The Clinical Spectrum and Disease Course of DRAM2 Retinopathy

Pathogenic variants in DNA-damage regulated autophagy modulator 2 gene (DRAM2) cause a rare autosomal recessive retinal dystrophy and its disease course is not well understood. We present two Slovenian patients harboring a novel DRAM2 variant and a detailed review of all 23 other patients described to date. Whole exome and whole genome sequencing were performed in the two patients, and both underwent ophthalmological examination with a 2-year follow-up. PubMed was searched for papers with clinical descriptions of DRAM2 retinopathy. Patient 1 was homozygous for a novel variant, p.Met1?, and presented with the acute onset of photopsia and retina-wide retinopathy at the age of 35 years. The patient was first thought to have an autoimmune retinopathy and was treated with mycophenolate mofetil, which provided some symptomatic relief. Patient 2 was compound heterozygous for p.Met1? and p.Leu246Pro and presented with late-onset maculopathy at the age of 59 years. On review, patients with DRAM2 retinopathy usually present in the third decade with central visual loss, outer retinal layer loss on optical coherence tomography and a hyperautofluorescent ring on fundus autofluorescence. Either cone–rod or rod–cone dystrophy phenotype is observed on electroretinography, reflecting the importance of DRAM2 in both photoreceptor types. Non-null variants can result in milder disease.


Introduction
Inherited retinal dystrophies (IRD) are a group of clinically and genetically heterogenous conditions that are characterized by a progressive degeneration of the photoreceptors and the retinal pigment epithelium (RPE) cells [1][2][3][4]. Pathogenic variants in >250 genes are responsible for these diseases [5]. Among those, DRAM2 retinopathy is a rare, relatively recently discovered autosomal recessive IRD associated with pathogenic variants in DRAM2. DRAM2 (MIM 613360) encodes a DNA-damage regulated autophagy modulator 2 (DRAM2), also known as transmembrane protein 77 (TMEM77), a 266 amino acid transmembrane protein containing six putative transmembrane domains [6,7]. It is ubiquitously expressed in numerous tissues, including the lymph nodes, spleen, heart, and placenta [8,9], where it localizes to lysosomal membranes and is thought to play a role in the autophagy process and tumor suppression [6,7,10,11]. Immunohistochemical analysis of the retina revealed that DRAM2 localizes to the inner segments of the photoreceptors and the apical surface of the RPE cells [12]. Since the first description of DRAM2 retinopathy in 2015 [12], fewer than 30 patients [12][13][14][15][16][17] have been reported with DRAM2 retinopathy, and the disease is still not well understood [12][13][14][15][16].   Abbreviations: VUS-variant of unknown significance; BCVA-best corrected visual acuity; RE-right eye; LE-left eye; OCT-optical coherence tomography; FAF-fundus autofluorescence; ERG-electroretinography; mfERG-multifocal ERG; PERG-pattern ERG; ffERG-full-ffield ERG; MS-mean sensitivity; ISe-inner segment ellipsoid. Fundus, OCT and FAF characteristics, visual field and ERG results were highly symmetrical between the eyes. Exam findings from the first and last exam are stated separately in case of differences. Variant calling was based on the GRCh37/hg19 genome assembly. Table 2. Pathogenic variants in DRAM2 associated with retinopathy from our study (Patient 1 and Patient 2) and previously published cases [12][13][14][15][16]. Patient 2 presented at the age of 61 years with a two-year history of the slow bilateral deterioration of vision with reading difficulties, problems recognizing faces and photophobia. At exam, her BCVA was 0.2 in BE, and she had normal color vision (Ishihara 14/15 in RE and 15/15 in LE). She had central scotoma on static perimetry (MS in RE 27 dB and in LE 26 dB) and slightly narrowed visual field on kinetic perimetry (Campus Goldmann). Slit lamp exam revealed nuclear cataract in BE, whereas fundus examination revealed an absent foveal reflex, narrow blood vessels, the granular appearance of the macula and yellow dots around the vascular arcades in BE ( Figure 4). OCT showed absent photoreceptor layer in the fovea in BE, and FAF showed foveal hypoautofluorescence delineated by a hyperautofluorescent ring in BE ( Figure 4). ERG findings were consistent with macular dysfunction; mfERG showed reduced responses, especially in the inner two rings in BE, whereas ffERG was still within the normal range ( Figure 2). After 25 months of follow-up, there was no significant deterioration of visual function, her BCVA was 0.2 in BE, color vision was still preserved (13/15 in RE and 14/15 in LE) and the fundus was similar to the first exam. OCT and FAF showed minimal progression of the disease (Figure 4), and visual field showed reduced MS (BE 24 dB) ( Figure 3). coherence tomography (OCT) images at the age of 35 years; (E,H) more peripheral OCT images at the age of 35 years; (I,K) FAF and (J,L) OCT images at the age of 37 years. Green lines in the FAF images show the location of the corresponding OCT scans in images (D,G,J,L). Yellow lines on the FAF images show the location of the corresponding OCT scans in images (E,H). Note the intraretinal cysts in images (D,E,G,H) (white arrows). On follow-up, the disappearance of the foveal photoreceptors (J,L) on and widening of the hypoautofluorescence in the fovea (I,K) is seen. Abbreviations: RE-right eye; LE-left eye. Full-field ERG and multifocal ERG findings in Patient 1 (last visit) and Patient 2 (first and only visit) compared with an age-matched control subject. Abbreviations: DA-dark-adapted; LA-light-adapted; Amp-amplitude; a-a-wave, negative ERG component that arises mostly from photoreceptoral activity; b-b-wave, a positive ERG component originating mostly from bipolar cells; OP2-oscillatory potential, second wave. Arrows indicate reduced (vertical arrow) and delayed (horizontal arrow) full-field ERG responses in Patient 1. Note the reduced mfERG responses in Patient 1 over the whole macular region. The mfERG responses in Patient 2 were reduced, especially in the inner two rings.  Patient 2 presented at the age of 61 years with a two-year history of the slow bilateral deterioration of vision with reading difficulties, problems recognizing faces and photophobia. At exam, her BCVA was 0.2 in BE, and she had normal color vision (Ishihara 14/15 in RE and 15/15 in LE). She had central scotoma on static perimetry (MS in RE 27 dB and in LE 26 dB) and slightly narrowed visual field on kinetic perimetry (Campus Goldmann). Slit lamp exam revealed nuclear cataract in BE, whereas fundus examination revealed an absent foveal reflex, narrow blood vessels, the granular appearance of the macula and

