From Stem (cell) to Branch: growing a renal tree

Significant strides have been made in the last few years towards the goal of engineering human renal tissue. Protocols to differentiate nephrons in vitro either on their own (and here) or in a mixed population with other renal cell types have being published. The ability to induce and maintain nephron progenitors in vitro has also recently been established, an important tool for both tissue engineering and developmental studies. While the formation of nephrons in vitro is a significant step forward and may be useful for some applications including basic research and toxicology, the architecture of the endogenous organ is key to function, and therefore any real strides towards physiologically relevant organoids will need to include appropriate gross morphology.

In the mixed population of renal progenitors and structures that results from the Takasato method, renal organoids can be formed that do appear to contain Gata3-containing epithelial tubules, however, these epithelial tubules do not branch or respond to the nephron progenitor signals in the same way as normal ureteric bud. In addition to the lack of higher-order structure comprising a central collecting duct, then, is the lack of branch-competent ureteric bud (UB) formation. There may be several reasons for this lack of branching. The stromal progenitors are important for maintenance of Ret expression in UB tips and so perhaps there is a problem with this population in the differentiated organoids. In addition, applying the same set of factors to all the tissue ignores the normal, localised antero-posterior differences that exist within the developing embryo.

In their recent paper, Taguchi and Nishinakamura have taken a “root and branch” approach, by paying close attention to the developmental cues and localised differences in normal development. They wanted to first develop a protocol for differentiating branch-competent ureteric bud from stem cells, and then try to add a single branching UB rudiment to metanephric mesenchyme (MM) in order to form a more anatomically correct organoid.

Using a reverse engineering method, the authors started by working backwards in development to identify signalling pathways essential for differentiation of the UB in vivo in mice. While it is true that this assumed the information gleaned from mice would be transferable to human, it was sensible in that mouse tissue is more readily available at precise developmental timings, has fewer ethical issues associated with it, and much is already known about the signalling pathways during mouse embryonic development.

The authors began by identifying the point at which branching capacity is achieved in mouse development. Using a transgenic Hoxb7-GFP mouse line and established reaggregation methods, they placed anterior or posterior GFP+ Wolffian Duct from E9.5, E10.5 and E11.5 with isolated metanephric mesenchyme, and also a GFP+ ureteric bud tip from E11.5 embryos with isolated MM. This showed them that branching capability was attained progressively as development went on, with robust branching from both WD and UB by E11.5, and less branching at earlier ages. The anterior or posterior WD branched equally in the presence of MM, driving home that the more anterior mesonephric mesenchyme is functionally different to the MM in its ability to induce branching.

They then used gene expression array analysis to identify useful markers for monitoring progression of differentiation. Looking at the gene expression analysis, they diligently worked their way back, identifying important factors for each stage of progression and then testing the ability of each factor to induce the next stage. For the progression from E9.5 WD to E11.5 UB, they isolated E9.5 GFP+ WD by cell sorting and then tried out combinations of factors identified in the gene analysis. Fgf9, retinoic acid (RA) and Wnt (in this case a Wnt agonist, CHIR) were identified as essential for progression of WD to UB, and addition of GDNF was needed to maintain Wnt11 expression and induce branching.

Having worked out the factors required for WD to branching UB, they then moved a step further back in developmental timings. The Hoxb7-GFP transgenic line shows clear expression of GFP at E8.75 in the anterior of the embryo suggesting that there are already WD-committed localised cells at this stage. They again identified factors from the array data and then tested their ability to maintain WD gene expression profile at E8.75, and to induce differentiation to UB. FGF9, CHIR and RA were again found to be required for maintenance of E8.75 WD gene profile, with lower levels of Wnt required and RA being essential. Interestingly, GDNF was dispensible at this stage to maintain Wnt11 expression, suggesting that different gene modules are active in the earlier embryo for WD maturation. Adding GDNF and higher Wnt concentration during a 3-day induction period from E8.75 led to branching UB-like cells.

