A new PIP-FUCCI construct that contained only the first 17 amino acids of Cdt1 linked to mVenus, along with amino acids 1–110 of Geminin linked to mCherry, was shown to precisely distinguish G1/S and S/G2 in U2OS cells [24], and here we examined primary endothelial cells expressing the reporter via lentivirus infection (Fig. 1A). Live image analysis revealed sharp transitions for Cdt11-17mVenus, with degradation at G1/S and re-expression at S/G2 (Fig. 1B–C, Supp. Movie 1), and mitosis always followed G2 in cells imaged to this stage transition. Analysis of multiple endothelial cells revealed that G1 phase averaged 8.1 h, S phase averaged 5.8 h, and G2 averaged 1.8 h, adding to a total endothelial cell cycle average of 15.7 h (Fig. 1D), in good agreement with total cell cycle transit of 15.9 h, as scored by time between mitoses of H2B-CFP expressing HUVEC (Fig. 1E). As described in Grant et al. [24], the degradation of Cdt11-17mVenus at G1/S was considered an exact measure of the start of S phase (as defined by formation of PCNA foci), and different cell lines sometimes showed a lag of Gem1-110 accumulation; a slight lag was documented in HUVEC, likely due to signal accumulation for Gem1-110. The G2/S ratio was 0.3, consistent with G2 being shorter than S in the cell cycle. Thus, primary endothelial cells regulate the PIP-FUCCI reporter in a temporal manner consistent with their cell cycle transit.

We next tracked labeled primary endothelial cells over time and asked whether cell cycle stage affected migration dynamics. We found that endothelial cells in G1 showed increased velocity and migrated distance/time (normalized to the average G2 interval) compared to cells in S-phase or G2 (Supp. Fig. 1A–C; Supp. Movies 2–4). This finding is consistent with another study showing enhanced G1 migration in cancer cells [32] and suggests that once G1 is complete, endothelial cell migration slows but does not completely stop under culture conditions, perhaps to accommodate activities supporting DNA synthesis and preparation for mitosis.

To generate a mouse carrying an inducible PIP-FUCCI allele, the PIP-FUCCI construct was placed 3’ to a standard cassette and built into the ROSA26 locus via CRISPR/Cas9-mediated insertion (Supp. Fig. 1D, see Methods). Mice that were either heterozygous or homozygous for the allele C57BL/6J-Gt(ROSA)26Sorem1(CAG-LSL-PIP-FUCCI)Vb/Vb, hereafter called PIP-FUCCI (PF/+ or PF/PF) were bred to Tg(Cdh5-cre/ERT2)1Rha (hereafter referred to as Cdh5-CreERT2) mice to excise the lox-STOP-lox cassette and induce reporter expression in endothelial cells of early post-natal retinas (Fig. 2A, Supp. Fig. 1E–F). Overlay of the PIP-FUCCI reporter signal with Isolectin B4 (IB4, vascular-specific, Supp. Fig. 2C–C’) or ERG (vascular endothelial-specific, Fig. 2B) staining of P6 retinal vessels showed that the PIP-FUCCI signal labeled only IB4- or ERG-positive cells, indicating endothelial cell-specific expression of the reporter in the vasculature in vivo.

PIP-FUCCI mouse reports cell cycle status in postnatal retinal vessels in vivo. A Breeding scheme and schedule for PF/PF;Cdh5-CreERT2/+ and PF/+;Cdh5-CreERT2/+ pups. B Representative images of one leaflet of a PF/PF;Cdh5-CreERT2/+ retina stained for ERG. Boxed areas in far left panel (scale bar, 200 μm) are magnified (scale bar, 100 μm) in middle (Angiogenic Front) and far right (Mature Region) panels. C Quantification of whole retina endothelial cell cycle phase analysis from PIP-FUCCI labeled P6 retinas stained for ERG (blue). n = 4 pups. D–F Representative images (D–E) and quantification (F) of P6 PIP-FUCCI labeled retinas stained for IB4 (blue) and EdU-labeled (purple or blue). E Representative cells (arrowheads) PIP-FUCCI labeled as in G1/G0 (green), S (red), or G2 (yellow) cell cycle phase, magnified from white boxed areas in D’ (top 3 rows of E) and another area of the same retina (bottom row). Scale bar (D) 100 μm; (E) 25  μm. F Indicated quantification, n = 3 pups. ****p < 0.0001 by Two-way ANOVA & Sidak’s multiple comparisons test comparing PF+EdU+ and all PF+ cells. G–I Representative images (G–H) and quantification (I) of PIP-FUCCI labeled P6 retinas stained for IB4 (blue) and Ki67 (purple). H Representative cells (arrowheads) PIP-FUCCI labeled as in G1/G0 (green), S (red), or G2 (yellow) cell cycle phase, magnified from white boxed areas in G’. Scale bar (G) 100 μm, (H) 25 μm. I Indicated quantification, n = 2 pups. ****p < 0.0001 by Two-way ANOVA & Sidak’s multiple comparisons test comparing PF+Ki67+ and all PF+ cells; ###p < 0.001, ##p < 0.01 comparing PF+Ki67- and All PF+ cells