Review of Previously Published Cases with Variant in DRAM2
The search in the previously published cases of retinopathy caused by biallelic variants in DRAM2 identified 6 reports with overall 24 cases from 14 families [12][13][14][15][16][17]. The genetic and clinical findings of these cases including ours are summarized in Tables 2 and  3, respectively.

Review of Previously Published Cases with Variant in DRAM2
The search in the previously published cases of retinopathy caused by biallelic variants in DRAM2 identified 6 reports with overall 24 cases from 14 families [12][13][14][15][16][17]. The genetic and clinical findings of these cases including ours are summarized in Tables 2 and 3, respectively.

Genetic Findings
Including this report, 19 unique variants in DRAM2 associated with retinopathy caused were identified (

Genetic Findings
Including this report, 19 unique variants in DRAM2 associated with retinopathy caused were identified (

OCT Characteristics
OCT data were available for 20 patients, 19 symptomatic and 1 asymptomatic. Outer retinal layer loss was observed in most of the symptomatic patients (84%; 16/19). One symptomatic patient (Patient 10) presented with reduced foveal thickness at the age of 21, which progressed to outer retinal layer loss at the age of 25. Foveal thinning due to severe atrophy of the photoreceptor and RPE layers was observed in Patient 22, and disrupted ellipsoid zone with thinning of the outer retinal layers at the macula was seen in Patient 23. OCT images of the asymptomatic individual (Patient 8) showed reduced foveal thickness with perifoveal disruption of the ellipsoid zone. Among the patients with available OCT images (N = 11), 18% (2/11; Patient 1 and Patient 4) had preserved integrity of the Ise band and/or external limiting membrane. In Patient 1, the integrity of ISe was disrupted after 2 years.