Finally, the authors wanted to identify the differentiation steps between mouse ES cells and E8.75 committed WD cells. Using their microarray analysis, they identified two cell-surface proteins that could be reliably used to specifically sort E8.75 WD, giving them a tool to quantify successful differentiation from mESC to WD committed cells. Using a similar reverse engineering approach, and drawing on existing knowledge about early embryogenesis, they were able to define a protocol for differentiation of UB from mouse ESCs. The authors had previously discovered that shorter exposure to Wnt leads to anterior intermediate mesoderm (AIM) identity, whereas prolonged exposure to Wnt leads to posterior intermediate mesoderm (PIM) identity. This is because Wnt signalling maintains T+ nascent mesoderm, and the longer the cells remain in this state the more likely they are to take on PIM identity. In fact, they found that there is a very narrow window, at day 4.5 of differentiation, for Wnt exposure in this early stage of differentiation in order to induce AIM identity and subsequent WD/UB progression. Of course, Takasato et al. already showed that limiting Wnt exposure tips the balance to an anterior fate and produces more Gata3-positive epithelia in their organoids, in agreement with these data. In identifying the signals required for differentiation from ESC to WD, they also found that cell fate patterning of nephron progenitors and UB starts even before formation of immature mesoderm at the epiblast/primitive streak stages, in agreement with previous work (and here) showing early stage mesoderm patterning in the avian embryo.

The authors now had a method for differentiating WD from mouse ESCs. But, like most things, the proof is in the pudding. Could these induced WDs (iWDs) go on to differentiate into branching UBs? They established mESCs from Hoxb7-GFP mice, used their protocol to induce WD, and then sorted the differentiated WD cells. These were reaggregated and cultured with the factors identified by “reverse engineering”, namely, Fgf9, CHIR and RA, adding in GDNF for the final stage. These produced cells with UB identity that, crucially, were able to branch robustly. Using the Ganeva method, they isolated single UB nodules from these reaggregates, and added E11.5 metanephric mesenchyme dissected from mouse embryos. This produced branching organoids with a single collecting system, where the collecting duct tissue was differentiated from ESCs. Nephron progenitor gene expression was maintained and the distal tubules of differentiated nephrons were linked up to the collecting ducts. Strikingly, these induced UBs could also be induced to branch using a previously-established branching protocol without MM, showing that they truly are fully branch-competent. In a further step, the authors used their earlier published protocol for inducing nephron progenitors from mESCs, and added these to the induced UBs (instead of dissected MM), along with sorted E11.5 stromal progenitors. When both stromal progenitors (from the embryo) and induced nephron progenitors were included, the induced UBs could branch nicely, but removing either the stroma or NPs abolished or significantly reduced the branching, further underlining the importance of stromal progenitors in UB branching and renal differentiation.

Making branching UBs from ES cells was a major step forward, opening up the tantalising possibility of making renal organoids fully derived from human stem cells that had realistic macro-anatomy as well as functional nephrons. The authors applied their differentiation protocols to human iPS cells. With some amendments (necessary due to interspecies developmental and maturation differences), they were able to successfully induce branching tubules in 3D gel from human iPS cells, in an MM-free culture system, although it was much slower. This likely reflects the longer differentiation timeline of human renal maturation vs mouse.

In an attempt to make a fully engineered human organoid, they then used human iPSC-derived nephron progenitors combined with hiPSC-derived UB – this induced some weak evidence of initial branching, but did not progress past 7 days. This is likely consistent with the absence of human primary stromal progenitors although it is of course possible that the human protocol is not yet optimised in the same way as the mouse protocol. Interestingly, there appeared to be robust nephrogenesis in these cultures (seen in Supplementary Figure S7C) based on the brightfield images, suggesting that the human induced UB can at least induce MET in the nephron progenitors.

It is clear that Taguchi and Nishinakamura have contributed significantly to the field of renal tissue engineering with this work. It is now even more urgent to develop a protocol for inducing stromal progenitors from mouse and/or human stem cells, in order to test the hypothesis that the missing link for anatomically realistic organoids in human is stromal cells. There is no doubt that the field is progressing rapidly, and once the hurdle of human stem cell-derived organoids with adequate macro-anatomy has been broached, the field can move on to issues such as vascularisation and innervation, both of which may be crucial to the full maturation and growth of these engineered kidneys.


