ON cryosections (Fig. 3E,F). ED1 is a widely used monoclonal antibody clone which is directed against CD68 and is used to identify macrophages, monocytes and activated microglia in rat tissues. While there appeared to be more ED1 + cells in the ON after b-ITON, the numbers did not reach statistical significance in any group. For example, there were 2.09 ± 0.68 ED1 + cells per mm 2 were present in sham ONs, we detected 10.11 ± 3.69, 21.79 ± 13.21, and 28.73 ± 12.21 ED1 + cells per mm 2 in the uninjected b-ITON, "post-blast" siEGFP and siCASP2 injected b-ITON ONs, respectively (P = 0.2255, ANOVA). These results demonstrate that markers of ON and retinal inflammation, including ED1 + macrophage-infiltration and vitreous haze increased in concert after b-ITON followed by intravitreal injection, with a possible additive effect when the intravitreal injection of siCASP2 knocked down caspase-2 levels.
ON axonal degeneration and RBPMS + RGC after "pre-blast" siCASP2 and siEGFP intravitreal injections. The study detailed thus far was performed with siCASP2 or siEGFP intravitreal injections at 5 h after the first 2 × 15 psi blast exposure, into a potentially vulnerable and traumatically inflamed eye, which we have demonstrated caused vitreous and retinal inflammation. To determine if intraocular injections at this time point were potentially detrimental to the eye, we next performed a further study in which we injected siRNA at 48 h before the b-ITON, "pre-blast" injection. Pre-injury siCASP2 intravitreal injection is unlikely to be clinically translatable, however, we aimed to derive information on pre-injury caspase-2 knockdown and the mechanistic role of caspase-2 in the b-ITON model. Therefore, we injected siCASP2 and siEGFP at 48 h before b-ITON and continued the study for 14 dpi (n = 5 mice per group).
Immunostained ED1 + cells in the retina were more frequently localised in the outer plexiform layer (OPL), IPL and GCL in b-ITON with siEGFP and siCASP2 injections compared to sham and b-ITON with no injections, suggesting increased inflammatory cell infiltration was caused by "pre-blast" intravitreal injections (Fig. 5F). GFAP immunostaining levels remained constant between all groups (Fig. 5F). Consistent with our findings from "post-blast" intravitreal injections, there was a trend for an increase in the number of ED1 + cells in the ON of mice receiving b-ITON compared to sham controls, however it did not reach statistical significance (8.40 ± 2.20 vs 0.75 ± 0.17 ED1 + cells per mm 2 ; P = 0.075, ANOVA; Fig. 5G,H) due to the low numbers and significant variation. The number of infiltrating ED1 + cells were also high in b-ITON eyes injected with siEGFP and those injected with siCASP2 (12.96 ± 3.82 vs 9.77 ± 3.37 ED1 + cells per mm 2 ; P = 0.5586 and P = 0.9742 post-hoc Tukey, both compared to b-ITON with no intravitreal injections), suggesting increased inflammatory cell infiltration was caused by blast in these groups (Fig. 5F-H). These results suggest that intravitreal injections administered 48 h before b-ITON caused less retinal and vitreous inflammation than when given 5 h after b-ITON.

Discussion
In this study we show differential responses to intravitreal injections given 48 h before compared with 5 h after b-ITON, with the latter causing retinal and ON inflammation compared to sham controls and b-ITON alone. Receiving "post-blast" siCASP2 intravitreal injections did not affect the numbers of degenerating ON axons or the number of intact axons. In contrast, "pre-blast" siCASP2 intravitreal injections resulted in fewer degenerating ON axons compared to b-ITON with no intravitreal injections. Unexpectedly, "pre-blast" intravitreal injections of control siEGFP also resulted in similar reductions in degenerating ON as siCASP2. Also surprisingly, the extent of axon loss at 28 dpi and the extent of axon degeneration at 14 dpi were much less in this study than in other studies using this model 40,42 which could potentially be due to the effect of the injections and the portion of the optic nerve that was examined 58 . Furthermore, there was no effect on RBPMS + RGC survival after b-ITON injury with or intravitreal injections of siCASP2 or siEGFP in both "pre-blast" and "post-blast" at the time points assessed.