FAF Characteristics
FAF data were available for 14/25 patients, all of whom were symptomatic. The most frequent pattern observed on FAF was a central hypoautofluorescent area surrounded by a hyperautofluorescent ring (57%; 8/14). Patient 4 had a paracentral hyperautofluorescent ring and central thinning of the photoreceptor layer. Perifoveal hypoautofluorescence was seen in Patient 22 and Patient 24, which in both progressed to more diffuse through followup. Patient 25, who was older than 22 and 24, had greater retinal hypoautofluorescence that was seen in the macula and periphery. FAF images from Patient 1 and Patient 20 showed a mosaic pattern with hyperautofluorescence and hypoautofluorescence. Of the patients, 36% (9/25) had follow-up data after the median of 4 (range 2-8) years. FAF images from follow-up visits revealed progression of the disease (enlargement of the central defect towards midperiphery and periphery) in six individuals (67%; 6/9) with median age of 55 (range 38-54) years. In patients with available FAF images (N = 10), we measured the area of definite decreased autofluorescence. The correlation between age and lesion area on FAF (at the last exam) was not statistically significant (Pearson correlation, p = 0.97) ( Figure 7A), and the correlation between lesion area on FAF and visual acuity was also not statistically significant (Pearson correlation, p = 0.15) ( Figure 7B).

Genotype-Phenotype Correlations
We aimed to determine if any genotypes result in a notably milder phenotype. The most frequent genotype was Gly47Valfs*3 in homozygous state, present in 11 patients (Patients 8-18) from the same family of Pakistani origin [12]. The patients with this genotype (baseline group) had relatively more severe course of disease in comparison and other patients (others) (Figures 5, 6 and 8). The genotype-phenotype correlation was statistically significant for several parameters.
Disease onset was significantly earlier in the baseline group (median 24; range 16-28 years) than in others (median 32; range 19-59 years) (Mann-Whitney test, p < 0.05) (  Figure 5). BCVA at last exam for patients homozygous for variant Gly47Valfs*3 was worse than for other patients when corrected for age (multiple regression, p < 0.01) (Figure 8). Accounting for visual acuity decline with age, five patients had BCVA above the 95% confidence interval (CI) of the baseline group (Figure 8).  (Figures 6 and 8) but was at the last exam within the 95% CI of the baseline group (Figure 8).

Discussion
This study provides a comprehensive overview of the disease spectrum of DRAM2 retinopathy. We extend the cohort of 16 patients described by El-Asrag et al. [12] and Sergouniotis et al. [13] with an additional 10 patients, including two Slovenian patients with a novel variant, one with an acute onset of widespread retinopathy mimicking autoimmune retinopathy and the other with a mild, late-onset maculopathy. Additionally, we performed an analysis of all reported FAF images and propose genotype-phenotype correlations.

Disease Onset
Based on the review all (N = 25) DRAM2 patients with clinical data, the first symptoms most often appear in the third decade [12][13][14][15][16], although the onset varies from teenage years to the sixth decade of life ( Figure 5). The earliest reported onset was by Patient 18 at the age of 16 [12], whereas the Slovenian Patient 2 had the latest onset so far, at the age of 59 years ( Figure 5). This finding expands the phenotypic spectrum of DRAM2 retinopathy with a late-onset presentation, as previously the latest reported onset was at the age 45 years [15]. If the disease occurs late in life and is mild, as in Patient 2, in whom the peripheral retina was spared, it may be confused with age-related macular degeneration. It is therefore possible that late-onset patients are undiagnosed and that the prevalence of DRAM2 retinopathy is higher than currently thought. A large range in disease onset has also been observed in other retinal dystrophies, including the most frequent monogenic disease, ABCA4 retinopathy [19]. There, the variability in onset is to some extent linked to the type of genetic defect [20], and it is likely that the same is true for DRAM2 retinopathy. Slovenian Patient 2 with the latest disease onset harbored a combination of a start loss (p.Met1?) and a missense variant (p.Leu246Pro). The other six patients with relatively late disease onset (30-39 years) also harbored presumably non-null variants, namely p.Met1?; p.Leu246Pro (Patient 2), p.Ala22del homozygous (Patient 5), p.Gly57Arg homozygous (Patient 19), p.Arg74His homozygous (Patient 20), p.His121Leu homozygous (Patient 21) and p.Ala210GlufsTer16 homozygous (Patient 22). We presume that certain alleles, such as p.Leu246Pro, retain some residual function of DRAM2 protein; however, a larger cohort of patients with the same genotypes and/or in vitro studies are needed to confirm this. The first symptom in DRAM2 retinopathy is usually central visual loss, which was reported by 63% (15/24) patients and is consistent with early macular impairment [12,13,21]. In addition, some patients (17%; 4/24) reported photophobia. This is a relatively frequent symptom in patients with IRDs and is thought to reflect an early involvement of photopigment-containing cells [15,22]. It is suggested that light hypersensitivity occurs due to impairment of either S-cones in the parafoveal region, rods, and/or photosensitive retinal ganglion cells, but the mechanism is not fully understood [22]. Interestingly, a subset of DRAM2 patients (12%; 3/24) described difficulty seeing in dark environments and dark adaption difficulties as their initial symptoms, suggesting an early rod impairment [14,16]. Their genotypes were p.Trp3del homozygous (Patient 3) and p.Val16Ala; p.Gly95Val (Patient 4). Although nyctalopia is often observed in patients with CRD, it is not expected to be the first symptom [16,23]. Patient 4 had reduced dark-adapted and light-adapted responses, whereas Patient 3 had ERG findings in the pattern of RCD.