Melanie Lawrence






A question of identity

The mammalian Pax genes are homologues of the Drosophila gene paired, and are a family of transcription factors harbouring a highly conserved DNA-binding paired-box domain. The Pax genes have been found to be involved in organizing numerous aspects of development, including development of the central nervous system and diverse organs such as the pancreas and the kidneys. Pax2 belongs to this family and has for a long time been known to be important in mammalian renal development (as well as other tissues).  It is expressed in the pro- and mesonephros as well as the earliest condensing metanephric mesenchyme, and thereafter in the cap mesenchyme, renal vesicles, developing nephrons and ureteric bud. The essential requirement for Pax2 in kidney development is underlined by renal agenesis in Pax2 null mice, and by human mutations that cause pathology including renal malformations. Although known to be involved in normal renal development, the function of Pax2 in different renal cell types has not been established.

Two recent papers, one in mouse and one in human iPSC-derived organoids, now begin to shed light on the function of Pax2 in the metanephric mesenchyme.

A loss of function mutation for Pax2 leads to renal agenesis, making it impossible to dissect the contribution of Pax2 to the specification of different renal cell types. To solve this, Naiman et al created a floxed conditional compound heterozygote, with a non-functional truncated second allele. By crossing to a Six2-GFP-Cre line, Pax2 was conditionally removed only in the Six2-positive population that marks a subset of the metanephric mesenchyme – the cap mesenchyme. In this system, the kidneys were severely hypoplastic and only developed a small ureteric bud tree, underlining the importance of Pax2. Perhaps surprisingly, however, the cap mesenchyme did initially form, and CM cells persisted for 2 days. It could be that Pax2 is not directly required for specifying this subset of metanephric mesenchymal cells. However, deleting Pax2 in cells that are expressing Six2 means that the nephron progenitor cells have already begun to be specified as a subset of the MM when Pax2 is removed (because Six2 is a marker of the cap mesenchyme). This leads to a kind of chicken and egg question and therefore this study does not yet answer whether or not Pax2 is required for specification of the cap mesenchymal cells.

Nevertheless, the persistence of the cap cells for some time before apparently disappearing begged the question what was happening to the cap mesenchyme cells when Pax2 was absent. The authors surmised that there were two possibilities: either they were dying, or they were changing their identity. Transdifferentiation is a process whereby cells that are already committed to a cell fate can be persuaded to take on the identity of a different cell type, without first being taken back to an earlier state. To test whether the Six2-positive cap mesenchyme cells were transdifferentiating or simply dying, Naiman et al used a LacZ-Cre reporter to lineage trace specifically the Pax2-deficient, Six2-expressing cells. What they found was that the cap mesenchymal cells in the Pax2 mutants were not simply dying: they were changing their identity towards renal interstitial cell fate. This is striking because previous work has shown that the lineage of Six2+ nephron progenitor population and the Foxd1+ stromal progenitor population are committed very soon after the onset of metanephric development. Furthermore, the authors show that in wild type metanephric mesenchyme, Pax2 and Foxd1 are usually not co-expressed. Yet, analysis of the expression pattern of the transdifferentiating cells showed that cells starting out as Six2+Foxd1-, are briefly Six2+Foxd1+ before settling into their new identities as stromal progenitors and gaining a Six2-Foxd1+ signature.

So what does this tell us about the role of Pax2 in renal development, at least in mice? Clearly, Pax2 is required for maintenance of the nephron progenitor population. Six2 is also required for maintenance of this population by repressing premature MET. Unlike Six2, however, Pax2 appears to maintain this population by repressing stromal identity. With Waddington’s epigenetic landscape for cell fate determination in mind, this means that removal of Pax2, even after the marble has rolled down the cap mesenchyme “trough” leads to cells reverting to a default identity of renal interstitium. It seems as though cap mesenchyme cells are desperate to leave their identity behind, but this is thwarted by Pax2 and Six2 as they block escape to a new identity, one stromal and one epithelial.