Compared to intravitreal injection of siEGFP, "post-blast" intravitreal injections of siCASP2 increased vitreous inflammation, assessed by OCT vitreous intensity, while siCASP2 intravitreally injected 48 h before b-ITON decreased vitreous inflammation. The hyper-reflective dots observed qualitatively were similar to previous studies which correlated these cells on OCT with histological mononuclear cells 59 , suggesting this vitreal haze could represent macrophage infiltration in response to the combination of b-ITON and intravitreal injections. These results suggest complex regulation of RGC survival and inflammation after b-ITON treated with intravitreal injections, which is dependent on timing of intravitreal injection. In addition, the intravitreal injection itself induced differential responses in b-ITON-treated mice, dependent on the timing of injection, with "pre-blast" siCASP2 reducing inflammation and "post-blast" treatment increasing inflammation. We also observed strong evidence for a reduction in ON axon density when siCASP2 was injected after the initial blast wave compared to siEGFP, and a trend towards a reduction in axon density when it was injected 48 h before, possibly due to ON gliosis known to be associated with this model 41 .
The delivery of siCASP2 by intravitreal injection at different time points around the blast wave exposures had varied effects on retinal and ON degeneration. Our first "post-blast" injection study, intravitreally injected siCASP2 or siEGFP at five hours after the initial 2 × 15 psi blast wave exposure, which was followed by two further 2 × 15 psi blast waves on consecutive days. We chose this time point to ensure that we still captured caspase-2 activation occurring < 24 h after injury allowing time for retinal siCASP2 penetration and knockdown (16 h) 16 while remaining acceptable for clinical translation as an injured soldier or civilian may receive specialist treatment within this time frame. Notably, we observed increased vitreous inflammation when an intravitreal injection was given after the initial blast wave exposure, which was independent of the compound injected and comparable to our previous observation with necroptosis inhibitor, Necrostatin-1s 48 . There was a greater vitreous intensity detected in eyes with siCASP2 injected after b-ITON, but not in eyes receiving "pre-blast" injections compared to siEGFP control, possibly due to caspase-2 knockdown preventing apoptosis of infiltrating macrophages 60,61 , with greater macrophage infiltration and higher persistent vitreous levels of siCASP2 in the group injected after b-ITON.
In the two different treatment paradigms (injection before and after injury), the number of RBPMS + RGC 62-64 at 14 and 28 dpi were not different, suggesting that b-ITON within the 14-28 days time-frame of our experiments alone, is not enough to cause RGC death. This is perhaps not surprising as we and others detect axon degeneration prior to RGC death in models of ITON, in fact in other models the RGC death is delayed for months    41,49 . For example, others have reported delayed and progressive RGC degeneration that results in reduction of GCL thickness, between 4 and 10 months after a single 26 psi blast-wave exposure 49,50 . In support of this assertion, decreased DAPI + cells in the GCL was observed at 2 days after b-ITON which remained low at 28 days 41 . We have however, previously demonstrated RGC degeneration caused by intravitreal injection 5 h after b-ITON 48 with comparable vitreous inflammation. In our current study intraocular injection after b-ITON did not affect the number of RBPMS + RGC but may reflect either neuroprotection induced by both siCASP2 and off-target effects of the siEGFP or greater macrophage infiltration with neuroprotective effects. Although the same siCASP2 had little effect on RGC survival in this model, we reported site-specific caspase-2-mediated RGC death peripheral to the injury site after blunt ocular trauma but not central to the impact site, but CASP2 did not drive photoreceptor death, suggesting that caspase-2-mediated apoptosis is both cell and site specific 17,65 .
The number of degenerating ON axons was increased 28 days after b-ITON alone and was unaffected by siCASP2 and siEGFP intravitreal injections administered after b-ITON. In contrast, intravitreal injection of either siCASP2 or siEGFP given 48 h before b-ITON reduced both the number of degenerating ON axon profiles and the axon density, with a lesser effect on total axon number, suggesting some interference with the process of axonal degeneration likely due to ON gliosis associated with this model 41 . Again, despite injecting equivalent treatments, the time point of intraocular injection surrounding the blast caused different ON responses. We have previously shown degenerating ON and reduced electroretinography recordings and elevated levels of pro-inflammatory cytokines when administering an intravitreal injection at 1 day after the b-ITON, indicating that intravitreal injections may be injurious to the ON when delivered at this acute stage of ON injury 58 . The "pre-blast" study ended at 14 dpi and the "post-blast" study ended at 28 dpi, which could be viewed as a limitation of our study, but we did not intend to compare the two protocols to each other. These timepoints of analysis were chosen since we previously showed that 2 weeks was the peak of axon degeneration in this model and at 4 weeks significant axon loss was detected 41 . Thus, the 2-week time point was used as the most robust anatomical assessment of protection by quantification of axon degenerative profiles, whilst at the 4-week time point the most robust anatomical assessment of protection was quantification of total axons.
