BASIC RESEARCH www.jasn.org

Developmental Programming of Branching Morphogenesis in the

Rosemary V. Sampogna, Laura Schneider, and Qais Al-Awqati

Division of Nephrology, Department of Medicine, Columbia University College of Physicians and Surgeons, New York, New York

ABSTRACT The kidney developmental program encodes the intricate branching and organization of approximately 1 million functional units (). Branching regulation is poorly understood, as is the source of a 10-fold variation in number. Notably, low nephron count increases the risk for developing hypertension and renal failure. To better understand the source of this variation, we analyzed the complete gestational trajectory of mouse kidney development. We constructed a computerized architectural map of the branching process throughout fetal life and found that organogenesis is composed of two distinct developmental phases, each with stage-specific rate and morphologic parameters. The early phase is characterized by a rapid acceleration in branching rate and by branching divisions that repeat with relatively reproducible morphology. The latter phase, however, is notable for a significantly decreased yet constant branching rate and the presence of nonstereotyped branching events that generate progressive variability in tree morphology until birth. Our map identifies and quantitates the contribution of four developmental mechanisms that guide organogenesis: growth, patterning, branching rate, and nephron induction. When applied to organs that developed under conditions of malnutrition or in the setting of growth factor mutation, our normative map provided an essential link between kidney architecture and the fundamental morphogenetic mechanisms that guide development. This morphogenetic map is expected to find widespread applications and help identify modifiable targets to prevent developmental programming of common diseases.

J Am Soc Nephrol 26: 2414–2422, 2015. doi: 10.1681/ASN.2014090886

The mammalian kidney is composed of a ramified of the branching tree is needed. Here we report the network of epithelial tubes arranged in intricate and complete spatial-temporal trajectory of kidney de- reproducible patterns. The structure of the ureteric velopment throughout intrauterine life in the form tree is established via ongoing interactions between of a computerized architectural map. theuretericbud(UB)andmetanephricmesen- We developed a stereohistologic method that chyme (MM) (Supplemental Figure 1). Human utilizes confocal imaging and reconstruction tools kidney development occurs entirely in utero. Pre- to resolve three-dimensional branching struc- maturity, maternal stress, and nutrient deficiencies tures throughout the entire course of gestation at are among factors that increase risk of fetally pro- grammed disease.1,2 Brenner et al. postulated that such influences might reduce nephron count and Received September 12, 2014. Accepted December 3, 2014. numerous studies have since shown that low neph- Published online ahead of print. Publication date available at ron number is associated with increased risk of www.jasn.org. essential hypertension.3,4 Moreover, a marked 10- Correspondence: Dr. Rosemary V. Sampogna, Division of Ne- fold range in nephron number has been observed phrology, Department of Medicine, Columbia University College between various patient populations undergoing of Physicians and Surgeons, New York, 630 W. 168th Street, autopsy for unexpected death.5,6 To understand PH4-124, New York, NY 10032. Email: [email protected] the source of this variation, a comprehensive map Copyright © 2015 by the American Society of Nephrology

