Supplemental Figure 1. Recombination pattern of Six3-cre in the adult retina and brain. The conditional reporter line, R26R, commences expression of β-galactosidase upon Cre-mediated recombination. The recombination pattern of the Six3-cre transgene in the retina on P21 reveals widespread recombination in all layers (A-C). At the retinal margin, areas devoid of recombination are apparent, as depicted by gaps in the X-gal staining pattern in this region (C, arrowheads). Co- immunolabeling using anti-Isl1 and anti-β-galactosidase reveals that most Isll+ cells in the INL co-localize with Six3-cre's lineage (D-F), although several Isl1+ cells are devoid of detectable reporter expression

(D-F, arrowheads). At the retinal margin, many more Isl1+ cells devoid of detectable reporter expression are encountered (G-I, arrowheads). The recombination pattern of Six3-cre in the brain of P21 animals reveals widespread recombination in the hypothalamus, septum and striatum (J).

Scale bar equals 50 µm in B (applies to B-C), in F (applies to D-F), in I

(applies to G-I), and 1mm in J.

Supplemental Figure 2. Additional cell marker expression in the Isl1-null retina. Isl1-null retinas display a 71% reduction in the expression of the

RGC marker Pou4f1 (B) compared to control (A; average number of

Pou4f1+ cells ± standard deviation: controls: 35 ± 7, n=3; Isl1-nulls: 10

± 2, n=3, p<0.01, Student's t test). A slight 17% increase in the numbers of the horizontal cell marker, Calbindin-28K, in Isl1-null retinas is observed, although this change is not significant (compare C to D; average number of Calbindin-28K+ cells ± standard deviation: controls:

10 ± 2, n=4; Isl1-nulls: 12 ± 1, n=4, p=0.05, Student's t test). A 27% reduction in the Müller cell marker, Cdkn1b, is observed in the Isl1-null retina (F) compared to control (E; average number of Cdkn1b+ cells ± standard deviation: controls: 41 ± 4, n=3; Isl1-nulls: 30 ± 4, n=3, p<0.05, Student's t test). Toluidine blue stained sections revealed equivalent numbers of rods and cones in Isl1-null retinas (H) compared to controls (G), of which representative examples are highlighted by black and white arrowheads, respectively (average number of rods ± standard deviation: controls: 663 ± 28, n=3; Isl1-nulls: 660 ± 49, n=3, n.s.,

Student's t test; average number of cones ± standard deviation: controls:

30 ± 5, n=3; Isl1-nulls: 31 ± 2, n=3, n.s., Student's t test). Scale bar equals 50 µm in F (applies to A-F) and in H (applies to G-H).

Supplemental Figure 3. Remaining reporter-positive cells in Isl1-null retinas co-localize with residual Isl1 expression. Immunolabeling using anti-β-galactosidase demonstrates a substantial loss of reporter-positive cells in Isl1-null retinas (D) compared to controls (A). The remaining Isl1- immunoreactivity co-localizes with the remaining reporter gene expression (D, overlay in F), and represents incomplete deletion of Isl1 using the Six3-cre transgene in Isl1-nulls. Scale bar equals 50 µm in F

(applies to A-F).

Supplemental Figure 4. Partial retinal ganglion cell loss alone does not affect bipolar subtype development. Confocal micrographs of rod and cone bipolar marker expression in mature retinas from Pou4f2-null and control littermates (n=3). Prkca+ rod bipolar have similar characteristics in Pou4f2-null and controls, with prominent, high-intensity staining of rod bipolar endfeet in both (arrowheads in A and B). The number of rod bipolars is not significantly different in Pou4f2-null retinas compared to controls (average number of Prkca+ cells ± standard deviation: controls:

63 ± 6, n=3; Pou4f2-nulls: 55 ± 3, n=3, n.s., Student's t test). Isl1+ bipolar cells, which identify rod bipolar and ON-cone bipolar cells, are also not significantly different in Pou4f2-nulls compared to controls