To quantitatively analyze endothelial cell cycle in the postnatal retina with PIP-FUCCI, we created an ERG mask for all retinal endothelial cells (Supp. Fig. 1G) that allowed for determination of mVenus and mCherry nuclear intensity in each individual endothelial cell. We then developed a semi-automated image analysis protocol (Supp. Fig. 1H) that assigned each endothelial cell to a cell cycle phase (G1/G0, S, or G2) based on PIP-FUCCI reporter fluorescence. This analysis showed that most endothelial cells (82% for PF/PF homozygous retinas and 62% for PF/+ heterozygous retinas, Supp. Fig. 1I) were labeled, with the small proportion of unlabeled cells likely a combination of unexcised lox-STOP-lox cassette and a few cells in very early G1 or right at the G1/S transition. Unlabeled endothelial cells were not assigned a cell cycle status or used in the quantification. Manual validation of cell cycle phase assignment using a subset of endothelial cells showed that our pipeline was accurate 96% of time (Supp. Fig. 1J). Using this pipeline, we identified the overall proportion of endothelial cells in G1/G0 vs. S vs. G2 in the retina to be 86:9:5 (Fig. 2C), consistent with previous reports [11]. G1/G0 is longer than S and G2, as we found in cultured HUVEC (Fig. 1), although the retinal analysis revealed a lower ratio of G2 (0.06) or S (0.1) to G1/G0 cells, likely due to numerous endothelial cells that were in extended G1/G0 (quiescence) in vivo but not found in actively cycling cultured endothelial cells.

We next examined the relationship of the PIP-FUCCI reporter with EdU-labeling, which identifies cells in S-phase during the labeling period (Fig. 2D–F, Supp. Fig. 2A). Among PIP-FUCCI labeled retinal cells, mVenus+, mCherry− (green, G1/G0) endothelial cells were exclusively EdU−, showing that the Cdt1-mVenus reporter does not label S phase cells in vivo (Fig. 2D–D’ Fig. 2E top row, Fig. 2F, Supp. Fig. 2A). In contrast, mVenus−, mCherry + cells (red, S) dominated (97.9%) PIP-FUCCI-labeled EdU-labeled cells (Fig. 2D–D’, Fig. 2E second row; Fig. 2F), showing that the Gem-mCherry reporter faithfully labeled endothelial cells in S phase in vivo. While most (89%) of mCherry+ endothelial cells are EdU+, some mCherry+ cells score as EdU− (Supp. Fig. 2A), perhaps reflecting that detection of EdU labeling in tissues is likely not as sensitive as the expressed reporter. mVenus+, mCherry+ endothelial cells (orange, G2) were primarily EdU-, although a small proportion (5%) were EdU+ (Fig. 2D’, Fig. 2E bottom row, Fig. 2F, Supp. Fig. 2A). Since EdU labeling occurs over 2 h, some endothelial cells were likely labeled in S but transitioned to G2 prior to harvest.

Ki67 reactivity identifies cells in late G1, S, and G2, although expression is heterogeneous rather than uniformly negative in G0 and G1 [33]. To further validate PIP-FUCCI readouts, we examined the relationship of the PIP-FUCCI reporter with Ki67 (Fig. 2G–I; Supp. Fig. 2B). Among PIP-FUCCI labeled retinal endothelial cells, mVenus−, mCherry+ (red, S) and mVenus+, mCherry+ (orange, G2) endothelial cells were almost exclusively Ki67+, consistent with our expectation that these cells were in the cell cycle in vivo (Fig. 2G’; Fig. 2H second and bottom row; Fig. 2I, Supp. Fig. 2B). In contrast, PIP-FUCCI-labeled Ki67− cells were almost exclusively mVenus + , mCherry− (green, 99.2%) (Fig. 2G’; Fig. 2H top row green arrowheads; Fig. 2I), consistent with our prediction that mVenus labels G1/G0 endothelial cells. While a majority (76%) of mVenus+, mCherry− cells were Ki67− (Supp. Fig. 2B), indicating they are likely in G0 or early G1 in vivo, some (24%) mVenus + , mCherry- (green) endothelial cells were Ki67+ (Fig. 2G’; Fig. 2H top row, white arrowhead; Supp. Fig. 2B), likely due to the accumulation of Ki67 in late G1 and/or its perdurance in early G0.

Closer inspection of retinal images revealed that endothelial cells in the angiogenic front exhibited signal consistent with G1/G0 (green), S (red), or G2 (orange) cell cycle stages, while cells in the mature region were largely G1/G0 (green) (Fig. 2B; Supp. Fig. 2C–C’), suggesting spatial differences in the distribution of cell cycle stages. One advantage to postnatal retinal angiogenesis analysis is that a temporal gradient of remodeling (optic nerve outward) vs. angiogenic expansion (distal to remodeling) exists, with a clear definition of tip cells at the front vs. stalk cells behind the tip vs. non-tip/non-stalk angiogenic front cells right behind the tip/stalk area [4]. Spatial domains for large arteries, arterioles, and large veins and venules that form upon remodeling are also well-defined. To better understand how the cell cycle changes with time and vascular maturation, we developed a novel semi-automated zonation pipeline that identified the proportion of endothelial cells in G1/G0 vs. S vs. G2 in primary arteries (PA), Arterioles (Art), primary veins (PV), venules (Ven), mature capillaries (MC), stalk cells (Stalk), tip cells (Tip) and angiogenic front capillaries (AFC) (Fig. 3A (yellow dotted lines, zones used in pipeline; red boxes, regions shown at higher resolution in panel B); Supp. Fig. 2D, customized scripts on GitHub, see Methods).

