*3.3. Ultrastructural Characterizations*

Next, we performed ultrastructural characterization by TEM of the membranes extracted from three subjects belonging to the PDR and MH groups. The membranes from RD cases were not included due to insufficient numbers.

The retinal cell-types were characterized based on previously published morphological details [13–18]. Varying thickness of collagen fibrils, the main component of the ERMs was observed in both PDR and MH. We frequently detected astrocytes, Müller glia with long processes, microglia like cells with thick dark heterochomatic cell body and a few dark pigmented cells (Figure 2) in the PDR versus the MH specimens.

**Figure 2.** Transmission electron microscopy images of fibrocellular membranes from; (**A**–**C**); macular hole, and (**D**–**E**); proliferative diabetic retinopathy membrane showing the presence of astrocytes, Müller glia processes and microglia. The cell body of a Müller cell can be identified by its nucleus and dark cytoplasm (red color filled triangular shape). An astroglial cell process with flattened intermediate filaments (yellow colored asterisk) can be observed next to the Müller cell. The microglia can be seen with round shaped dark cell bodies (brown arrow) and glial cells processes with skyblue star shaped structure. (**B**,**E**) show collagenous fibrils of membrane indicated by black arrow and; (**C**,**F**) show their corresponding magnified image.

#### *3.4. Characterization of Di*ff*erent Cell Types in Each Group*

We next validated the presence of the specific cell types by indirect immunofluorescence using appropriate marker antibodies. We used nine samples, with three samples from each group (MH, RD and PDR). The MH samples were used as negative controls.

Our analyses confirmed the presence of both macroglial and microglial cells along with other proliferative spindle shaped cells. While GFAP and ALDH1L1, markers for astrocytes (and gliosis) were mostly detected in all specimens, the PDR samples showed greater number of GFAP positive cells present in the membrane as compared to MH and RD specimens (Figure 3 and Figure S2). CRALBP staining did not significantly differ in the Müller glia among PDR and RD samples. The number of CD11b (microglia marker) positive cells was higher in the RD membranes as compared to the others.

**Figure 3.** Representative images showing IHC of membranes from pathological conditions MH, RD and PDR with intense cytoplasmic and membranous expression of GFAP, CRALBP, and CD11b proteins. (Scale bar; 100 μm).

Quantification of the cells present on the membrane validated the higher number of CD11b positive cells in RD than MH and PDR. On the other hand, increased number of GFAP positive cells was seen in PDR as compared to all cases (Table 1).


**Table 1.** The number of different cell types identified by immunohistochemistry using different specific markers where GFAP; astrocytes/gliotic changes, CRALBP; Müller glia, CD11b; microglia are used.

#### *3.5. Evaluation of Oxidative Stress among Di*ff*erent Pathological Condition*

To further explore the involvement of oxidative stress in the ERMs, we performed immunofluorescence analysis using anti-OXR1 antibody in the same tissue specimens (*n* = 3) of MH, RD and PDR cases. OXR1 is expressed at elevated levels in cells undergoing oxidative stress [19]. We found a greater number of OXR1-positive cells in the RD sample than PDR (*p* = 0.04) shown in Figure 4; none of the cells was positive for OXR1 in the membranes from MH.

**Figure 4.** Representative images of IHC showing the expression of OXR1 in MH, RD, and PDR. The graph shows the mean percentage of OXR1 positive cells for each pathological condition (*n* = 3) (scale bar; 100 μm) \*\* *p*-value < 0.05.

#### *3.6. Gene Expression Analysis of Oxidative Stress and Inflammation-Related Pathway Genes*

Oxidative stress though being closely interlinked to inflammation, could also be the cumulative effect of continued insults to the eye such as ischemia and hyperglycemia [20]. To further quantitate the underlying oxidative stress and inflammation in membrane formation and DR pathogenesis, a targeted gene expression profiling was undertaken by semi-quantitative real time PCR using SYBR green chemistry. We used MH as control and the expression was measured among MH, PDR and RD. Initially, to measure the underlying oxidative stress, we used *Nrf2*, and *HIF1-*α while *MMP9*, and *IL1-*β were used as markers of inflammation. Expression of *CD11b* and *VEGF* was also measured for activated microglia and angiogenesis. In comparison to MH, the gene expression of *Nrf2*, *HIF1-*α and *MMP9* was higher in the RD (*Nrf2*: 31.81 ± 0.76, \*\*\* *p* = 0.00066; *HIF1-*α: 13.423 ± 0.63, \*\*\* *p* = 0.005; *MMP9*: 8.851 ± 0.86, \* *p* = 0.04) followed by PDR (*Nrf2*: 2.76 ± 0.16, \*\* *p* = 0.007; *HIF1-*α: 3.88 ± 0.47, ns, *p* ≥ 0.05 and; *MMP9*: 4.901 ± 0.86, \* *p* = 0.03). Consistently, we noticed increased expression of *IL1-*β in RD (8.648 ± 0.43, \* *p* = 0.03) and PDR (5.282 ± 0.27, ns, *p* ≥ 0.05) compared to MH. *VEGF*, an angiogenic marker showed almost similar elevated level of expression in both RD (2.542 ± 0.63, \* *p* = *0.031*) and PDR (1.42 ± 0.13, \* *p* = 0.01). Likewise, *CD11b*, also showed significantly higher expression in RD (10.333 ± 0.23, \*\*\* *p* = 0.000076), and PDR (3.092 ± 0.25, \* *p* = 0.011) than MH (Figure 5a).