Whatever light has been shed by the Naiman paper, however, some questions remain about the role of Pax2 in the MM. Hot on the heels of the work by Naiman et al is another paper, demonstrating that PAX2 appears to be dispensable for nephron formation, and therefore nephron progenitor cell maintenance, in human iPS cells. Kaku et al have taken human iPS cells, knocked out PAX2 (replacing it with GFP), and used their protocol to differentiate to metanephric mesenchyme. This mesenchyme exhibited robust tubulogenesis when co-cultured with murine spinal cord, despite the absence of PAX2. Using cell type specific markers, the authors then show that this differentiated cell population is indeed metanephric mesenchyme and identify two distinct sub-populations within it: ITGa8+/PDGFRA-, and ITGa8-/PDGFRA-. PDGFRα is a cell surface receptor that is associated with renal interstitial cells and which has been identified previously by the authors as a negative selection marker for nephron progenitors. The ITGa8+/PDGFRA- population sorted from both wild type and PAX2-deleted hiPSCs-derived MM was shown to induce tubulogensis whilst the ITGa8- population did not. Proximo-distal polarity was not impaired, and glomerular structures were observed. The only phenotype they observed in their PAX2 knockout organoids was a morphological change in the parietal epithelial cells of the glomerulus.

On the one hand, Naiman et al show quite convincingly that Pax2 is essential for maintaining the cap mesenchyme population and repressing a shift towards a different identity. On the other hand, Kaku et al suggest that it might be dispensable for nephron progenitor maintenance in human cells in vitro. One possible explanation to reconcile these opposing findings is that there is a species difference. It would be interesting to see if Pax2 remains dispensable for nephron formation from mouse iPS or ES cell-derived MM. Another possibility raised by Kaku et al is that PAX8, which is thought to have some redundancy with PAX2 based on the more severe phenotype of the double Pax2/Pax8 knockout mice, might be compensating for the lack of PAX2. They show that PAX8 expression is expanded in the epithelia of their hiPSC-derived PAX2 null organoids. If this is the case then there must indeed be some species differences since PAX8 remained intact in the Naiman conditional knockout. And in humans with mutations in PAX2, the serious kidney hypoplasia seen does not support a strong compensatory role for PAX8.

Clearly, in vitro generation from iPS cells and in vivo development are very different developmental processes. Another possible explanation for the different outcomes upon removal of PAX2 is that an organoid derived from a single renal cell type is not equal to an organ developed in vivo. In the human iPSC PAX2 knockout, the protocol differentiates only MM without UB and spinal cord is used to induce tubulogenesis. Could there be a signal from the UB telling the MM to become stromal progenitors, with PAX2 repressing this signal? If so, then in the absence of UB, PAX2 would indeed be dispensable for nephron formation since the message to follow the stromal lineage would be non-existent. Or perhaps the protocol for nephron formation in vitro does not faithfully reproduce the situation in vivo; there may be compensatory mechanisms at play due to this difference and therefore extrapolation to human in vivo development should be approached cautiously.

What is clear is that the role of Pax2 in mammalian renal development in vivo is critical. Its role in maintaining the identity of the nephron progenitor population within the developing renal environment in mice is a fascinating developmental mechanism, without which the cells undergo a sort of identity crisis leading to failure of the kidney to develop. Many questions persist including possible species difference, the mechanism of repression of stromal identity by Pax2 in mice, and the role of Pax2 in the collecting duct. It will be interesting to see how the story of Pax2 unfolds and the differences between species and in vitro/in vivo development.

Melanie Lawrence

More information on the 20th UK/EU nephrogenesis workshop

Registration for the nephrogenesis workshop on 22 June in The Roslin Institute, University of Edinburgh is still open. A preliminary program is now available. Registration and coffee will start at 9 pm, with the first talk at 9:30. The day is expected to end at 16:30 pm.

Submission of abstracts, from which the majority of talks will be selected, will be closed on 9 June 2017.  Two keynote talks will be given by Andreas Schedl and Barry Denholm.

For all other information about the workshop, please see this previous blog entry.

The easiest kidney research fundraising you’ve ever done…

Not directly linked to kidney development, but still pretty close… On 22 June the International Kidney Cancer Coalition is organising the World Kidney Cancer Q&A Day. As part of this, the Q&A day quiz is now online. Not only very informative but for every completed quiz $5 is donated to kidney cancer research. So go there now, spend 2 minutes on the quiz, and help kidney cancer fundraising.