2414 ISSN : 1046-6673/2610-2414 JAmSocNephrol26: 2414–2422, 2015 www.jasn.org BASIC RESEARCH unprecedented resolution. Our developmental map identifies The former spans E11.5 to E15.5 and is characterized by a two distinct phases characterized by specific branching rates, burst in branching rate that peaks at E13.5. The late phase growth, and variation in the reproducibility of branching spans E15.5 to E19.5 and is defined by a constant and consid- pattern. A critical window is also identified when intrauterine erably slower rate. Overall, we calculate a 149-fold and a 5.6- stressors may maximally affect kidney development. Whereas fold increase in the number of new tips generated during the previous studies of organogenesis proposed genetic models early and late branching phases, respectively. The maximal that encode robust and stereotyped branching events,7,8 our branching rate falls within midpregnancy, a vulnerable period study outlines that the late developmental program, previously when critical structures are forming, and is consistent with not analyzed, specifies inherent asymmetry, variability, and studies that correlated midgestation starvation with impaired progressive complexity in branching architecture. These re- adult renal function.10,11 sults provide important new insights into the factors that de- termine nephron number and, more generally, that control Binary Stereotyped Branching Dominates the Early Branching branching rate and trajectory. Phase We applied our normative map to analyze organogenesis Within the context of these differential rates, we analyzed as- under various pathologic conditions. In each case, effects on sociated morphologic processes that shape organogenesis. A branching rate, branch segment number, tree morphology, and stereotyped, repeating bifurcating branching pattern predom- efficiency of nephron induction were deficiency specificand inates during the early phase. Specifically, initial branching quantifiable. Because congenital and intrauterine defects are a events are highly reproducible because most divisions produce common cause of renal failure, our detailed models of kidney two daughters. This is shown at E12.5 in the three-dimensional development and its sensitivity to environmental insults may tracing (Figure 3A) and in the form of a dendrogram (Figure 3B). help to identify therapeutic targets to aid organogenesis. The Most divisions are bifurcating (Supplemental Figure 2A) although temporal-spatial analysis presented here provides an essential some are trifurcating (Supplemental Figure 2B), consistent with link between kidney architecture and the fundamental de- previous reports.12 velopmental programs that regulate branching and nephron As branching rate rapidly accelerates during the early phase, formation throughout fetal life. not all parent nodes produce two daughters; a subset of pro- cesses will branch out to higher order. These notably populate the upper and lower poles, requiring a greater number of branch RESULTS iterations to extend farther from the origin. This is shown on day E13.5 (Figure 3, C and D) in which, within the branches colored A Map of Branching Morphogenesis throughout in blue, each and every node divides to produce two daughter Gestation segments (although occasionally three). Supplemental Figure 3 To analyze the sequence of events that forms the mammalian outlines a similarly sized tree that comprises a full binary tree kidney, weconstructedaspatialmap ofnormal kidneydevelop- defined by height h and (2h+121) vertices.13 Thus, significant ment throughout gestation. Morphologic measurements in- stereotypy is observed early on in age-matched kidneys, pri- cludingbranch segment length,diameter andvolume, branching marily because the majority of lower-order branching divisions angle, branching lineage and tree geometry were computed for utilize a bifurcating replication pattern. Segments that divide each specimen. beyond this full binary tree are shown to branch with more variability (in red). Some of these divisions are binary but re- Branching Rate Varies over Two Developmental Time Windows peat to higher order than others and some are trefoil. These set Figure 1 presents daily three-dimensional reconstructions of the stage for progressive asymmetry required to sculpt the ir- the ureteric tree and provides a trajectory of branching mor- regular kidney shape. phogenesis throughout intrauterine development. Relevant parameters are summarized in Figure 2A and Supplemental The Developmental History of Kidney Modularity Table 1. We detected 587 tips at E15.5 compared with similar From an anatomic perspective, kidneys are modular. Human reports of 362 and 534 that utilized immunofluorescence mi- studies have outlined the existence of six and four conserved croscopy and optical projection tomography, respectively.8,9 lobes in the upper and lower poles, respectively.14 Our results These differences may arise from variation in genetic back- show that this modularity can be traced to early branching ground, imaging, and counting techniques. In addition, the events. In all kidneys studied, a specific branch consistently standard overnight mating period introduces slight age variation. arises from the upper main stem and establishes gross asym- To analyze intrauterine branching rates, we calculated dif- metry by E12.5 (Figure 1, Supplemental Figure 2C). This buds ferences in tip numbers between consecutive days and nor- off in a lateral orientation from a preexisting segment instead malized these over gestation (see the Concise Methods). The of a tip, a motif previously described in culture and kidney resultant biphasic rate curve is composed of two distinct de- explants.12 Branch descendants can be traced from E12.5 to velopmental windows of approximately 4 days each, which we E13.0 to E19.5 (Supplemental Figure 4) as this segment de- designate as the early and late branching phases (Figure 2B). velops into a lateral lobe (yellow). Overall, the developmental

J Am Soc Nephrol 26: 2414–2422, 2015 The Kidney Branching Program 2415 BASIC RESEARCH www.jasn.org

Figure 1. Reconstruction of the ureteric tree throughout intrauterine development. One representative three-dimensional tracing is shown for each gestational day. Each successive branching generation is represented by a different color. Note that the color scheme is repeated after 14 branching orders starting at E15.5. Scale bar, 100 mm. program of the kidney provides instructions that generate number and segment lengths over terminal branch orders in gross asymmetry. E13.5 littermate (identically aged) kidneys. Low variance at orders 1–5 is due to the dominant binary stereotyped pattern, but increas- Stochastic Events Modulate Patterning during the Late ing noise between 6 and 10 reflects greater variation in node oc- Branching Phase cupancy. Corresponding branches are characterized by increasing The late branching phase initiates at E15.5 with a marked decrease stochasticity in developmental pattern. More detailed comparison in branching rate that remains active over an increasing population of these littermate kidneys shows that terminal branching patterns of tips. This precludes division of each and every tip. Stochasticity are similar but not identical (Figure 4, Supplemental Figure 5). in branching pattern becomes more prominent during this phase Even right and left kidneys dissected from a single embryo display as terminal segments are progressively added in variable patterns unique branching patterns (Supplemental Figure 6). Thus the ure- and play a role in sculpting the kidney’suniqueshape.Thecurvein teric tree, particularly at the periphery, is composed of an array of Figure 3E shows that throughout intrauterine life, the maximum asymmetrically branched processes of varying order, length, and branching order of segments that comprise a full binary tree lineage. In other words, there does not appear to be a “wave” of follows a linear pattern (blue curve). However, the maximum uniform branching per tip. Rather, each process (from origin to branching order of segments that are added in variable spatial tip) is characterized by its own specific morphology. patterns begins to outweigh the progress of the full binary tree at E15.5 (red curve), notably timed with onset of the late phase. Nephron Induction Occurs in the Late Branching Phase An example of the degree of variability is outlined in Figure Mature glomeruli (a marker for nephron formation) begin to 3, F and G, and shows increasing statistical variance in node appear at E13.5, a delay that reflects UB and MMcomaturation.

