To more fully define the molecular character of in vivo
podocytes we conducted a series of gene expression profiling experiments. The purpose was to globally define the changing gene expression states of this remarkable cell from stage E13.5 of development to adult. To this end we made use of the MafB-GFP
BAC transgenic mouse from the GENSAT project 
. We found that these mice showed highly restricted GFP expression in podocytes in both the developing and adult kidney.
The podocyte specificity of MafB-GFP label was clearly demonstrated by fluorescent microscopy. Even as early as E13.5 in the S-shaped bodies the prospective podocytes were uniquely labeled by GFP (, left panels). At this stage of development the immature podocytes form a single layer of cells adjacent to the glomerular cleft. As development progresses a capillary loop forms within the cleft and the early glomerulus is encircled by podocytes (, right panels). The MafB-GFP transgenic kidneys did not show GFP fluorescence in cell types other than podocytes. In addition, the MafB-GFP fluorescence pattern was observed to exactly match expression patterns of known podocyte marker genes, as discussed in more detail later.
MafB-GFP mice show restricted expression of GFP in podocytes.
We used a strategy of rapid enzymatic cell dissociation followed by fluorescent activated cell sorting (FACS) to isolate the MafB-GFP positive cells from embryonic kidneys, at stages E13.5 and E15.5. For analysis of adult podocytes we first sieve purified glomeruli, which provided a significant enrichment for podocytes. The glomeruli were then further subjected to enzymatic dissociation and FACS, in order to obtain a pure population of podocytes. The MafB-GFP GFP fluorescent label was quite strong, allowing a stringent gating during FACS, which resulted in very low levels of contamination.
The resulting podocyte microarray data was further screened to monitor for possible contamination. One method used for estimating podocyte cell purity was to determine the transcript level for a marker of a flanking cell type. In particular, we examined expression levels for Tie2
), which is specifically expressed in endothelial cells, which abut the podocyte and are therefore the most likely source of contamination. Samples with significantly above background levels of Tie2
expression were removed from further analysis (see Material and Methods). In addition we have previously defined cell type and compartment level specific markers for most of the elements of the developing kidney (GUDMAP.ORG, 
). The podocyte microarray data was carefully examined to insure against contamination by these multiple components.
Defining the adult podocyte gene expression state
The podocyte has a unique multifunctional character that we wished to better define by examining its complete gene expression state. We used two strategies to screen the expression data. First, to discern the adult podocyte specific components we subtracted out the gene expression profile of the total adult kidney cortex. That is, we sought genes with transcripts specifically enriched in the podocytes compared to total cortex. To this end we used GeneSpring software, with summarization algorithm RMA16, filtered on raw expression minimum of 150 in at least three samples, performed ANOVA P<0.05, and required minimum three fold enrichment in adult podocytes compared to total cortex, giving a total of 436 probesets.
In a second strategy we searched for genes that were active in the adult podocyte compared to the developmentally early E13.5 podocyte. The underlying hypothesis is that the E13.5 podocyte is largely undifferentiated, and as the differentiation program ensues the set of genes that distinguish podocytes from other cells will become active. A similar analysis of the data was performed, requiring a minimum three fold enrichment, this time between adult podocytes and E13.5 podocytes. This approach yielded 739 probesets, with the two strategies having an overlapping set of 281 probesets.
Each strategy alone is imperfect. Some genes that are of key importance in the podocyte might also show significant expression in other regions of the kidney cortex. These genes would be unfortunately subtracted in the screen looking for enrichment compared to kidney cortex. And, some genes that are specific markers of the podocyte, such as MafB, are already active at E13.5. These genes would be eliminated by a screen that looks only for genes with transcripts enriched in the adult podocyte compared to E13.5. Nevertheless it is reassuring that the two strategies yield similar sets of genes, with about 2/3 of the 436 probesets found by comparison to kidney cortex also identified by the screen comparing adult to embryonic.
We first cast a wide net, looking to create a comprehensive catalog of genes whose expression defines the adult podocyte. We therefore combined the lists made with the two screening strategies creating a set of 894 probesets. A heatmap provides a visual representation of the relative abundances of transcripts of adult podocytes compared to E13.5, and E15.5 podocytes, as well as total kidney cortex (). Many of the genes show a graded expression level, weakest at E13.5, stronger at E15.5, and then strongest in the adult podocyte. also illustrates how some of the adult podocyte probesets are enriched compared to E13.5 but not total cortex, while others are enriched versus total cortex but not E13.5. For a complete inventory of the 894 genes, along with fold enrichments, see Table S1
. Of interest, and validating the screen, a large number of genes previously associated with podocytes showed the greatest enrichments. These results confirm the purity of the podocytes used for array analysis. For example, comparison of adult podocytes to total kidney cortex showed very strong fold changes (FC) for Nphs1
, 23 FC), Nphs2
, 20 FC), Wt1 (Wilm's Tumor1
, 19 FC), Foxc2 (forkhead box c2
, 19 FC), nes
, 12 FC) and pdpn
, 17.5 FC) and synpo
, 7.5 FC) (Table S1
Heatmap of 894 probesets with elevated expression in adult podocytes.