Visual Acuity
The median BCVA at first exam (median age 34; range 19-61 years) in the better eye was 0.4 Snellen decimal, deteriorating to 0.1 at last exam (median age 41; range 24-63 years) ( Figure 6). This reflects the progressive nature of DRAM2 retinopathy, with mostly poor visual outcomes after the fifth decade. The best preserved BCVA was noted in Slovenian Patient 2, who at the age of 63 years still had BCVA of 0.2 Snellen decimal (0.7 LogMAR) ( Figure 6). Several other patients also had notably better preserved BCVA for their age in comparison with patients homozygous for p.Gly47Valfs*3 (baseline group) (Figure 8).

Fundus Appearance
Fundus examination in both Slovenian patients and previously reported cases [12][13][14][15][16] reveals that degeneration initiates in the macula (Figures 1 and 4). First signs were in most cases described as macular atrophy and/or granular macular appearance (Tables 1 and 3). Tiny white or yellow dots were observed in 48% (12/25) of patients, including in the Slovenian patients. Considering the DRAM2 role in autophagy, the dots could represent the residual components of the visual cycle [11]. Although probably not pathognomonic, they could potentially help differentiate DRAM2 retinopathy from other dystrophies. At first exam, 32% (8/25; median age 50; range 32-61 years) patients had pigmentary depositions (bone-spicule degeneration and/or widespread pigment clumping), and it was noted in most patients with follow-up exams (82%; 9/11), suggesting progressive peripheral degeneration [15,16]. According to the fundus appearance, disease progression is rather slow, with changes towards periphery observed at a median age of 50 (range 32-71) years. Only one (8%; 1/12) patient older than 45 years had no peripheral changes (Patient 4).

OCT
Outer retinal loss was observed on OCT in most (84%; 16/19) of the symptomatic patients as well as the one asymptomatic patient. This is consistent with the finding that DRAM2 localizes to the outer retina, i.e., the photoreceptor layer and the apical surface of the RPE [12]. The absence of DRAM2 in the retina is thought to reduce the effectiveness of autophagy, leading to diminished photoreceptor renewal [12]. In the asymptomatic individual (Patient 8), OCT images revealed perifoveal disruption of the ellipsoid zone, which suggests that the initial impairment began in the photoreceptor layer [12]. On the other hand, considering that autophagy in RPE is important for the degradation of the outer segments of photoreceptors, the degeneration could also begin in the RPE [13]. Interestingly, in Patient 1, the loss of photoreceptor layers also began in the parafoveal region with relative foveal sparing, which coincided with her good BCVA at the first exam. However, after 2 years, her vision deteriorated, and the loss of photoreceptor in the foveal and perifoveal regions was observed. In total, 18% (2/11) of patients with available OCT images had preserved photoreceptors in the fovea when parafoveal degeneration had already occurred. It is possible that patients who had loss of central vision and outer retinal layer loss on OCT at the first presentation also had perifoveal impairment a few years before and only reported to the ophthalmologist when degeneration spread to the fovea. Kuniyoshi et al. [16] showed pericentral scotoma in two patients that expanded to involve the whole macula in the course of disease (Patient 3 and Patient 20 in this review). The pattern of initial foveal sparing and/or initial perifoveal degeneration is not specific for DRAM2 retinopathy and has been noted in several other monogenic and multifactorial diseases. These include macular diseases such as age-related macular dystrophy and cone-rod dystrophies such as foveal sparing ABCA4 retinopathy and certain PRPH2 retinopathies [24][25][26]. RCD may also begin with an annular scotoma and in late stage also often exhibits perifoveal RPE atrophy, surrounding the preserved RPE in the fovea [27]. These patterns of degeneration are thought to be influenced by disease independent factors such as metabolic differences between regions of the macula, rod-derived cone viability factor, variations in macular pigment and peak distribution, cone density, increased vulnerability in certain parafoveal photoreceptors, factors related to RPE and choroid, etc. [28].