2416 Journal of the American Society of Nephrology J Am Soc Nephrol 26: 2414–2422, 2015 www.jasn.org BASIC RESEARCH

Figure 2. Growth and branching rates throughout gestation. (A) Growth curves for total tip and count at each day (semi-log2 plot). The rate of glomerulus formation lags behind tip formation. Vertical bars represent variance. (B) Observed tip branching probability per day calculated from the time course of the mean in A. Early and late branching phases are indicated.

Average branch tip to glomerulus stoichiometry is 5:1 until E19.5, the curve is characterized by minimal variability simply because when approximately 3000 tips and 1000 morphologically mature most of the branching events are dichotomous and almost each glomeruli are formed. At birth, the tip count in newborn mice and every node produces two daughter segments. At higher is estimated by extrapolation to be 4000. This is consistent with orders, total branch numbers smoothly decline reflecting the other estimates at E19.5 ranging from 3000 clusters of con- decreasing quantity of high (from origin to tip) order branch densed mesenchyme8 to 8000 glomeruli in all stages from the processes. The area under the curve (AUC) is calculated by S-shaped body to maturity.9 summing the number of branches over each order and repre- sents the total number of branch segments in the entire tree. Global Remodeling and Sculpting of the Tree Occurs during the From this family of curves, we are able to extract growth and Late Branching Phase branching rates, the number and distribution of branch seg- At E15.5, a central cavity forms at the origin (Supplemental ments per generation, and maximal branching order. Figure 7A), coinciding with initiation of urine flow.15 Supple- mental Figure 7B shows simultaneous proliferation and apo- Application of the Normative Map to Mutation and ptosis as the initial branch segments are remodeled. By E16.5, Maternal-Fetal Malnutrition segments comprising the first two branching orders begin to To explain the observed variation in human nephron number be resorbed into the expanding (Supplemental and to determine whether our model is consistent with known Figure 7C). At E19.5, further erosion creates seven indepen- nephron defects, we assayed effects of genetic and nutritional dent subtrees (Supplemental Figures 4 and 7D), indicating that deficiencies known to affect nephron number at E15.5, an age approximately three orders (23) have undergone significant when major features have formed and branching rate has remodeling. stabilized. We analyzed induced changes in growth curves, Despite the finding that branching rate markedly decreases branch lineage, and complete geometries in an attempt to cor- during the late phase, the kidney’s upper-to-lower pole length relate specific morphogenetic abnormalities with mechanisms increases 3-fold. On subsequent days, differential elongation of branching and nephron formation. of low-order branch segments was observed (Supplemental Figure 7E). This directed growth in the longitudinal axis is Fibroblast Growth Factor-7 Knockout timed to rapid cranial-caudal elongation of the fetal body.16 Fibroblast growth factor-7 (FGF7) plays a key role in branching and nephron number specification.17,18 We found that gross Mapping of Growth Curves throughout Kidney Development organ shape and relative branch segment lengths and angles To delineate the developmental trajectory, we mapped the are maintained in mutants, thereby preserving wild-type ge- growth of the branching network throughout gestation. Normal ometry. The major defects, however, arise from a delay in de- growth curves for four select ages are analyzed in Figure 5A. In velopment. The traced ureteric tree is compared with the wild each, the left curve edge shows a rapid rise in the total number of type in Figure 5B. Although the overall shape of the mutant segments per order (n=order, slope approximately 2n). Thus, the and wild-type growth curves are alike (Figure 5C), the FGF7(2/2) subset of the tree represented within this exponential portion of profile (red) lies directly between the E14.5 (blue) and E15.5

J Am Soc Nephrol 26: 2414–2422, 2015 The Kidney Branching Program 2417 BASIC RESEARCH www.jasn.org

Figure 3. Stereotyped and variable branching patterns. (A) Three-dimensional tracing representing the topology and lineage of binary (blue) and variable (red) branching patterns at E12.5. (B) Dendrograms are tree diagrams that are generated from the manual tracing of each specimen that outline branch lengths and lineage. Horizontal lines represent measured branch lengths, and intersections represent branch nodes. Vertical line lengths are arbitrary. Numbers represent centrifugal order assignment of nodes where the root is shown by order 0. Early branching events follow a stereotyped pattern such that, on average, two daughters are produced per division. These (shown in blue) comprise a full binary tree (also see Supplemental Figure 3). (C) The tracing at E13.5 localizes blue segments to the interior and red to the periphery and tips. (D) Representative dendrogram at E13.5. Segments shown in red represent branch processes outside the full binary tree that are composed of asymmetric and variable patterns. (E) The graph shows maximum branch orders occupied by the stereotyped and variable patterns during gestation. At each day, the height (maximum order) of the full binary tree generated by stereotyped branching events increases linearly (blue). However, the curve for subtrees that divide beyond the full binary tree significantly increases at the beginning of the late branching phase (red). This slope reflects progressive contribution of variable pattering until birth. (F) Average total node number per order in four kidneys taken from littermates at E13.5. Ureteric tree branching for orders 1 to about 5 primarily form by bifurcation of each tip and show an exponential growth phase. Thereafter, variance (vertical bars) increases as branches from orders 6–10 occupy increasingly different configurations. (G) The curve for branch segment lengths summed over each order shows analogous shape and behavior. Similar curves were found on all gestational days (data not shown).

