Angiogenin expression in human kidneys and Wilms' tumours: relationship with hypoxia and angiogenic factors

Abstract

Angiogenin (ANG) is a potent angiogenic factor that is up-regulated by hypoxia. ANG expression is well documented in normal tissues and in common tumours, but its expression has not been reported in the normal human kidney or in Wilms' tumours (WT). We examined ANG expression in WTs, human fetal kidney (FK) and childhood kidney (NK) samples and studied its relationship with microvascular density (MVD) and with three other hypoxia-induced angiogenic factors: lactate dehydrogenase A (LDHA), vascular endothelial growth factor (VEGFA) and BHLHE40 (basic helix-loop-helix transcription factor E40). Total ANG protein levels were significantly lower in WTs when compared with those in 15 matched-paired NKs. ANG immunoreactivity was observed in the glomeruli, proximal tubules and vessels in the FKs and NKs, indicating that ANG plays a physiological role in the human kidney. ANG cellular localization and distribution in 27 WTs reflected the pattern observed in the FKs. ANG colocalized with LDHA in the perinecrotic areas of untreated WTs suggesting up-regulation by hypoxia. There was a significant correlation between CD31-MVD and ANG-MVD. ANG, CD31, VEGFA and BHLHE40 mRNA levels were significantly lower in 15 WTs compared with matched-paired NKs. Univariable and multivariable statistical analyses showed significant correlations between ANG and CD31, ANG and BHLHE40 mRNAs and a weaker relationship between ANG and VEGFA mRNAs. ANG expression in WTs recapitulates that seen during nephrogenesis, and correlation with CD31-MVDs and mRNAs is consistent with a contribution to angiogenesis in WTs. Our study contributes to the understanding of angiogenesis during development and in WTs.

Wilms' tumours (WTs), the most common malignant tumours of the paediatric kidney, occur at a frequency of eight cases per million in children younger than 15 years (Kaste et al. 2008; Davidoff 2009). WTs are developmental tumours that arise as a result of abrogated nephrogenesis (Rivera & Haber 2005). Nephrogenesis involves spatially and temporally regulated interaction of various transcription and growth factors that coordinate the differentiation of metanephric mesenchymal (blastemal) to epithelial (glomerular and tubular) structures. Approximately, 30–44% of WTs show foci of abnormally persistent nephrogenic cells called nephrogenic rests (NRs) in the adjacent renal parenchyma (Beckwith et al. 1990). They are considered to be precursors lesions of WTs (Beckwith et al. 1990). The two main types, perilobar nephrogenic rests (PLNR) and intralobar nephrogenic rests (ILNR), display different topographical, genetic and histological features (Fukuzawa & Reeve 2007). PLNRs are associated with perilobar WTs, which are composed predominantly of differentiated epithelial glomerular and tubular structures. ILNRs are generally associated with intralobar WTs, which contain predominantly stroma and heterologous elements, such as skeletal muscle (Beckwith et al. 1990).

Vascularization of the developing kidney is a complex process involving vasculogenesis (differentiation in situ from angioblasts) and angiogenesis (sprouting from pre-existing vessels) (Freeburg & Abrahamson 2003; Takano et al. 2007). Angiogenesis is also essential for the growth of solid tumours (Weis & Cheresh 2011). The angiogenic growth factor, vascular endothelial growth factor (VEGFA), plays a critical role in nephrogenesis (Tufro et al. 1999) and also correlates significantly with increased microvascular density (MVD) in WTs (Ghanem et al. 2003, 2011).

Angiogenin (ANG), a 14.1-kDa protein, plays an important role in tumour angiogenesis (Kishimoto et al. 2005; Gao & Xu 2008). Overexpression of ANG correlates with increased MVD in many epithelial malignancies, reviewed by (Tello-Montoliu et al. 2006) and the embryonic tumour, neuroblastoma (Dungwa et al. 2012). ANG is up-regulated by hypoxia in human cell lines (Hartmann et al. 1999; Koga et al. 2000; Pilch et al. 2001; Rajashekhar et al. 2005; Nakamura et al. 2006; Sebastia et al. 2009; Kishimoto et al. 2012). Furthermore, ANG expression correlates with hypoxia-inducible factor (HIF-1α) and basic helix-loop-helix transcription factor E40 (BHLHE40, also known as DEC1/Eip1/SHARP-2/Stra13/Clast5), a target gene of HIF-1α, (Ivanova et al. 2001; Yamada & Miyamoto 2005) in breast cancer (Chakrabarti et al. 2004; Campo et al. 2005).