FAF
The most frequent FAF pattern in DRAM2 patients was central hypoautofluorescent area surrounded by a hyperautofluorescent ring, present in 57% (8/14) of patients, corresponding with the area of photoreceptor loss on OCT (Figure 4). The hyperautofluorescent ring may be found in several IRDs and usually delineates the border between the preserved and affected retina. In cone dystrophy (CD)/CRD, the retina is affected inside the ring, whereas in RCD (retinitis pigmentosa (RP)), the retina is affected outside the ring [29][30][31]. The source of hyperautofluorescence is thought to be photoreceptor outer segment loss overlaying the still intact RPE, resulting in the increased detection of the normal RPE autofluorescence due to the decreased blockage of photoreceptors [13]. Another cause of increased autofluorescence is possibly the increased accumulation of lipofuscin in the degenerating photoreceptor inner segments and/or the RPE, potentially contributing to photoreceptor damage [29,30,32].
The hypoautofluorescence within the ring in DRAM2 corresponded to the RPE atrophy, which presumably followed photoreceptor loss, as has been observed in other CRD patients [30]. FAF images from follow-up visits showed the enlargement of the hyperautofluorescent rings, a clinical feature that suggests disease progression over time and that has also been reported in patients with other CRD [31,[33][34][35]. The enlarging of central hypoautofluorescence is consistent with expanding macular atrophy, and the expansion of the hyperautofluorescent ring probably reflects the disorganization of the photoreceptors during the progress of the disease [31,33]. Slovenian Patient 1 did not exhibit a clear hyperautofluorescent ring but instead, larger hyperautofluorescent lesions in the macula, corresponding with photoreceptor loss on OCT (Figure 1). From the available images for published cases (N = 14), Patient 20, who is also the oldest patient (71 years old; p.Arg74His homozygous), exhibited this finding. Furthermore, 43% (6/14) of patients with at least 55 • imaging showed hypoautofluorescent lesions along the midperiphery, representing widespread retinal disease. Interestingly, their median age (44; range 38-54) was similar to the patients without midperipheral lesions (45; range 25-71). As in other retinal dystrophies, FAF and OCT are useful imaging tools for evaluating the structural damage of the retina and following the disease progression in patients with DRAM2 retinopathy.

ERG
First reports of DRAM2 retinopathy described findings consistent with CRD [12][13][14][15][16]. Later, Kuniyoshi et al. [16] observed that some patients exhibit a RCD phenotype. Macular dystrophy was also reported [12,13]. On review, among patients with ERG data (N = 14), 50% had ERG features of CRD, 21% of RCD and 29% of macular dystrophy with normal full-field ERG. In CD/CRD, cones are primarily targeted, which results in central visual loss, a remarkable decrease in visual acuity and central scotoma [1]. In CRD, rod impairment and rod-related symptoms such as loss of peripheral vision and night blindness appear at subsequent stages [1,3]. On the other hand, RCD (RP) is characterized by primary rod degeneration followed by cone degeneration. These patients usually present with night blindness and progressive loss of the peripheral visual field that is followed by loss of central vision due to cone impairment [1]. It is important to note that ERG diagnosis of a RCD does not always align with the clinical diagnosis. For example, typical RCD (RP) is supported with structural findings of peripheral retinal involvement and relative sparing of the macular area. None of the DRAM2 patients were reported to exhibit this phenotype as they all exhibited early macular involvement except for initial foveal sparing in some cases. Patient 23 presented with difficultly in night vision at the age of 19 years, which could suggest primary rod impairment. However, she lost peripheral visual field only at 43 years, at the same time as her ffERG responses were undetectable. A couple of years before that, she already had a central scotoma, while ffERG responses were still in normal range. Considering clinical and ERG findings together, her presentation is more consistent with CRD. In keeping with that, among the two Slovenian patients, Patient 1 (homozygous for p.Met1) had a greater reduction of rod-than cone-specific function on ffERG ( Figure 2) but with a predominant macular involvement. At presentation, the ERG findings were not typical of CRD but were more like retinopathy of inflammatory origin. Typically, CRD patients have severely abnormal mfERG and PERG that corelates with poor visual acuity [36], whereas she still had good visual acuity, although perifoveal structural loss was seen. In addition, ffERG in CRD typically shows preserved but reduced rodand cone-specific responses [37], whereas the patient had reduced rod-specific responses while the responses of cone system were only delayed. This suggests that cones were structurally still largely preserved at that time but that their functioning was disturbed. Intraretinal cysts that were observed at that time in regions of preserved photoreceptors in the macula (Figure 1) are also in concordance with the existence of an inflammatory process. On the other hand, Patient 2 (p.Met1?; p.Leu246Pro) had abnormal mfERG and normal ffERG (Figure 2), a finding suggestive of macular dystrophy. The homozygous nonsense variant (p.Met1?) that Patient 1 carries seems to lead to CRD, whereas the same variant in the biallelic state (p.Met1?; p.Leu246Pro) that Patient 2 carries seems to lead to macular dystrophy. However, studies on a larger cohort of patients will be needed to determine whether there exists some genotype-phenotype correlation or if any genetic, epigenetic modifiers and environmental factors affect the phenotype. Considering all these observations, DRAM2 retinopathy most commonly leads to CRD, less often to atypical CRD with early rod dysfunction or macular dystrophy and never to typical RCD. A similar observation has been proposed in EYS, which mostly causes RP, i.e., RCD, but has also been described as a cause of CRD and macular dystrophy that further confirms the clinical heterogeneity of IRDs [38].