Peter Hohenstein

Registration UK/EU nephrogenesis workshop now open

On 22 June 2017 the 20th UK/EU nephrogenesis workshop will be held at The Roslin Institute, University of Edinburgh. Registration is now open.

As always, we have tried to keep registration fees as low as possible. PhD students and technicians will get free registration, post-docs and PIs will have a registration fee of £30. Registrations include all tea, coffee breaks and lunch breaks. These low registrations costs have been made possible by generous support from Kidney Research UK,  Transnetyx and IDT Integrated DNA Technologies. Note that if you want to use the free registration option you will need to supply a letter from your supervisor or line manager to confirm your status. This letter needs to be uploaded at the time of registration.

Keynote speakers will be Andreas Schedl from the University of Nice in France and Barry Denholm, Centre for Integrative Physiology, University of Edinburgh. Other talks will be selected from submitted abstracts. Abstracts can also be uploaded at the moment of registration.

The Roslin Institute is on Easter Bush Campus, south of Edinburgh. It is served by bus routes 15, 37, X47 and 67 from Lothian Buses. Taxi costs from the city centre to the institute are approximately £20 and take 20-30 minutes. Finding accommodation in Edinburgh is not a problem, and hotels in all price categories are available.


Peter Hohenstein

Kidney reaggregation and improved kidney imaging in the special issue on organoids in Development

The journal Development published a special issue this week on organoids. Besides a wide variety of papers on organoid systems for different tissues, two papers are focusing on the developing kidney.

Since the demonstration by Unbekandt and Davies in 2010 that the classic Auerbach and Grobstein aggregation system could be adapted to only require cell types from the embryonic kidney itself, this aggregation method or derivatives from it have formed the basis of many seminal kidney organoid papers. However, the mechanism underlying this remarkable self-organization have remained unclear. Now Lefevre et al from the Little lab, using time-lapse and confocal imaging as well as mathematical modelling, study the dynamics of this reaggregation process.They use time-lapse imaging of reaggregated kidneys to study the kinetics of the spontaneous nephron initiation. Clusters of ureteric epithelium cells formed after 8 hours, which after 48 hours was followed by the formation of cap mesenchyme clusters around them. Mathematical modelling based on these time-lapse data suggested that at least in the first 24 hours differential adhesion between cells could account for the formation of the UE clusters. Finally, the authors hypothesise that homophilic cadherin interactions could explain this clustering, and using blocking antibodies they show that P-cadherin, but not E-cadherin, is likely involved in this clustering of UE cells in the first 24 hours. In all, this is a very interesting study on the mechanism behind these aggregation experiments.

A second kidney paper in this special issue is from the lab of Seppo Vainio. Saarela et al present a method to improve the confocal time-lapse imaging of kidney rudiments which they refer to as ‘fixed z-direction’ or ‘FiZD’ imaging.  They limit the growth of the embryonic kidney in the z-direction by growing the kidney under a porous membrane but on a glass slide, with the two separated by glass beads that determine the space the growing kidney can get:


The system allows for very good development, including the formation of Loops of Henle, as had previously been observed in the Sebinger low-volume culture method. It would be interesting to see how these two different culture conditions allow this better development that the conventional method does not allow. The (confocal) imaging using the FiZD cultures is indeed superb and allows for computer-assisted cell segmentation and morphometric analysis. The FiZD system will no doubt find useful uses in the time-lapse and confocal imaging of developing kidneys.

Peter Hohenstein

Serum-free UB cell maintenance, proliferation and branching in vitro

From patterning of the vasculature and neuronal networks to the development of diverse organs such as the lungs, mammary glands, and kidneys, branching morphogenesis is a common phenomenon observed during animal development. Perturbations in normal branching caused by genetic or exogenous factors can lead to disease, or abnormalities in surrounding tissues. In the kidney, correct patterning of the collecting duct is central to renal function, allowing appropriate filtration and drainage into the bladder. In addition, the ureteric bud and tip cells that will become the mature collecting duct are essential for the development of the kidney’s main processing unit, the nephron. Thus, perturbations in the process of branching morphogenesis in the developing kidney can lead to numerous pathologies and so studying the mechanisms behind this process are of great value. A platform to study branching of ureteric bud cells in vitro is a useful way of dissecting some of these mechanisms, and may also be a key part of advancing renal regenerative medicine.