Skoldenberg et al. (2001) showed that mean serum ANG levels are slightly elevated at the time of diagnosis in children with WTs compared with healthy controls. However, these authors did not study ANG expression in WT or non-neoplastic kidney tissue samples. Serum ANG levels are elevated as a non-specific acute phase response (Olson et al. 1998). ANG is secreted by a variety of non-neoplastic cell types including the liver (Olson et al. 1998), fibroblasts, smooth muscle cells, normal vessels and immune cells (Moenner et al. 1994). To our knowledge, there is no information regarding ANG expression in normal human kidney tissue or in WT samples. The aims of this study were as follows: (i) to perform a comprehensive analysis of ANG expression in fetal and childhood kidneys and in WTs and their precursor lesions, nephrogenic rests; (ii) to assess the potential contribution of ANG to MVD; and (iii) to examine the relationship of ANG with the hypoxia-inducible gene products lactate dehydrogenase A (LDHA), VEGFA and BHLHE40.

Methods

Tissue samples

All samples were obtained from the Paediatric Pathology files of the University Hospitals Bristol NHS Foundation Trust, Bristol, UK. The study was approved by the South Bristol and North Somerset Research Ethics Committee (09/H0106/4) for the use of anonymised tissue that was surplus to diagnostic requirements. Formalin-fixed and paraffin-embedded (FFPE) blocks of WTs and adjacent uninvolved/non-tumoral kidneys (NKs) were selected for CD31 and ANG immunostaining (IHC). Primary nephrectomy (n = 12) or nephrectomy after standard chemotherapy (n = 15) was performed according to the appropriate UKWT and Société International d'Oncologie Pédiatrique protocols (Kaste et al. 2008). The pathological characteristics of the cohort including the WTs with the unfavourable feature, diffuse anaplasia (Pritchard-Jones et al. 2012) are shown in Table 1. Sections from five fetal kidneys (FKs) ranging from 12 to 30 weeks of gestation were also examined by IHC.

Table 1.

Pathological characteristics of Wilms' tumours

Characteristics	Number
Chemotherapy status
Pre-CT	12
Post-CT	15
Size (cm)
Pre-CT median (range)	13 (5–18)
Post-CT median (range)	9 (3–20)
Weight (g)
Pre-CT median (range)	519 (265–1735)
Post-CT median (range)	291 (86–1848)
Stage
Pre-CT: 1,2,3,4,5	2,6,4,0,0
Post-CT: 1,2,3,4,5	3,4,4,2,2
Histological type
Pre-CT: Blastemal, epithelial, stromal, mixed	2,1,0,9
Post-CT: Blastemal, epithelial, stromal, mixed, R	2,1,2,5,5
Anaplasia
Focal	1
Diffuse	4
Risk group post-CT
High risk	4
Intermediate risk	11
NRs
Perilobar NR	8
Intralobar NR	4
Combined	2

Pre-CT, prechemotherapy; post-CT, postchemotherapy; R, regressive-type; NRs, nephrogenic rests.

Areas of necrosis were included with the viable areas for immunohistochemical (IHC) staining in 10 of 12 untreated WTs because markers of perinecrotic localization had been studied previously to elucidate their regulation under hypoxic conditions in WTs (Dungwa et al. 2011) and other tumours (Vleugel et al. 2005; Jubb et al. 2010).

Fifteen snap-frozen WTs and matched-paired NK samples were available for mRNA and protein assays. Only tumours that displayed >65% viability on frozen sections were chosen for mRNA and protein extraction. Human FK mRNA and human FK protein were purchased from AMS Biotechnology Limited (Abingdon, UK).

Enzyme-linked immunosorbent assay (ELISA)

Angiogenin tissue concentrations were measured using the human ANG Duoset enzyme-linked immunosorbent assay kit (DY265; R&D Systems, Abingdon, UK) as described previously (Dungwa et al. 2012). Standard curves were constructed using twofold serial dilutions of recombinant ANG (range 0–500 pg/ml). Optical densities were determined at λ = 450 nm with a corrective reading at λ = 570 nm using a spectrophotometer. Each protein extract was measured three times, and each test was performed in triplicate. The intra-assay and interassay variations were <10%.

Immunohistochemistry

Conventional 4-μm sections of FFPE tissue blocks were stained using a BOND-III automated immunostainer (Leica Biosystems, Milton Keynes, UK) for ANG (5 μg/ml rabbit anti-ANG polyclonal antibody (AB10603; Merck Millipore, Watford, UK), (Dungwa et al. 2012), LDHA (1 μg/ml rabbit anti-LDHA polyclonal antibody; Abcam 47010, Cambridge, UK) and CD31 (Dungwa et al. 2011). Endothelial cells in the vessels served as internal positive controls in each sample. Staining without a primary antibody was used as a negative control.