Genotype-Phenotype Correlations
We identified a novel variant c.3G > A (p.Met1?) in the DRAM2. The variant is predicted to cause the loss of the start codon with a novel start codon downstream. Both Slovenian patients carried this variant. A priori predictions of start loss variant effect on the final protein structure is difficult to determine. It has been shown that there are non-canonical or non-AUG translation initiation sites that are used by the cell to warrant protein translation. These sites are generally not as effective as the canonical ones, but they still guarantee a minimum production of the protein. In addition, different protein isoforms may have a different start codon and can make up the lack of one of the others. Therefore, functional in vitro studies are necessary to demonstrate the actual biological effect of start-loss variant [39][40][41]. Patient 1 was homozygous and presented with a widespread retinopathy, while Patient 2 was compound heterozygous for c.3G > A (p.Met1?) and a missense variant c.737T > C (p.Leu246Pro) and presented with a late-onset maculopathy. We presume that the disease in Patient 2 was milder and delayed mostly due to the residual function conferred by the missense variant. However, it is possible that p.Met1? also retains some DRAM2 function. Compared to the baseline group of patients with two null variants, the Patient 1, homozygous for p.Met1? had a delayed disease onset (35 years vs. median 24 years) and significantly better preserved visual acuity ( Figure 8). Nevertheless, the disease was still affected the whole retina and progressed relatively quicky, thus the retained function, if present, is not enough to prevent a severe disease.
Interestingly, although the 26 patients (25 with clinical data) had 19 different variants, 81% (21/26) patients were homozygous. The major reason for this seems to be consanguinity, which was noted for Patients 8-18 from the same large family of Pakistani origin [12,13] as well as Patient 5 [12,13], Patients 22 and 25 [15], Patient 3 and Patient 23 [16]. The other possible reason is that different variants are founder variants in different geographic regions. Slovenian homozygous patient had no history of consanguinity but does come from a small geographic region. It is possible that the variant p.Met1? is a founder variant for the Slovenian region. A similar observation was made for Usher patients in Slovenia, where variant c.11864G > A (p.Trp3955*) in represent the majority (84%) of pathogenic alleles in Slovenian USH2A Usher syndrome population, and is otherwise a rare variant [42]. In the previous studies of DRAM2 retinopathy it was suggested that transcripts of variants in Patient 7 (p.Ser44Asn; p.Trp165*), Patients 8-18 (p.Gly47Valfs*3 homozygous), Patient 22 (p.Ala210GlufsTer16 homozygous) and Patient 24 (c.518-1G > A homozygous) create premature termination codons (PTC), which are eliminated by the nonsense-mediated mRNA decay (NMD) to avoid aberrant gene expression [12,13,15,43]. In addition to that, Abad-Morales et al. [15] observed that DRAM2 mRNA expression in Patient 22 and Patient 24 was decreased, coinciding with PTC and promoting NMD transcript degradation. They had similar phenotypes with mid-peripheral RPE disturbances. In Patient 25 no differences in DRAM2 mRNA levels were detected, because skipping exon 8 does not affect the open reading frame. Interestingly, the Patient 25 had more severe and widespread degeneration in this study and the author proposed it is due the requirement of exon 8 for the correct biological function of the protein [15]. Sergouniotis et al. [13] concluded that individuals harboring at least one loss-of-function variant present first symptoms earlier than patients harboring only missense variants or in-frame deletions. Similar observations were made in other retinal dystrophies. For example, in ABCA4 retinopathy, the BCVA and lesion area involvement were significantly more severe in the patients with two null variants than in patients harboring null and missense variant or two missense variants [44]. In our review, the patients with biallelic variant p.Gly47Valfs*3 (baseline group) BCVA at first exam had significantly worse phenotype in comparison to most patients with other biallelic variants (Figure 8). One option is that some stop variants escape the NMD and some splicing variants partially retain normal splicing. Considering that patients with p.Gly47Valfs*3 were from the same family, other non-DRAM factor could influence their phenotype. The genotypes that were consistently better than the baseline group were namely p.Met1? homozygous (Patient 1), p.Met1?; p.Leu246Pro (Patient 2) p.Gly57Arg homozygous (Patient 19) and p.Ala210GlufsTer16 homozygous (Patient 22). On the other hand, some genotypes were relatively similar to the baseline group, e.g., c.518-1G > A homozygous (Patient 24) and c.693 + 2T > A homozygous (Patient 25). The missense variant of Patient 7 p.Ser44Asn; p.Trp165*) also likely does not retain much DRAM2 function considering the patient's severe phenotype.