(upper brackets in C and D; average number of Isl1+ bipolar cells ± standard deviation: controls: 108 ± 6, n=3; Pou4f2-nulls: 113 ± 10, n=3, n.s., Student's t test). Isl1+ cholinergic amacrine cells are not significantly different either (lower brackets in C and D; average number of Isl1+ amacrine cells ± standard deviation: controls: 17 ± 6, n=3;

Pou4f2-nulls: 15 ± 6, n=3, n.s., Student's t test). CABP5+ bipolar cells are not significantly changed in Pou4f2-nulls compared to controls (E and

F; average number of CABP5+ cells ± standard deviation: controls: 111 ±

10, n=3; Pou4f2-nulls: 105 ± 22, n=3, n.s., Student's t test). CABP5+ endfeet from OFF cone bipolar, ON cone bipolar and rod bipolar cells are all detectable in Pou4f2-null retinas (arrowheads in F). The OFF cone bipolar marker, Vsx1, is not significantly changed in Pou4f2-null retinas

(H) compared to control (G; average number of Vsx1+ cells ± standard deviation: controls: 40 ± 1, n=3; Pou4f2-nulls: 45 ± 14, n=3, n.s.,

Student's t test). Anti-Pou4f2 immunolabeling in Pou4f2-nulls confirms

'null' status of animals as no detectable Pou42 labeling is observed (J).

Scale bar equals 50 µm in J (applies to A-J).

Supplemental Figure 5. Co-localization of remaining Isl1+ cells with markers of bipolar and retinal ganglion cells in the Isl1-null retina. Co- immunolabeling using anti-Isl1 with various cell type markers were performed to assess the fate of remaining Isl1-expressing cells in the

Isl1-null retina. A predominant fraction of remaining Isl1+ cells (B) co- localizes with the pan-bipolar marker Chx10 (C, arrowheads in C).

Isl1+/Chx10- negative cells (C), positioned lower in the INL represent amacrine cells. As in the mature wildtype retina, Isl1+ bipolar cells co- localize with ON-bipolar markers, such as Goα (D-F, arrowheads in F).

The ON/OFF-bipolar marker CABP5 partially co-localizes with Isl1+ cells p(G-I). CABP5+ OFF-bipolar cells terminating nearest to the INL (asterisk in G and I) do not co-localize with Isl1, and CABP5+ ON-bipolar cells terminating closer to the GCL (white arrowheads in G-I) did co-localize with Isl1. Some Isl1+/CABP5- cells are also observed (yellow arrowheads in G-I). The ganglion cell marker Pou4f1 co-localizes with residual Isl1+ cells in the GCL (J-L). Scale bar equals 50 µm in L (applies to A-L).

Supplemental Figure 6. Immunolabeling for cleaved Caspase-3 in Isl1- null retinas. Fluorescence micrographs of anti-activated caspase-3 immunolabeling are shown for control and Isl1-null retinas from P5-P7.

Caspase-3+ cells in the INL were scored at these different ages. While numbers of caspase-3+ cells at P5 were not significantly different

(compare B to A; n=3), there was a significant reduction in the number of caspase-3+ cells at P6(n=3) and P7(n=3) in Isl1-nulls compared to controls (compare D to C and F to E). Summary of quantitation is shown in G. Scale bar equals 50 µm in F (applies to A-F). *p<0.05

Supplemental Figure 7. Residual Isl1 expression fails to establish Prkca+ rod bipolar with 'wildtype' morphology. 10 µm stacked confocal images of Prkca-labeled rod bipolar cells in control, Pou4f2-null and Isl1-null retinas were traced in Photoshop and analyzed subsequently in Image J.