**Figure 5.** (**<sup>a</sup>**,**b**) Differential gene expression analysis of oxidative stress, inflammatory markers and their associated pathway involved in the pathogenesis of epiretinal membrane of RD, PDR compared to macular hole (\* *p* ≤ 0.05, \*\* *p* ≤ 0.01, \*\*\* *p* ≤ 0.001) (**c**) Heat map showing the expression pattern of the differentially expressed genes in PDR and RD.

We then analysed the expression of the genes associated with major signaling pathways including *NOTCH1, DKK1* and *ERK1* in the ERMs from RD and PDR. Besides oxidative stress and inflammation, the additional genes and pathway chosen for this analysis were selected based on the existing proposed pathogenic theories in the literature and those known to be involved in retinal cellular proliferation and maintenance. There was significant upregulation of *NOTCH1* gene expression in RD (8.427 ± 0.64, \*\* *p*-value=0.008) and PDR (6.459 ± 0.81, \* *p*-value = 0.02) as compared to control (MH) (Figure 5b). Although *DKK1* and *ERK1* were downregulated, the effect was not statistically significant.

Further, the heatmap, a clustering based method that groups genes and/or samples together based on the similarity of their gene expression pattern, showed the differential expression profile of the genes for PDR and RD shown in Figure 5c. Except for *NRF2* all other genes that clustered together seems to be regulated by common processes.

To further validate the presence of activated microglial, we again performed immunofluorescence analyses of both resting microglia and activated microglia markers: F4/80 and Iba1 respectively. The results showed the presence of both resting as well as activated microglia in all pathologies except MH as shown in Figure 6.

**Figure 6.** Validation of activation of microglia on the membrane shown by immunohistochemistry using marker for resting; F4/80 and activated stage; Iba1 in different pathological conditions MH, RD and PDR; scale bar; 100 μm.

#### *3.7. Protein-Protein Interactions for the Di*ff*erentially Expressed Genes by In-Silico Analysis*

To further understand functional overlap among the significantly expressed genes in different pathways, we performed bioinformatic analysis using STRING (v 11.0). The examined genes (*HIF1-*<sup>α</sup>, *Nrf2*, *VEGF*, *IL1-*β, *CD11b*, *NOTCH1*, *MMP9*, *ERK1*, and *DKK1*) showed interaction among each at the protein level (Figure 7a). Several other proteins were also identified that interact with the studied proteins to modulate specific signaling pathways. Genes differentially expressed in the ERMs of PDR and RD are also shared with the genes belonging to the innate immune system, hypoxia related signaling, inflammatory pathways mediated by microglia, VEGF signaling, *TNF-*α signaling, MAP kinase signaling and extra cellular reorganization (Table 2).

Further, functionally associated proteins involved in the disease pathogenesis were also determined by studying their co-expression. We found *MMP9*, *IL1-*β and *ITGAM* (alias for CD11B) are functionally associated and co-expressed. Similarly, *HIF1-*<sup>α</sup>, and *Nrf2* were co-expressed together showing their strong association in disease pathogenesis (Figure 7b).

**Figure 7.** (**a**) Prediction of protein-protein interaction for the differentially expressed genes among PDR and RD in the present study. Protein 3D structure are enclosed in the circles. The colors of bond between the proteins indicate the evidences for their interaction (known interactions; skyblue-curated database, purple-experimently determined, other predicted interactions; lime-textmining, black-co-expression). Protein of interest studied are enclosed in dotted black circle (**b**) triangle-matrices, where the intensity of color indicates the level of confidence of two proteins which are functionally associated.


**Table 2.** List of the important pathways involved in ERM pathogenesis.