Immunological Component of Retinal Dystrophies
Retinal dystrophies are believed to have an immunological component, most likely due to the reaction of the immune system to the decaying retinal tissue [45]. The acute onset in Patient 1 effectively mimicked an autoimmune retinopathy and interestingly, the patient reported some improvement on anti-inflammatory treatment. Nevertheless, this warrants further studies and the often serious side effects of immunosuppressive therapy for this inevitably progressive disorder must be weighed.

Evaluation of Slovenian DRAM2 Patients
The study included two female patients with retinal dystrophy in whom genetic analysis identified biallelic DRAM2 variants, ascertained from Eye Hospital, University Medical Centre Ljubljana, Slovenia. The patients were from two unrelated families, aged 37 (patient 1) and 62 years (patient 2). All investigations were carried out in accordance with the Helsinki Declaration on Biomedical Research in Human Beings. Informed written consent was obtained from the patients.

Genetic and Bioinformatic Analysis
Genetic analysis included whole genome sequencing in Patient 1 and whole exome sequencing in Patient 2. Genomic DNA was extracted from blood samples according to the standard procedure. Sequencing of the defined clinical target was performed using next-generation sequencing in the isolated DNA sample. Briefly, the fragmentation and enrichment of the isolated DNA sample were performed according to the Illumina Nextera Coding Exome capture protocol, with subsequent sequencing on Illumina NextSeq 550 in 2 × 100 cycles (Illumnia, San Diego, CA, USA). After duplicates were removed, the reads were aligned to the UCSC hg19 reference assembly using the Burrows-Wheeler aligner algorithm (BWA) (v0.6.3), and variant calling was performed using a GATK framework (v2.8). Only variants exceeding the quality score of 30.0 and depth of 5 were used for downstream analyses. Variant annotation was performed using the ANNOVAR and snpEff algorithms, with pathogenicity predictions in the dbNSFPv2 database. Reference gene models and transcript sequences are based on the RefSeq database. Structural variants were assessed using the CONIFER v0.2.2 algorithm. Variants with population frequency exceeding 1% in gnomAD, synonymous variants, intronic variants and variants outside the clinical target were filtered out during analyses. An in-house pipeline was used for the bioinformatic analyses of exome sequencing data, in accordance with the GATK best practice recommendations [46].The interpretation of sequence variants was based on ACMG/AMP standards and guidelines [18]. When sequencing the DNA sample, we reached a median coverage of 67× and covered over 99.9% targeted regions with a minimum 10× depth of coverage [47]. The presence of the pathogenic variant in the population was evaluated in the gnomAD database (gnomad.broadinstitute.org, accessed on 31 May 2022). Genetic characteristics of patients were classified into three types based on variant effect predictor (VEP) (https://gnomad.broadinstitute.org/gene/ENSG0000015 6171?dataset=gnomad_r2_1 (accessed on 31 May 2022)).