Long-lasting morphological and functional consequences in the eye have also been observed in models of repetitive mild traumatic brain injury (r-mTBI) 66 . For example, r-mTBI in a mouse model caused a decrease in ON diameters, increased cellularity and areas of demyelination in the ON. This was consistent with areas of decreased cellularity in the GCL and 67% reduction in Brn3a + RGC. Furthermore, SD-OCT demonstrated thinning of the inner retina whilst ERG demonstrated a decrease in the amplitude of the photopic negative response without changes in a-or b-wave amplitudes. In a separate single blast TBI model, the authors also found decreases in RNFL thickness and reduced cellularity in the GCL at 3-months with accompanying changes in retinal function 50 . However, r-mTBI led to more profound and widespread damage to the RNFL. These studies suggest that visual system dysfunction might be a common feature after blast and repeated blunt mTBI.
As we have previously reported 48 , there were infiltrating ED1 + cells in the ON at 28 days after b-ITON alone and in eyes receiving pre-blast and post-blast injections and have now shown infiltrating ED1 + cells at 14 days. ED1 + cells are likely to be infiltrating inflammatory macrophages to clear myelin debris from degenerating ON axons 67 , and would be consistent with previous findings of CD68 + cells infiltrating the brain after blast injury 68 , and macrophage ON infiltration after ultrasonic injury 69 . However, macrophages recruited into the ON may exert polarised effects since they are not only toxic to neurons and glia but can also promote CNS axon regeneration 70 . Vitreal inflammation, induced by lens injury or injection of zymosan has long been known to cause release of oncomodulin, promoting RGC neuroprotection and axon regeneration [71][72][73][74] . In this study, RGC and ON degeneration may reflect a complex balance between pro-degenerative ON macrophage infiltration and neuroprotective retinal and vitreous macrophage infiltration, with more neurodegeneration in the "pre-blast" injection group, which displayed less vitreous inflammation and neurodegeneration, than the "post-blast" injection group which had more vitreous inflammation.
Intravitreal drug delivery has become routine for the delivery of drugs, suspensions and intraocular implants into the vitreous cavity. It is the main route to deliver macromolecules to the posterior segment of the eye. Although the technique leads to targeted delivery of therapeutics, it is invasive since it requires the penetration of the globe and is associated with complications such as endophthalmitis, retinal detachment, cataracts and intraocular haemorrhage 75 . Intraocular injections can be uncomfortable, may have limited patient compliance and often require multiple injections which can increase the risk of side effects such as infectious endophthalmitis and retinal detachment 76 . However, repeated intravitreal injections of anti-angiogenic agents such as vascular endothelial growth factor (VEGF) inhibitors have become the first-line treatment for exudative age-related macular degeneration (AMD) 77 . Our study which detected additive effects of combined b-ITON and intraocular injections 48 , suggests that intraocular injection may not be the optimal therapeutic delivery method for treating ocular blast injury and that this should be considered in development of treatments for humans. However, caution needs to be exercised when projecting adverse effects of intraocular injections in mouse eyes to substantially larger human eyes since an intraocular injection in a mouse eye may inflict a greater degree of injection-related ocular injury than in the larger human eye with a ~ 1000-fold larger vitreous volume 78,79 . Hence, the overall effects of intraocular injections into the human eye may be negligible and needs further investigation.
In conclusion, despite evidence of caspase-2 activation after b-ITON, we did not detect a neuroprotective effect of caspase-2 knockdown in this model, possibly because of the limited loss of RGC soma. We demonstrate that intravitreal injection and b-ITON combined cause retinal and vitreal inflammation, with a greater effect when the injection was administered after b-ITON compared to before the injury. Depending on the timing of injection, intravitreal injection may also induce RGC axonal loss, although this was minimal. Similarly, the timing of the intravitreal injections with respect to b-ITON also determined whether siCASP2 reduced or increased vitreous inflammation. The time point of intravitreal injection surrounding ocular trauma can induce varied