(gray) wild-type curves. The age of this mutant is 0.5 days time via deceleration in branching rate, a finding that behind the normal trajectory, consistent with a primary devel- supports a timing function for FGF7. The equivalent decrease opmental delay. Comparable shapes of all curves reflects that in glomerular number suggests that each tip, although late in relative tree geometries and proportions are maintained in the appearing, is equally potent in inducing nephrons. mutant; the number of glomeruli is reduced by 30%, the volume of the ureteric tree by 25%, and the total number of branch Vitamin A Deficiency segments (AUC) by 25% in the mutant (P,0.01 for all three). Kidney development requires normal retinoid levels and low Most notably, the branching rate is significantly reduced, be- fetal vitamin A status is associated with decreased human kidney cause the average maximum branch order at E15.5 is 12 in the size and nephron number.19 To examine branching effects, we mutant as compared with 15 in the wild type (P,0.01). used a previously published protocol to induce vitamin A de- These data indicate that the observed decrease in nephron ficiency in rats to levels compatible with gestation.20 Animals number in FGF7(2/2)arisesfromaprimarylossofdevelopmental were maintained on a low vitamin A diet for 3 weeks before

2418 Journal of the American Society of Nephrology J Am Soc Nephrol 26: 2414–2422, 2015 www.jasn.org BASIC RESEARCH

Figure 4. Symmetry, asymmetry, and stochasticity in kidney development. (A) Dendrogram for E13.5 wild-type kidney showing lineage of each tree. Colored ellipses outline four lobes in the upper and three lobes in the lower half of the kidney. These are defined by clusters of associated nodes and branch segments, are conserved in all kidneys studied, and correlate with kidney anatomy. (B) The dendrogram adjusted radially to show two-dimensional lobe arrangement. (C) The average number of tips present within each lobe is indicated. n=4 kidneys, each extracted from each of four (identically aged) littermates at E13.5. mating then throughout pregnancy. At E15.5, we found that during pregnancy and found that E15.5 kidneys from protein- despite significant growth impairment, normal branching rate restricted dams have a 50% decrease in total segment number, is maintained (Figure 5, B and C). Although tree volume is de- 75% decrease in glomerular number, and 90% decrease in tree creased by 32% (P,0.01), there is no significant difference in the volume (P=0.001, P=0.001, and P,0.001, respectively) (Figure total number of branch segments (AUC) and in branching rate 5, B and C). However, there is no significant difference in the compared with the wild type. However, maximum branching maximum branch order. Although the stereotyped binary order increases from 15 to 18 (P,0.05) accompanied by a branching pattern remains intact over the first 5 orders, there change in shape of the normal growth profile (Figure 5C). The is a marked loss in the sum of branches per order between 7 and initial curve shows that stereotyped binary branching is dimin- 12 (Figure 5C). The tracing (Supplemental Figure 8C) shows ished to four generations compared with seven in the wild type that although the origin-to-tip order of branch processes (P,0.05), indicating a significant shift toward increased pattern that span the longitudinal axis remain unchanged, there is a variation. directed reduction in maximal order in those that delineate Most importantly, despite maintenance of normal branch- circumference. The latter branch to fewer generations and ing rate and segment number, glomerular count is decreased by consequently have a significant reduction in both total length 55% (P=0.002) (consistent with animal studies19,21). Im- and in pattern variation. paired ability of each tip to induce glomeruli is possibly due These gross anatomic changes correlate with human ul- to retinoid-dependent signals that act directly on the UB or, trasonography studies that detected “sausage-shaped” kidneys based on reciprocal molecular signaling, indirectly via the of reduced diameter in growth-restricted fetuses.26 The pres- stroma and MM.20,22,23 Such effects also might contribute to ence of distinct longitudinal and circumferential growth axes increased variability in branching pattern. during development also has been demonstrated.9 Because the radius of each branch normally tapers with each successive Protein Deficiency division, the decrease in overall branch order also results in Protein deficiency is known to adversely affect kidney size.24 the loss of more finely branched segments (Supplemental Fig- Starvation during midgestation in humans resulted in low ure 8, A and B). In summary, the generalized insult of protein birth weight (a marker for low nephron number) and in- deficiency affects most, if not all, of the steps that specify creased risk of proteinuria in adults.1,25 We placed animals branching morphology and complexity, nephron number, onalowproteindietfor3weeksbeforematingandthen growth axes, and organ size.

J Am Soc Nephrol 26: 2414–2422, 2015 The Kidney Branching Program 2419 BASIC RESEARCH www.jasn.org

Figure 5. Growth curves and architecture of wild-type and mutant kidneys. (A) Wild-type segment distribution at embryonic days E13.5, E15.5, E17.5, and E19.5. The total number of branch segments per order is plotted. The left edge of each curve increases exponentially between orders 1 and 6 for E13.5, 1 and 7 for E15.5, 1 and 8 for E17.5, and 1 and 9 for E19.5. The right tail represents the average maximal number of branching generations at each age. The AUC represents the total number of segments per kidney. Characteristic curve shape is maintained at each day of gestation. (B) E15.5 ureteric tree tracings for each condition (to scale). N is the maximum generation number and glom represents the mean glomerular count (minimum of three kidneys studied per condition). In each case, glomerular count is reduced compared with the wild type. The number of branching generations is specific to each condition. (C) Growth curves for each deficient condition are in red at E15.5. The E15.5 wild type is shown in gray. In the case of the FGF7 mutant, the E14.5 wild type is shown in blue. WT, wild type; Vit, vitamin.