Clinical Examination
Patients underwent a complete ophthalmological exam, including BCVA (Snellen), color vision (Ishihara plates), slit lamp biomicroscopy and dilated fundus examination. Visual field testing was performed using manual kinetic Goldmann perimetry and Octopus G2-top static perimetry. Imaging included color fundus photography with conventional fundus color photographs (Topcon, Tokyo, Japan), FAF and OCT (Spectralis, Heidelberg Engineering, Dossenheim, Germany). The diameter of the hypoautofluorescence lesion (lesion area with a level of darkness of almost 100% in reference to the optic nerve) was measured using ImageJ (U.S. National Institutes of Health, Bethesda, MD, USA). The scale was set using the known approximate 15 • distance between the center of the optic nerve and the fovea. ERG was performed according to the standards and guidelines of the International Society for Clinical Electrophysiology of Vision (ISCEV) [48][49][50], using Espion (Diagnosys LLC, Lowell, MA, USA) or RETI scan (Roland Consult Stasche & Finger GmbH, Germany) visual electrophysiology testing systems with HK-loop active electrodes. ffERG was used to assess the general retinal function with the following recording protocols: dark-adapted 0.01 ERG (DA 0.01 ERG; rod-specific response, driven by on-bipolar cells), dark-adapted 3 ERG (DA 3 ERG; DA 3 ERG; combined responses from photoreceptors and bipolar cells, mostly rod-specific), dark-adapted oscillatory potentials (DA osc. pot.; responses mostly from amacrine cells) light-adapted 3 ERG (LA 3 ERG; cone-specific response; a-waves originates from cone photoreceptors and cone off-bipolar cells, while the b-wave arises from on-and off-cone bipolar cells), light-adapted 30 Hz flicker ERG (LA 30 Hz; cone-specific response) [48]. MfERG; patient 1 and 2) [49] and PERG (Patient 1) [50] were used to assess the function of the macula. mfERG testing was performed with the stimuli of 60 • in the diameter, presented on a cathode-ray tube monitor. The stimulus included an array of 61 hexagons, which were modulated between light (L) and dark (D) with 96-98% contrast according to a binary m-sequence (511 samples of the sequence: LDDDD). PERG was elicited with 0.8 • checkerboard pattern, presented on a 21.6 × 27.8 • CRT screen simulator. The checkerboard pattern reversed 1.8 times per second, and the contrast between the black and white fields was 99%. The signals were amplified and stored on a hard disc on the computer for further analysis. Head MRI, chest X-ray and PET-CT were performed in Patient 1. Blood screening in Patient 1 included paraneoplastic panel, tumor markers, rheumatology screening and the exclusion of common infectious ethology.

Review of Previously Published Cases with Variant in DRAM2
The electronic database PubMed was queried (on 20 March 2022) to review previously published cases with retinopathy caused by a pathogenic variant in DRAM2. There were no publication year or language restrictions. Inclusion criteria consisted of biallelic variants in DRAM2 and a description of the phenotype. When FAF images were available, the diameter of the hypoautofluorescence lesion was measured as stated above (Chapter 4.1.2.)

Genotype-Phenotype Analysis
For the purpose of distinguishing whether any genotype results in a milder phenotype, we used the relatively large group of patients with the same, presumably null, genotype (homozygous p.Gly47Valfs*3; N = 11) as the baseline cohort. The p.Ala210GlufsTer16, present in homozygous state in Patient 22, was not considered null, as the variant occurs later on in the gene, potentially escaping the NMD [15], which could result in residual protein function. Similarly, p.1Met? (Patient 1) possibly results in a start codon later on and was therefore also not considered null for this analysis. Patient 26 (p.Leu246Pro homozygous) had no clinical data.

Conclusions
In conclusion, a novel start loss variant in the DRAM2 (p.1Met?) was identified in two Slovenian patients, causing severe RCD in homozygous state and a mild, late-onset macular dystrophy in trans with p.Leu246Pro. On the review of all published DRAM2 cases, patients usually present with central visual loss in the third decade and macular abnormalities on fundus examination. OCT findings most commonly reveal outer retinal layer loss, whereas FAF usually shows hyperautofluorescent ring that enlarges towards the periphery during the progression of the disease. ERG findings are most commonly in the pattern of CRD, although macular dystrophy and atypical rod-cone pattern with early macular involvement are also possible. Certain non-null variants such as p.Leu246Pro may result in milder disease.