Anti-Prkca and anti-Isl1 co-immunolabeling was performed in Isl1-null retinas to document Isl1 expression status. Areas of dendritic and terminal arborizations of Prkca+ rod bipolars in Isl1-null retinas are significantly decreased compared to control cells (compare C to A, summarized in D and E), except at the peripheral margin, where a clustering of Isl1+ cells is observed (compare C to A, summarized in D and E). Pou4f2-null rod bipolar cells on the other hand do not significantly differ from control cells in this respect (compare B to A, summarized in D and E). The length of the bipolar cell on the other hand is significantly decreased in both Isl1-null and Pou4f2-null retinas compared to control (compare B and C to A, summarized in F). Two- tailed Student t tests were performed adjusting the significance level to

0.017 from 0.05 because of the multiple comparisons. ****p<0.0001, ** p<0.01, n.s. p0.017. Control (n=6 cells); Pou4f2-null (n=5 cells); Isl1- null (n=14 cells from central retina; n=5 cells from peripheral retina).

Supplemental Figure 8. A reduction in late retinal progenitor proliferation does not accompany the reduction in restricted late born cell types in the Isl1-null retina. Fluorescence micrographs of anti-BrdU immunolabeling following 30-minute BrdU pulses at P5 and 7 in controls

(A,C,E,G) and Isl1-null retinas (B,D,F,H) were performed to gather whether reductions in proliferation of late retinal progenitors in mutants could underlie later reductions in cell numbers. While a large number of BrdU- incorporating cells are encountered in the peripheral retina of control (A) and Isl1-null retinas at P5 (B), markedly fewer BrdU+ cells are encountered in both controls (C) and Isl1-nulls centrally (D). The number of BrdU+ cells in Isl1-nulls, however, is greater in the central retina at P5

(number of BrdU+ cells: controls: 29 and 18, n=2; Isl1-nulls: 64 and 62, n=2). By P7, a marked down regulation of BrdU+ cells peripherally is observed as well, as evidenced by a trailing edge of BrdU+ cells at the peripheral margin (arrowheads in E and F). Despite the general reduction in BrdU+ cells in controls and Isl1-nulls, the trailing edge of BrdU+ cells in the periphery extends more centrally in mutants at P7, suggesting the general reduction in proliferation is delayed in the Isl1-null retina.

However, the increase in BrdU+ cells in the peripheral retina of Isl1-nulls at P7 observed did not reach significance (average number of BrdU+ ± standard deviation: controls: 43±10, n=3; Isl1-nulls: 67±25, n=3, p≤1,

Student's t test). Furthermore, this delay is transient, as central portions of the retina display a similar absence of BrdU-incorporating cells in both control (G) and Isl1-null retinas (H). Scale bar equals 50 µm in H (applies to A-H).

Supplemental Figure 9. BrdU pulse labeling reveals a bias for ONL cell fates in the postnatal Isl1-null retina. The fate of BrdU-incorporating cells following a BrdU pulse on P3 was evaluated on P12, when the majority of cells in the retina have been produced and proliferation has essentially ceased. The relative proportions of BrdU cells in the outer nuclear versus inner nuclear layers in control (A, n=3) and Isl1-null mice (B, n=3) were compared. While the majority of BrdU+ cells in controls reside in the inner nuclear layer (A, C), in Isl1-nulls, a significant shift in the fates of

BrdU+ cells to the outer nuclear layer is observed (B, C; χ2 = 9.31; 1,

N=6, p<0.01). In total, a 22% shift in the disposition of BrdU+ cells is observed in Isl1-null retinas. Total numbers of BrdU+ cells are not significantly different (C, n=3, Student's t test), however, consistent with BrdU incorporation not being significantly different at P3 in control and

Isl1-null retinas (Figure 1).

Supplemental Figure 10. Schematic summarizing the transcriptional regulation of major bipolar subtypes. (A) Factors known to specify bipolar cells are required for pan-bipolar marker expression. Isl1, not affecting the expression of pan-bipolar markers initially, likely acts downstream of bipolar specification, either before (B) or after (D,E,F) bipolar subtypes are specified. The coincident expression of Lhx3, Lhx4 and Isl1 is spatiotemporally poised to affect bipolar subtype establishment (C). The requirement for Isl1 in the differentiation of bipolar subtypes includes rod bioplar (D), ON-cone bipolar (E) and OFF-cone bipolar cells (F).