DISCUSSION that guide organogenesis. During the early phase, stereotyped branching divisions occur at maximal rate and construct the Epidemiologic studies have shown that fetal programming plays a core scaffolding. This denotes a vulnerable gestational period significant role during kidney development and progression of when environmental influences can have profound implications disease later in life.3,4 Here we present a high-resolution map on organ structure and nephron number. During the late that analyzes the trajectory of normal kidney development phase, a marked increase in tip population is accompanied throughout gestation. We identify two distinct temporal phases, by a significantly reduced branching rate. This relative mis- each characterized by stage-specific morphogenetic mechanisms match prohibits division of each and every tip and comparison

2420 Journal of the American Society of Nephrology J Am Soc Nephrol 26: 2414–2422, 2015 www.jasn.org BASIC RESEARCH of age-matched ureteric trees shows variation in spatial overlap morphogenetic effects of different insults. Growth is affected in specifically in peripheral branching pattern. As branch and tip all, but the reduction in glomerular number is condition specific. numbers multiply during late gestation, this variation became These developmental perturbations can be attributed to sepa- more prominent. The degree of pattern variability seen at E18.5 rable morphogenetic processes that include primary develop- and E19.5 was based on examination of three specimens each, mental delay, defects in nephron induction, changes in growth but loss of stereotyped branching appears much earlier in de- axis, and alterations in balance between stereotyped and sto- velopment. Indeed, this same result also was found in kidneys at chastic processes. These produced up to a 3-fold decrease in glo- E13.5 and older (where more specimens were examined), also meruli and 2-fold decrease in total branching events. Although suggesting a breakdown of stereotyped pattern. With the caveat the range in human nephron number is even larger, we now that sample size may be small at any one stage, we believe that have two examples for genetically identical mice and one FGF7 this level of variation will continue to be observed. knockout model in which nephron number is significantly We found that initial stereotyped branching events transition variable. Together our findings support the notion that genetic into a stochastic branching process that becomes increasingly defects and environmental abnormalities such as fetal nutrient dominant during the second half of organogenesis. Short et al. delivery modulate final nephron count. The varied response to recently concluded that kidney architecture is structurally ste- these different insults suggests existence of a number of definable reotypic,8 a result that is consistent with our findings during mechanisms that control organogenesis. Most generally, we be- early development. They also noted that when comparing age- lieve that our model will facilitate correlation between organ matched kidneys between E11.75 and E15.5, the average tree architectureandthegeneticmechanismsthatregulateunderlying fraction contained within the maximum common subtree was developmental pathways. 0.8560.11 (1 represents perfect overlay). This degree of variabil- ity is also consistent with our findings during early organogen- esis. Although limitations in resolution using optical projection CONCISE METHODS tomography prohibited their analysis of the complete ureteric tree past E15.5, we were able to study global branching events at The Columbia Institutional Animal Care and Use Committee ap- later ages. It is during late gestation, notably between E15.5 and proved all animal procedures. Embryonic kidneys were obtained from E19.5, that we found increasing levels of pattern variation. Nat- timed C57BL/6 or FGF7 mutant mice (Jackson Laboratories). The ural spatiotemporal variation in lung branching configuration, morning of plug detection was designated E0.5 and embryos were particularly at later generations, has also been reported.27,28 harvested at noon. To preserve organ structure, intact embryos were The map presented here provides a refined quantitative assay fixed with 4% paraformaldehyde, equilibrated in 30% sucrose, snap- that outlines the entire trajectory of intrauterine kidney de- frozen in Tissue-Tek OCT compound (Sakura Finetek), and then velopment. It defines key developmental periods and associated sectioned to 150 mm using a Leica CM3050S cryostat. Nutrient- morphogenetic mechanisms that guide organogenesis. Although deficient diets (Harlan) were initiated 3 weeks before mating and con- the genetic program of epithelial branching has not yet been tinued throughout gestation; the diets contained trace vitamin A demonstrated, our analysis identifies specific time points that (TD.86143) and 6% protein (TD.90016). The control standard rodent mark transitions between major stages of organogenesis29,30 and diet contains 24.3% crude protein and 12.6 IU/g vitamin A (TD.8604). may guide further studies. Indeed, evolving global gene expres- Sections were blocked (PBS, 5% donkey serum, 0.1% Triton X-100) sion patterns during kidney development have been shown to for 1 hour at room temperature, incubated overnight at 4°C in mark critical transitions between separable developmental cytokeratin-8 (DSHB) and podocalyxin (R&D Systems) primary anti- stages.31 Our approach should also be applicable to study other bodies, washed with PBS, and were then incubated with Alexa Fluor organs that develop via repetitive branching and make tractable secondary antibodies (Invitrogen). Apoptosis was detected with the rigorous analysis of morphologic events at any age. Although we InSitu Cell Death Detection Kit (Roche) and proliferation with Ki-67 present a complete map of intrauterine kidney development, it antigen (DAKO). Z-stacks were obtained using a Zeiss LSM510 is important to note that in rodents, a significant portion of META scanning confocal microscope (Supplemental Figure 9). Succes- nephrogenesis continues for a short time after birth and .50% sive stacks were aligned and concatenated with the ImageJ34 Image of nephrons form after cessation of branching.8,32 In humans, Layering Toolkit (Supplemental Figure 10A). Reconstructed images Oliver discovered that the pattern of branching changes signif- were imported into Neurolucida (MBF Bioscience), manually traced, icantly during late nephrogenesis.33 During this period, bifid and analyzed (Supplemental Figure 10B). branching gives way to division of single terminal branches The number of wild-type kidneys analyzed per day is five for E11.5 that lead to the formation of many nephrons by lateral branch- (limb stage [ls] 5–7), five for E12.5 (ls8–9), five for E13.5 (ls10), six for ing. Further work may reveal whether the level of variation E14.5 (ls11), five for E15.5 (ls12), three for E16.5 (ls13), three for increases or stabilizes after birth. E17.5 (ls13), three for E18.5 (ls13), and three for for E19.5 (ls13). We used our normative map as a basis for examining kidneys Three to seven kidneys were studied for each deficiency. Sex of the that developed under three pathologic conditions and identified embryos was not determined. Data were analyzed by the t test. a spectrum of abnormalities. Our analysis is consistent with The observed daily branching probability was calculated using the previous studies but provides much new insight into specific following formula:

J Am Soc Nephrol 26: 2414–2422, 2015 The Kidney Branching Program 2421 BASIC RESEARCH www.jasn.org

= 15. Airik R, Kispert A: Down the tube of obstructive nephropathies: The ¼ niþ1 ni pi 9 importance of tissue interactions during development. Kidney ∑ n þ n i¼1 i 1 i Int 72: 1459–1467, 2007 16. Schoenwolf GC, Bleyl SB, Brauer PR, Francis-West PH: Larsen’s Human , Philadelphia, Churchill Livingstone/Elsevier, 2008 where ni+1/ni compares fold difference in tip number between days i+1 and i. Day 1 represents comparison of day E11.5 (two tips) to the initial 17. Qiao J, Uzzo R, Obara-Ishihara T, Degenstein L, Fuchs E, Herzlinger D: – FGF-7 modulates growth and nephron number in the ureteric budding event (n=one tip). Days 2 9 represent subsequent developing kidney. Development 126: 547–554, 1999 daysthroughE19.5.Ateachtimepoint,foldchangeintipsisnormal- 18. Qiao J, Bush KT, Steer DL, Stuart RO, Sakurai H, Wachsman W, Nigam ized by the sum of day-to-day fold difference throughout gestation. SK: Multiple fibroblast growth factors support growth of the ureteric bud but have different effects on branching morphogenesis. Mech Dev 109: 123–135, 2001 fl ACKNOWLEDGMENTS 19. Merlet-Bénichou C: In uence of fetal environment on kidney de- velopment. Int J Dev Biol 43: 453–456, 1999 20. Lelièvre-Pégorier M, Vilar J, Ferrier ML, Moreau E, Freund N, Gilbert T, We thank J. Barasch, F. Costantini, B. Levin, A. Monge, and J. Ross for Merlet-Bénichou C: Mild vitamin A deficiency leads to inborn nephron input and comments. deficit in the rat. Kidney Int 54: 1455–1462, 1998 This study was supported by a Research Career Award from the 21. Moreau E, Vilar J, Lelièvre-Pégorier M, Merlet-Bénichou C, Gilbert T: National Institute of Diabetes and Digestive and Kidney Diseases Regulation of c-ret expression by retinoic acid in rat metanephros: Implication in nephron mass control. Am J Physiol 275: F938–F945, (K08-DK078014) and a Carl W. Gottschalk Research Scholar Grant 1998 from the American Society of Nephrology (both to R.V.S.). 22. Batourina E, Gim S, Bello N, Shy M, Clagett-Dame M, Srinivas S, Costantini F, Mendelsohn C: Vitamin A controls epithelial/mesenchy- mal interactions through Ret expression. Nat Genet 27: 74–78, 2001 DISCLOSURES 23. Rosselot C, Spraggon L, Chia I, Batourina E, Riccio P, Lu B, Niederreither None. K, Dolle P, Duester G, Chambon P, Costantini F, Gilbert T, Molotkov A, Mendelsohn C: Non-cell-autonomous retinoid signaling is crucial for re- nal development. Development 137: 283–292, 2010 24. Moritz KM, Singh RR, Probyn ME, Denton KM: Developmental pro- REFERENCES gramming of a reduced nephron endowment: More than just a baby’s birth weight. Am J Physiol Renal Physiol 296: F1–F9, 2009 1. Luyckx VA, Bertram JF, Brenner BM, Fall C, Hoy WE, Ozanne SE, Vikse 25. Bertram JF, Douglas-Denton RN, Diouf B, Hughson MD, Hoy WE: BE: Effect of fetal and child health on kidney development and long- Human nephron number: Implications for health and disease. Pediatr term risk of hypertension and kidney disease. Lancet 382: 273–283, Nephrol 26: 1529–1533, 2011 2013 26. Konje JC, Okaro CI, Bell SC, de Chazal R, Taylor DJ: A cross-sectional 2. Bagby SP: Developmental origins of renal disease: Should nephron study of changes in fetal renal size with gestation in appropriate- and protection begin at birth? Clin J Am Soc Nephrol 4: 10–13, 2009 small-for-gestational-age fetuses. Ultrasound Obstet Gynecol 10: 22– 3. Brenner BM, Garcia DL, Anderson S: Glomeruli and blood pressure. 