Assessment of immunoreactivity and MVD

The intensity of ANG immunoreactivity was evaluated within each component (i.e. blastemal, epithelial and mesenchymal stroma) and was graded as weak, moderate or strong for each positively stained element. The ANG staining intensity was strongest in the glomerular and vascular endothelial cells, and these cells were used for comparisons with other components.

Vascular hot spots were identified by scanning CD31-stained WT slides at low magnification (100×). CD31-lined vessels (CD31-MVD) were counted in the most vascularized areas (hot spots). At least four hot spots at 200× magnification were counted within each tumour (Dungwa et al. 2012). Any endothelial cell cluster, even without a lumen, was counted. For ANG-MVD, matching hot spots in serial sections were identified in slides stained for ANG. Vessels with ANG-stained endothelium and/or smooth muscle were counted.

CD31-MVD and ANG-MVD quantification was performed simultaneously by two observers to reach a consensus. The median values for CD31-MVD and ANG-MVD were recorded.

RNA extraction, reverse transcription and quantitative PCR

Total cellular RNA was extracted and reverse transcribed as described previously (Dungwa et al. 2011). Real-time quantitative PCR was performed using a Roche Lightcycler 480 (Roche Diagnostic, GmbH, Mannheim, Germany). The PCRs were performed with SYBR green, and the products were detected on the FAM channel. The sequences of the genes and the sizes of their amplification products are shown in Table 2. The cycling conditions consisted of an initial denaturation step of 95 °C for 3 min, followed by 40 cycles of denaturation at 95 °C for 20 s, annealing at 60 °C for 20 s and extension at 72 °C for 20 s, culminating in a melting curve analysis. Each sample was analysed in triplicate, and samples without reverse transcriptase or template were used as controls. Conventional RT-PCR was performed to confirm the specificity of each assay. The relative expression of the transcripts was quantified after normalizing the expression levels with the geometric means of the endogenous reference genes PPIA, GUS and TBP (Dungwa et al. 2011). Data analysis was performed using genex Pro 5.0 (MultiD, Göteborg, Sweden), which uses the ΔΔC_t relative quantification model (Dungwa et al. 2011).

Table 2.

Primers used in the study with amplification product size

Genes	Sense and antisense sequences	Amplicon
ANG	5′-GTGCTGGGTCTGGGTCTGAC-3′ 5′-GGCCTTGATGCTGCGCTTG-3′	201
CD31	5′-AAGGAACAGGAGGGAGAGAGTATTA-3′ 5′- GTATTTTGCTTCTGGGGACACT–3′	79
VEGFA	5′- GAGCCTCCCTCAGGGTTTC-3′ 5′-GGTTTGGATTAAGGACTGTTCT-3′	129
BHLHE40	5′-GTGCCTGCGTGTTGGTATAG-3′ 5′-CCCAAGTCTCCTGATGTCAAG-3′	84

ANG, angiogenin.

Statistical analyses

A paired Student's t-test was used to make within-patient comparisons between NKs and WTs with respect to the ANG protein levels. ANG- and CD31-MVDs were not normally distributed and non-parametric Mann–Whitney U-tests were used to compare the results of (unpaired prechemotherapy vs. postchemotherapy samples, and a Spearman's correlation coefficient was used to study their correlation (graphpad prism version 5.04 for Windows; GraphPad Software, San Diego, CA, USA).

The ANG, CD31, VEGFA and BHLHE40 mRNA expression levels were log₂-transformed (fold-change). All such transformed expression data were normally distributed; therefore, paired Student's t-tests were used to compare NKs and WTs. Pearson's correlation coefficients and regression analyses were used to study their interrelationship. A 5% level of significance was used throughout.

Multivariable analyses were conducted to explore the interrelationships between the mRNA levels and the distinction between WT and NK samples; the pairing of WT and NK samples was ignored here and the 30 samples were regarded as being independent. Hierarchical clustering was performed using Ward's linkage and Euclidean distance (genex pro5.0; MultiD). A canonical discriminant analysis was used to explore the capacity of ANG, CD31, VEGFA and BHLHE40 together to distinguish NK and WT samples using the spss v14.0 software package for Windows (SPSS Inc, Chicago, IL, USA). Comparative receiver-operator characteristic (ROC) curves were plotted with stata v10.1 software package for Windows (Stata; Statacorp LP, TX, USA).