Interestingly, we also found unexpected genes expressed in podocytes. For example Foxd1, which has generally been considered a marker of the kidney interstitium, or stromal lineage, showed extremely robust expression in the podocyte.
To better define the molecular processes and biological functions carried out by the podocyte we analyzed the 894 gene list with the ToppGene web tool 
. This software application searches for gene enrichments associated with specific molecular functions and biological processes. An interesting view of the podocyte emerged, with an unusual mix of functions. Given the extraordinary structure of the podocyte it is not surprising that a number of enriched genes were associated with the cytoskeleton. There were 65 cytoskeletal binding proteins identified, and 39 genes involved in actin skeleton organization. Several other interesting molecular processes and biological functions emerged. Twelve genes encoded proteins involved in integrin binding, and another 44 were involved in calcium ion binding.
The top biological processes to emerge from the ToppGene analysis included vesicle mediated transport, with 72 genes involved, actin cytoskeleton organization (39 genes), regulation of signaling (99 genes), neurogenesis (74 genes), neuron projection development (52 genes), axon guidance (33 genes), biological adhesion (59 genes), response to oxygen levels (19 genes), neuromuscular junction (7 genes), chemotaxis (38 genes), phagocytosis (11 genes), striated muscle cell differentiation (16 genes), muscle contraction (20 genes). For complete gene lists see Table S2
. The biological process analysis again shows a strong neuronal character, but also some molecular features associated with muscle and phagocytes.
We examined the possible muscle character of the podocytes in more depth, in light of their possible contractile role in counteracting the perfusion pressure of the capillaries. It is interesting to note that podocytes did express several myosins, including myo6, myo1e, myo1d, myo10 and myl6. Nevertheless, these are generally unconventional myosins that are more associated with vesicle transport and other movements along actin filaments rather than muscle contraction. Podocytes also showed strong expression of Tpm1, tropomyosin, which binds actin filaments in both muscle and non-muscle cells.
It is interesting to compare the array results presented here with previous studies of the muscle nature of podocytes. Potential contractility of the podocyte has long been noted 
, and a more recent study examined in some detail the muscle characteristics of podocytes grown in culture 
. Three muscle markers, smoothelin, calponin and myocardin were detected by three methods, microarray, western blot and immunofluorescence. Several other genes associated with smooth muscle differentiation were also seen elevated in expression in podocytes by as measured by microarray. Surprisingly, our microarray analysis of in vivo podocytes provided somewhat disparate results. We observed Cnn1
(calponin) expression at essentially background levels in adult podocytes, although slightly above background at E13.5 and E15.5. Myocd
(myocardin) was expressed only at background levels in podocytes for all times examined. Similarly Smtn
, (Smoothelin), was just slightly above background (~100 raw signal) at all times. In addition for several other muscle function genes previously observed expressed in podocytes 
, we saw little, if any, expression. Neb
, and Ryr2
were at background levels, while Fhl2
were just slightly above background (~100 raw signal). Id2
showed moderate expression (~400) in developing podocytes, but was off in adult, and Aebp1
gave low level (~250) adult expression. We did, however, observe very robust expression for both Mbnl 1
(muslcebound like) in developing and adult podocytes. In summary we detected a distinct, yet rather weak, muscle gene expression signature. There are several possible reasons for the discrepancies. In particular, different microarrays were used, and the previous study primarily examined a podocyte cell line grown in tissue culture, compared to the in vivo
podocytes isolated by FACS in this study.
The gene expression profile analysis also provided a global view of the signaling pathways present in podocytes. A gene ontologies analysis of the 894 genes with podocyte elevated expression, using GeneSpring software, identified 44 genes in the receptor category, and 28 genes encoding proteins that bind receptors. These genes can be further screened by requiring a raw expression signal of at least 500, very roughly the transcription level required to be detected by in situ
hybridization. This reduces the numbers to 26 receptor and 17 receptor binding genes. See Tables S3
for complete gene lists. Table S5
shows another list of 116 genes expressed in podocytes involved in signal transduction.