26, 1997 Less of one, more the other? Am J Hypertens 1: 335–347, 1988 27. Boyden EA: A synthesis of the prevailing patterns of the broncho- 4. Barker DJ, Bull AR, Osmond C, Simmonds SJ: Fetal and placental size pulmonary segments in the light of their variations. Dis Chest 15: 657– and risk of hypertension in adult life. BMJ 301: 259–262, 1990 668, 1949 5. Puelles VG, Hoy WE, Hughson MD, Diouf B, Douglas-Denton RN, 28. Blanc P, Coste K, Pouchin P, Azaïs JM, Blanchon L, Gallot D, Sapin V: A Bertram JF: Glomerular number and size variability and risk for kidney role for dynamics in mouse lung branching morpho- disease. Curr Opin Nephrol Hypertens 20: 7–15, 2011 genesis. PLoS ONE 7: e41643, 2012 6. Hoy WE, Ingelfinger JR, Hallan S, Hughson MD, Mott SA, Bertram JF: 29. Shah MM, Sampogna RV, Sakurai H, Bush KT, Nigam SK: Branching The early development of the kidney and implications for future health. morphogenesis and kidney disease. Development 131: 1449–1462, J Dev Orig Health Dis 1: 216–233, 2010 2004 7. Metzger RJ, Klein OD, Martin GR, Krasnow MA: The branching pro- 30. Nigam SK, Shah MM: How does the ureteric bud branch? JAmSoc gramme of mouse lung development. Nature 453: 745–750, 2008 Nephrol 20: 1465–1469, 2009 8. Short KM, Combes AN, Lefevre J, Ju AL, Georgas KM, Lamberton T, 31. Tsigelny IF, Kouznetsova VL, Sweeney DE, Wu W, Bush KT, Nigam SK: Cairncross O, Rumballe BA, McMahon AP, Hamilton NA, Smyth IM, Analysis of metagene portraits reveals distinct transitions during kidney Little MH: Global quantification of tissue dynamics in the developing organogenesis. Sci Signal 1: ra16, 2008 mouse kidney. Dev Cell 29: 188–202, 2014 32. Rumballe BA, Georgas KM, Combes AN, Ju AL, Gilbert T, Little MH: 9. Cebrián C, Borodo K, Charles N, Herzlinger DA: Morphometric index of Nephron formation adopts a novel spatial topology at cessation of the developing murine kidney. Dev Dyn 231: 601–608, 2004 nephrogenesis. Dev Biol 360: 110–122, 2011 10. Roseboom T, de Rooij S, Painter R: The Dutch famine and its long-term 33. Oliver J: Nephrons and Kidneys: A Quantitative Study of De- consequences for adult health. Early Hum Dev 82: 485–491, 2006 velopmental and Evolutionary Renal Architectonics, New York, Hoeber 11. Schulz LC: The Dutch Hunger Winter and the developmental origins of Medical Division, Harper & Row, 1968 health and disease. Proc Natl Acad Sci U S A 107: 16757–16758, 2010 34. Schneider CA, Rasband WS, Eliceiri KW: NIH Image to ImageJ: 25 years 12. Watanabe T, Costantini F: Real-time analysis of ureteric bud branching of image analysis. Nat Methods 9: 671–675, 2012 morphogenesis in vitro. Dev Biol 271: 98–108, 2004 13. Rosen KH: Graph Theory and Its Applications, Boca Raton, FL, Chapman and Hall/CRC Press, 2006 14. Jennette JC, Olson JL, Schwartz MM, Silva FG: Heptinstall’s Pathology This article contains supplemental material online at http://jasn.asnjournals. of the Kidney, Philadelphia, Lippincott Williams and Wilkins, 2007 org/lookup/suppl/doi:10.1681/ASN.2014090886/-/DCSupplemental.