Previous work has shown that it is possible to coax ureteric bud tips in vitro to propagate and branch in the absence of metanephric mesenchyme (MM) or MM-conditioned media. This was in itself an important step forward, since proliferation and branching of the ureteric bud normally requires signaling from the MM. However, serum was still required in addition to growth factors and signaling factors that would normally be secreted by the mesenchymal population. This is a problem for studying the process of branching in vitro, since serum contains numerous known and unknown factors that can act as confounding variables.

In their recent Stem Cell Reports publication, Yuri et al. have used a systematic approach to reveal a defined set of factors for the maintenance, proliferation and branching of dissected UB tips as well as dispersed ureteric bud (UB) cells. Crucially, their culture method does not require serum, making it ideal for studying pathways involved in branching. The authors began by culturing UBs from e11.5 mouse kidneys in factors known to be important for UB cell maintenance – specifically, GDNF and FGF1. These two factors together, or FGF1 alone, allowed the UB cells to survive and proliferate but did not allow the survival of tip cells or allow branching – this was a problem since the tip cells represent the stem cells of the collecting duct. It has previously been shown buy Bridewater et al and  Marose et al that Wnt / β-catenin signaling is necessary for maintaining UB tip cell identity, so the authors hypothesized that addition of the GSK-B inhibitor CHIR99021 might encourage tip cell proliferation and maintenance. In fact, addition of all three factors (GDNF, FGF1 and CHIR) induced extensive branching and led to the enrichment of tip cells. Interestingly, direct inclusion of Wnts as a substitute for the WNT activator CHIR99021 did not recapitulate the effects, suggesting a more global pathway. To investigate this, the authors turned to a mouse knockout of an R-spondin protein receptor with abnormal UB branching. Rspondin (RSPO) proteins are known agonists of WNT-B-catenin signaling so they hypothesized that RSPO proteins may be able to activate WNT in their culture system similarly to CHIR99021.  In fact they found that substituting either RSPO1 or RSPO3 for CHIR99021 in the culture system preserved the maintenance of tip cell identity and ability to branch extensively.

Continuing their systematic approach to identifying the defined factors required for in vitro propagation of branching UBs, the authors then investigated the role or retinoic acid (RA). RA is known to be crucial to kidney development as shown by the Mendelsohn lab (Batourina et al and Rosselot et al) and indeed the authors found that culturing UBs in RA alone allowed the cells to survive – but not branch. Adding GDNF to the RA encouraged proliferation and maintenance of tip markers; this highlights the essential role of RA in maintaining tip cell identity and the interplay of RA and GDNF/Ret signaling in proliferation and branching. This is of potentially crucial importance for renal regenerative medicine; RA is mainly secreted by the stromal cells and so inclusion of this population of cells in any approaches to renal tissue engineering may be required.

In addition to defining serum-free factors that allow the proliferation and branching of UBs in vitro, the authors proceeded to identify a defined set of factors that would allow single (dispersed) UB cells to propagate. This would be a useful addition to the toolkit of renal biologists, allowing dispersed UB cells to be manipulated (e.g. transfected with plasmids) and then their proliferation and branching studied either alone or in the context of other cell types. Stunningly, the authors were able to combine the factors above – FGF1, GDNF, RA – to not only allow proliferation but also branching of single UB cells, and inclusion of CHIR99021 into the cocktail induced further branching in an additive manner. Just as impressively, the authors mixed UB branched structures formed from single, dispersed UB cells with metanephric mesenchyme, and showed that they were still able to induce the mesenchyme to form nephrons, demonstrating the potential power of this approach.

This work is an impressive step forward in the propagation of UB cells, opening up avenues for studies of branching morphogenesis and kidney development in a serum-free manner. It will be interesting to see if this approach will hold true in human cells differentiated from pluripotent stem cells – if so it could be a valuable addition to the field of renal regenerative medicine.

Melanie Lawrence