We found that podocytes express three semaphorins, Sema3g
. Semaphorins are a large family of secreted and membrane bound proteins that act as axonal growth cone guidance molecules, primarily as short-range inhibitors, deflecting axons from inappropriate regions. It is easy to imagine how they could contribute to the formation of the slit diaphragm. It has previously been shown that immortalized podocytes in culture express semaphorins 
, although many were reported expressed in near equal abundance, while we observed more restricted and distinct expression levels. It is also interesting to note that the Sema3a
gene knockout mouse shows very wide foot processes and foot process effacement 
. Our array data showed that Sema3a
is expressed during kidney development, but is turned off in the adult podocyte. In contrast, Sema3e
were off during development and expressed in the adult. Sema5a
was expressed at low levels during development and strongly in the adult.
Robo2, another molecule with a key role in axon guidance, showed strong expression in adult podocytes. Of interest, one of the ligands, Slit2, showed modest expression in the podocyte at E15.5, but was off in the adult.
gene, showed very robust expression, with a raw signal of around 4500 in the adult podocyte. This gene encodes a receptor for atrial natriuretic peptide, produced by the atria of the heart, with a number of vascular, renal and endocrine effects that are important in the maintenance of blood pressure and fluid volume. Npr3
mutant mice show reduced ability to concentrate urine, and exhibit mild diuresis, as well as surprising bone defects 
. Podocytes also expressed Npr1
) encodes a protein that functions as part of receptors for IL6, LIF, OSM, CNTF, IL11 CTF1 and BSF3. We also observed strong expression of interferon gamma receptor 1 (Ifngr1
) in adult podocytes. It has been previously observed that podocytes express this receptor and respond to gamma interferon with expression of HLA-DR, -DP and -DQ 
We also identified several integrin genes expressed by podocytes, including Itga2, Itga3, Itgav, Itgb1 and Itgb5. ITGB1, heterodimerizes with ITGA3 forming a receptor for fibronectin, laminin, collagen, epiligrin, thrombospondin and CSPG4. Integrin alpha-V/beta-5 is a receptor for fibronectin, recognizing the sequence R-G-D.
receptor gene showed modest expression in the adult podocyte. It was previously reported that cultured mouse podocytes express Bmpr1a
, as well as Bmp2
, as determined using RT-PCR methods that could detect very low levels of expression 
. Our array data shows only background levels of transcripts for Bmpr1b
, and Bmp2
, very low but microarray detectable expression levels for Acvr1
, modest expression for Acvr2
and a high level of expression for Bmpr2
, which is also expressed in embryonic podocytes, as well as the total kidney cortex.
Podocytes also expressed the Mertk
gene, encoding a receptor kinase, which is of interest in terms of the possible clearing role of podocytes, since this gene has been shown to be important in the phagocytic function of some cell types 
. Podocytes also expressed Colec12a
scavenger receptor that has been shown to be important in the mediation of zymosan phagocytosis by vascular endothelial cells 
. It has also been implicated in the clearance of amyloid beta in Alzheimers disease 
Other signaling molecules of particular interest expressed by podocytes included Spred2
, which is a sprouty related inhibitor of receptor tyrosine kinases, and Gmfb
, glial maturation factor, beta, which causes differentiation of brain cells, and inhibits proliferation 
Several genes related to calcium-regulated events showed elevated expression in podocytes, including Anxa5 and Anxa1, calcium dependent phospholipid binding proteins implicated in exocytic and endocytic pathways. S100a10, is also calcium binding and also implicated in endocytosis and exocytosis. Fyn encodes a receptor tyrosine kinase that helps regulate intracellular calcium levels, and plays an important role in the brain in regulating axon growth, axon guidance and neurite extension.
Of particular note, podocytes also expressed Vegf, Ctgf, Egf and Npnt (nephronectin). Angptl2 is a member of the VEGF family, and also showed very strong expression in the adult podocyte. Podocytes also expressed Efnb1 (ephrin B1), a membrane bound ligand for Ephrin related receptor tyrosine kinases, which has been implicated in the orientation of axons.
In summary, the receptor and receptor binding gene expression signature of podocytes provide an interesting picture, showing roles for calcium signaling, phagocytic/clearing function, and overall a striking neuron like character for these cells.
Transcription factor signature of the adult podocyte
Transcription factors generally play key roles in defining the identities of cells. Directed transcription factor expression can be used in some cases to drive the differentiation of stem cells towards defined cell types 
, or to trans-differentiate one cell type into another 
. Remarkably, this strategy can also be used to reprogram differentiated cells into the functional equivalent of embryonic stem cells 
. We were, therefore, particularly interested in defining the transcription factor gene expression code of the podocyte.