2422 Journal of the American Society of Nephrology J Am Soc Nephrol 26: 2414–2422, 2015 SUPPLEMENTAL MATERIAL

Branching Parameters During Kidney Development Embryonic Day Maximum Number Total Tip Number Total Glomeruli Of Generations (±SD) (±SD) E11.5 1 2 0 E12.5 5 ± 1 11 ± 1 0 E13.5 8 ± 1 116 ± 13 11 ± 2 E14.5 10 ± 1 298 ± 36 59 ± 12 E15.5 15 ± 1 587 ± 61 101 ± 22 E16.5 16 ± 1 873 ± 140 176 ± 34 E17.5 19 ± 2 1232 ± 256 228 ± 50 E18.5 22 ± 2 1772 ± 196 361 ± 82 E19.5 27 ± 3 3271 ± 425 956 ± 84

Table S1. Branching parameters during kidney development. The average number of maximal branching generations, total tip number and total glomerular number are shown for each embryonic day. N=5 for E11.5, N=5 E12.5, N=5 E13.5, N=6 E14.5, N=5 E15.5, N=3 E16.5, N=3 E17.5, N=3 E18.5 and N=3 for E19.5. Standard deviation (SD) is shown for total tip number and total glomeruli.

Figure S1. Schematic representation of kidney development and organization. Mouse intrauterine kidney development occurs between embryonic days 11.5 and 19.5. At E11.5, the ureteric bud (UB in green) is an outpouching that arises from the Wolffian duct and invades an adjacent collection of loose epithelial progenitor cells called the metanephric mesenchyme (MM in blue). Reciprocating signals induce the UB to branch while the MM simultaneously condenses at UB tips, epithelializes and organizes into nephron segments. Repeating UB branching events form the ureteric tree. One mature nephron is shown on the right with the ureteric tree in green and MM-derived structures in blue (proximal nephron) and red (glomerulus). Glomeruli are capillary networks that initiate the filtration process required to maintain homeostasis. In the adult kidney, on average one million nephrons are packed into a complex arrangement that also includes a parallel vascular supply. The final nephron count in the adult kidney is highly variable and reflects the efficiency of this branching and maturation process.

Figure S2. Bifurcating, trifurcating and lateral branching mechanisms. (A) Projection of confocal image stacks (generated using LSM Image Browser) showing the surface of an age E15.5 kidney. Circled areas indicate terminal branches that bifurcate. The average local branch angle is 56.3o +/- 27.7o. (B) Projection of E13.5 kidney demonstrates trifurcations at the surface. (C) Tracing of the E12.5 ureteric tree demonstrates a lateral branch (*) that arises from the upper main stem (order number 1 is shown in red). This branching event was preserved in every kidney studied and established the first element of asymmetry between upper and lower poles.

r

1

2

3

4

Figure S3. Definition of a complete binary tree. The tree shown has height h = 4. All circles represent vertices (or nodes) and the open circle represents the root that corresponds to the first division of the ureteric bud. Every level is filled in the figure shown, thus forming a complete binary tree with height h and 2h+1 -1 vertices.

Figure S4. Persistence of the first lateral branch throughout gestation. The first lateral branch noted at E12.5 is circled (yellow). Descendants of this branch are shown at E13.0 within the yellow ellipse and at E19.5 by the subtree traced in yellow. The color- coded lobular arrangement seen at E13.0 remains conserved throughout E19.5. At E19.5, each color represents the corresponding individual tree with independent origin that is established by remodeling and pruning of the first several branch orders.

Figure S5. Branching dendrograms of littermates at E13.5. Dendrograms show that basic kidney scaffolding and upper vs. lower pole asymmetry are conserved in kidneys extracted from three littermates. Patterns within ellipses demonstrate variability in node number, segment lengths and topology. Although the kidneys appear identical in shape and gross size, the branching patterns and lineage are variable.

Figure S6. Branching dendrograms of left and right kidneys from a single embryo at E13.0. Left and right kidneys dissected from a single embryo begin to show variability in branch patterns and lineage. The left and right have 88 and 85 total tips respectively.

F

A B

C D

E

Figure S7. Remodeling during the late branching phase. (A) A central cavity begins to form and expand at E15.5 (glomeruli shown in purple). (B) Apoptosis (TUNEL) and proliferation (Ki-67) markers show concomitant cell growth and apoptosis in the origin (first branching bifurcation) of the ureteric tree at E15.5 (inset; magnification 400x). (C) Kidney bisected at E16.5 showing further cavity expansion. Arrow indicates the origin of a resorbing branch. (D) Lateral view of the kidney at E19.5 shows some of the separate collecting ducts exiting the papilla. These are created by erosion of the first 2 to 3 branching orders. (E) Kidney at E18.5 shows segment lengthening in the longitudinal axis.

A B C

Figure S8. Protein deficient and wild type kidneys at E15.5. (A) A confocal section of protein deficient kidney at E15.5 shows truncated branching orders as compared to wild type (shown in panel B). Figures to scale. (C) Tracing colored by stereotyped (blue) and variable (red) branching patterns shows a deficit in segments that specify kidney circumference (compare to wild type pattern arrangement at E13.5 in Fig. 3C).

Figure S9. Series of z-stacks. Stacks are obtained from 150 micron sections of an E17.5 kidney stained with cytokeratin-8 (green) and podocalyxin (red).

Figure S10. 3-dimensional reconstruction and tracing of wild type E15.5 kidney. (A) 150 micron thick serial sections were imaged and layered using ImageJ34 to reconstruct the whole embryonic kidney. Rotated views are represented. (B) Corresponding tracings of branching processes generated by the Neurolucida program.