Spatial analysis reveals cell cycle differences in different retinal vascular zones. A A representative ERG (white)/IB4 (blue) labeled P6 retina image. Yellow dotted lines and labels, vascular zones used for pipeline quantification of endothelial cells cycle status; yellow arrows, endothelial cells defined as tip/stalk cells; red boxes, areas shown with additional markers and resolution in panel B. Label definitions to right. Scale bar, 200 μm. B High resolution views with additional markers of areas in (A) denoted by red boxes. ERG + endothelial cells labeled mVenus+, mCherry− (G1/G0), green arrowheads/arrows; mVenus-, mCherry+ (S), red arrowheads/arrows; and mVenus+, mCherry+ (G2M), yellow arrowheads/arrows; in labelled retinal vascular zones. In Tip/Stalk panel: arrows, stalk cells; arrowheads, tip cells. White dotted lines, PA, PV, Art and Ven outlines. Scale bar, 100 μm. C Quantification of % mVenus-, mCherry+ (S phase), ERG+ endothelial cells across vascular zones. All reporter-labelled ERG+ cells from each retina were quantified. ****p < 0.0001, ***p < 0.001, **p < 0.01 by one-way ANOVA & Sidak’s multiple comparisons test. D Quantification of endothelial cells (ERG+) labeled G1/G0, S and G2M in retinal vascular zones. All reporter-labelled ERG+ cells from each retina were quantified. E–G Comparison of AFC, Stalk and Tip cells. E Quantification of % mVenus+, mCherry+, ERG+ (G2) endothelial cells. **p < 0.01, *p < 0.05 by one-way ANOVA & Sidak’s multiple comparisons test. F Ratio of endothelial cells in G2 to S, G ratio of endothelial cells in G2 to G1/G0 in AFC vs. tip cells vs. stalk cells. *p < 0.05, ***p < 0.001 by paired t test. n = 4 pups for all quantifications

As expected, the proportion of mVenus-, mCherry + (S phase) endothelial cells was elevated in primary veins and venules compared to arterial counterparts (Fig. 3B third row, red arrowheads in PA, PV, Art, Ven; Fig. 3C), consistent with reports that veins are more proliferative [2, 9]. Interestingly, primary arteries and arterioles consistently have a small proportion (~ 2%) of S-phase labeled endothelial cells (Fig. 3C), suggesting that a small proportion of arterial endothelial cells are in the cell cycle and not quiescent, consistent with a recent report showing that 2.5% arterial endothelial cells are mitotic in neonatal coronary arteries [34]. The highest S phase labeling was seen in the AFC (angiogenic front capillaries, 15.5%), which is significantly higher than the more mature MC zone (mature capillaries, 7%, Fig. 3C; Fig. 3B third row, red arrowheads in MC, AFC) and stalk cells (8.7%, Fig. 3C; Fig. 3B third row, red arrow in Tip/Stalk). Very few tip cells are mVenus−, mCherry+, (S phase, 1%, Fig. 3C; Fig. 3B third row Tip/Stalk), consistent with a previous report that tip cells do not usually divide [15].

Further analysis of the high-resolution zonation of cell cycle status in retinal endothelial cells revealed a surprising increase in the proportion of mVenus+, mCherry+ endothelial tip cells (orange, G2, 17.3%) and stalk cells (18.9%) (Fig. 3D–E) compared to neighboring endothelial cells in the angiogenic front (AFC, 8.4%) (Fig. 3E; Fig. 3B third row, yellow arrowheads and arrows in AFC and Tip/Stalk). Additionally, while other vascular zones had 1–3 times more endothelial cells in S than G2 phase (Fig. 3D), tip cells showed a significantly higher G2/S ratio compared to the angiogenic front capillaries (AFC) just behind the tip/stalk (9.3 in tip cells vs. 0.6 in AFC), while stalk cells showed a non-significant increase in this ratio (2.3) over AFC due to increased S-phase cells (Fig. 3F). This trend is also seen for both tip and stalk cell categories in the G2/G1 ratio comparison that is significantly higher in tip cells compared to AFC (0.22 in tip cells vs. 0.11 in AFC, Fig. 3G) and trending for stalk cells (0.28) vs. AFC, consistent with the idea that G2-enrichment of tip cells is not simply due to decreased S phase tip cells. Thus, a portion of endothelial tip and stalk cells may be held in a G2 arrest, which suggests co-ordination of the cell cycle and cell behaviors such as sprout extension.