GeneSpring mediated gene ontologies analysis of the 894 genes with podocyte elevated expression identified 54 genes in the transcription category (Table S6
). We found 34 of these showed over 500 raw expression signal. Included were 14 genes with very high expression levels of over 1000. It is striking that three of these genes, Foxc2
, have been previously shown to play a key role in kidney, and podocyte development. Indeed, Foxc2
, coupled with early Notch signaling, have been shown to be capable of driving a podocyte development program 
, another highly expressed gene in the podocyte is a member of the forkhead family of transcription factors, like Foxc2
. Raw signal expression levels were observed to be over one thousand during development, at E13.5 and E15.5, and approximately 2,500 in the adult podocyte. Foxd1
is generally considered a marker of stromal, or interstitial cells. The dramatic Foxd1
mutant kidney phenotype, including reduced branching morphogenesis and decreased nephron numbers, has been attributed to a stromal cell function deficit in signaling to forming nephrons 
or in formation of the renal capsule 
. The robust expression of Foxd1
observed in the podocyte, however, suggests a possible function mediated through this cell type. This result reasons that previous studies investigating Foxd1
function in kidney development might be subject to re-interpretation. A podoctye specific knockout of Foxd1
would be useful to determine its function in this cell type.
, like Foxd1
, is highly expressed in the adult podocyte, with transcripts showing an approximate ten fold enrichment compared to total kidney cortex. It is also more widely expressed in the earlier developing kidney, but again including definite podocyte expression. The Dach1
mutant mice exhibit early postnatal death, although no developmental defects were detected in any organ system examined, including kidneys 
. The Eya
encoded proteins often interact in a conserved network in development 
. We observed only background levels of Eya1
transcripts in the podocyte, and only slightly above background expression for Eya3
. Similarly, Six1
were present at very low levels in the podocyte, while Six2
transcripts were present at modest levels in the early developing podocyte, and essentially absent in the adult. Dach1
in the podocyte therefore appears to act independent of Six
. Of particular interest, given the importance of preventing adult podocyte proliferation 
, DACH1 can inhibit Cyclin D1 and thereby inhibit cell proliferation 
. DACH1 can also inhibit JUN-mediated contact-independent growth 
, which could be of importance since we observe that Jun
is also expressed in prodocytes. JUN is a positive regulator of cell proliferation. Of interest, loss of E-cadherin mediated cell-cell contact can up-regulate Jun 
Podocyte expressed transcription factors also included Hoxc4
, and Mafb
. Of interest, Hoxc6
are both expressed strongly in stromal cells during development, like Foxd1
, as well as in forming podocytes (Genepaint). The targets of Hox genes can vary from cell type to cell type, but it is interesting to note that HOXC6 targets in prostate cells include elements of FGF, BMP, NOTCH and WNT signaling pathways 
. MAFB is a member of the Maf family of transcription factors, and has been previously shown to play a critical role in podocyte development, with mutant podocytes showing fused foot processes that did not interdigitate 
encodes a zinc finger E-box binding homeobox transcription factor and has previously been associated with epithelial mesenchymal transition, in particular as a repressor of E-cadherin 
Other transcription related genes of interest strongly expressed in the podocyte included Supt4h1, a regulator of transcription pausing, Sap18, part of a histone deacetylases complex, Smarca2, a member of the SWI/SNF family of proteins that regulate gene expression through chromatin remodeling, Setd7, a histone methyltransferase, Kdm4c and Jhdm1d, both histone demethylases, Txnip, which can induce cell cycle arrest, and Sync, which is generally expressed in skeletal and cardiac muscle.
It is important to note the striking success of this screen in identifying a number of transcription factors previously shown to play key functional roles in the podocyte. This provides an important historical validation of the analysis. It suggests that some of the newly identified podocyte expressed genes, reported herein, may also have yet to be discovered essential functions in podocyte biology. It is also important to note that the analysis of podocyte transcription factor expression presented in this report is far from exhaustive. Many other transcription factors, as listed in supplementary data, that are not necessarily strongly enriched in the podocyte, and/or are not expressed at high levels, are certain to play important roles in the podocyte.
A GeneSpring pathways analysis, looking for direct interactions, was carried out on the 894 podocyte probeset list (). The resulting diagram illustrates many of the known interactions among these gene products, and shows particularly strong interaction centers for EGF, JUN, and RHOA, as well as many interactions for ITGB1, H2-K1, FYN, ACTR2, VEGFA, APP, ANGPTL2 and HSPA1B. This begins to reveal the complex interplay of the podocyte specific gene products.
GeneSpring pathways analysis of the 894 genes with elevated podocyte expression.
Requiring greater podocyte stringency
The above analysis examined a broad set of podocyte enriched genes. There is, however, also utility in the identification of a more stringent set of podocyte expressed genes. This serves to further define the unique nature of the podocyte, provides a potentially valuable new set of molecular markers specific to the podocyte, and can serve to aid in the construction of novel transgenic tools for the study of the podocyte.
To this end we repeated the analysis of the array data, only this time requiring five fold enrichment in podocytes, instead of the previous three. This resulted in 204 probesets for the podocytes versus total cortex comparison, and 304 probesets for the adult podocytes versus E13.5 podocyte comparison, with an overlap of 143 probesets (). The combined entity list of 365 probesets (Table S7
) was then examined with ToppGene, finding similar sets of molecular functions and biological processes, compared to those identified with the 859 probeset list. These can be visually represented with cytoscape, with a subset of the 365 podocyte specific genes shown as hexagons in the center (). Associated molecular functions and biological processes are shown as green rectangles, linked by lines to related genes. Key functions identified include neuron development (), calcium binding (), and cytoskeletal protein binding (). This functional analysis of genes lists is based on statistically significant gene enrichments. For example Table S2
shows that there are 72 podocyte genes associated with vesicle-mediated transport, and this is highly statistically significant, with a P value of zero, arguing very strongly that podocytes carry out vesicle mediated transport.
Heatmap of 365 probesets with five fold elevated expression in adult podocytes.
Functional groupings of genes with elevated adult podocyte expression.
Podocyte specific markers
We further extended the adult podocyte specific analysis, this time screening for genes with higher expression levels as well as five fold podocyte enrichment compared to total kidney cortex. In this case we did not include genes with only strong enrichment in adult compared to embryonic podocytes, as several of these genes were not podocyte specific. The heatmap of shows the resulting 171 probeset list, with expression levels compared across 13 specific kidney compartments. A version of this heatmap with gene names included is provided in Figure S1
. After removing probeset duplications and ESTs the list reduces to 144 podocyte highly enriched genes (Table S8
Comparison of podocyte expressed genes across many kidney cell types.
These genes were then examined for podocyte expression using two public gene expression databases GenePaint (www.genepaint.org
) and Eurexpress (www.eurexpress.org
). These databases examine gene expression patterns by in situ
hybridizations, using E14.5 mouse embryos. During these early stages of development the podocytes encase the developing glomerulus, resulting in distinctive crescent or circular expression patterns. The Nphs
1 and Nphs2
genes, which are known to be podocyte specific in expression, served as positive controls. The result was striking, with 30 genes showing hybridization patterns suggestive of podocyte expression (,). This is a surprisingly large fraction of the 144 total, since the array data showed that many of the adult podocyte specific genes are not yet strongly expressed at early developmental stages ().
Expression of adult podocyte genes in the developing kidney.
In situ hubridizations suggesting podocyte expression.
To provide additional validation of the microarray data more in situ hybridizations were performed, at E15.5, and a third public database, GUDMAP.ORG, was used. The resulting gene expression patterns for twelve podocyte expressed genes are shown in . For some of these genes, such as Wt1, Synpo and Clic5, podoctye expression has been previously shown, It is important to note that they are expressed in the E15.5 kidney in cells that circle the capillary loop, exactly as observed for the MafB-GFP fluorescent cells. In addition this figure documents the podocyte expression of many additional genes. In some cases the expression is quite podocyte specific, as for Rhpn1, Gpsm3, Tdrd5 and Sgip1, while in other cases there is also expression observed in other cell types. Of particular importance, the in situ hybridization results confirm the podocyte expression of Foxd1, which has previously been assumed to be a stromal cell specific marker.
Expression of podocyte genes in E15.5 embryos, as well as P0 and adult.
In summary we have carried out an analysis of the complete gene expression states of the podocyte during development and in the adult. The gene expression profiles provide a comprehensive definition of all of the transcription factors, growth factors, receptors and cytoskeletal proteins expressed by this remarkable cell type. The results confirm and extend previous studies. A distinctly neuronal character was defined for these cells, consistent with their axonal like projections, but seemingly contrary to their mesodermal origin. The podocytes also expressed a number of genes commonly associated with muscle, although when examined in toto their muscle gene expression signature was rather weak. Podocytes were also found to display phagocytic like properties, perhaps related to their function in the clearing of the GBM. In addition, an exhaustive list of podocyte expressed genes involved in calcium signaling were identified. This study, to the best of our knowledge, provides the most comprehensive analysis to date of the molecular character of this most spectacular cell type.