Results

ANG protein levels in FK, NKs and WTs

Angiogenin levels were quantified by ELISA in the protein extracts of FK, 15 WTs and matched-paired NKs. The ANG protein level was 3.20 pg/μg in the FK at 20 weeks of gestation. The ANG levels were significantly decreased in the 15 WTs compared with the matched-paired NKs [median 13.27 (SD 4.5) pg/μg total protein vs. median 15.98 (SD 3.8) pg/μg total protein, respectively], Figure 1.

Figure 1.

ANG protein levels measured by ELISA in paired WT and NK samples (n = 15). Each data point represents the mean of triplicates from three separate experiments. ANG, angiogenin; WT, Wilms' tumours.

ANG immunolocalization in FKs, NKs and WTs

ANG is expressed in FKs and NKs

Angiogenin immunoreactivity was weak in the metanephric blastema and of moderate intensity in the comma- and S-shaped bodies and glomeruli of the FKs (Figures 2a and S1a). Vascular endothelial staining was also seen in the FKs (Figure 2a). The NKs showed ANG immunoreactivity in the glomeruli (Figure 2b). The lining of the capillary loops showed ANG immunoreactivity (Figure S1b). The cells at the periphery of the capillary loops also showed strong nuclear staining (Figure S1b). The proximal tubules showed a moderate intensity of cytoplasmic and nuclear staining, with strong labelling of the nucleoli (Figure 2b). Strong endothelial staining and weak-to-moderate smooth muscle and pericyte immunoreactivity in the arterioles and arteries were evident in the NKs (Figure 2b).

Figure 2.

Immunoreactivity in FK, NK and Wilms' tumours (WTs): (a) Fetal kidney shows strong angiogenin (ANG) expression in the developing glomerulus (left), the tubules show weak to moderate intensity of expression for ANG; the metanephric mesenchyme (black arrow) and a small vessel are weakly positive (red arrow). (b) Non-tumoral kidney in which the glomerulus (red arrow) shows strong and vessels (black arrows) show moderate intensity of staining. The proximal tubules show nucleolar reactivity (yellow arrows). (c) PLNR – the small immature glomeruli display ANG staining (inset), whereas the primitive tubules are negative. Normal kidney on the left shows strong ANG staining of a glomerulus (red arrow). (d) ILNR – low power: WT on right, ILNR, composed of variably sized tubular cysts separated by small amount of stroma on left. Inset: the WT displays strong ANG staining (inset, right), whereas the tubular cyst lining cells show weak ANG staining (left). (e) Triphasic WT reveals moderate intensity of ANG staining in blastema (B) and tubules (black arrows) with strong staining in the primitive glomeruli (red arrows). (f) WT, stromal-type, shows strong nuclear ANG staining in the vessels (red arrow), a rare tubule (black arrow) and skeletal muscle cells present in the vascular stroma. Inset shows striations in skeletal muscle cells. (g) WT blastemal cells (B) display ANG up-regulation in perinecrotic areas (N – necrosis). (h) WT blastemal cells (B) display LDHA up-regulation in perinecrotic areas (N – necrosis). FK, fetal kidney; LDHA, lactate dehydrogenase A; PLNR, perilobar nephrogenic rests; ILNR, intralobar nephrogenic rests.

ANG is expressed in NRs

The primitive glomeruli in the PLNRs showed strong immunoreactivity to ANG (Figure 2c). Most of the cysts in the ILNRs showed weak expression (Figure 2d).

ANG is expressed in WTs

Varying intensities of ANG immunoreactivity were observed in the three components of all WTs (Table 3). The intensity and distribution of ANG staining reflected that observed in FK tissue; that is, the glomeruli (Figure 2e) and endothelia (Figure 3a) displayed the strongest reactivity. The WTs displayed nuclear/nucleolar expression in the blastemal (Figures 2e and 3a, inset), immature glomerular (Figure 2e) and tubular (Figure 2e) components. The undifferentiated mesenchymal stromal cells displayed focal staining of weak-to-moderate intensity. The skeletal muscle cells in WTs showed strong ANG immunoreactivity (Figure 2f). There was no difference in the pattern or intensity of ANG immunoreactivity seen between the classical and anaplastic WTs.

Table 3.

ANG-immunoreactivity in human kidneys and pre- and post-CT WT samples

	FK	NK	WT
Tubules	+	++ to +++	+ to +++
Metanephric blastema	+ to ++	NP	+ to ++
Mesenchymal stroma	+	NP	++ to +++*
Glomeruli capillary EL	+++	+++	+++
Vascular EC, SMC, pericytes	− to +	++	+ to ++

CT, chemotherapy; ANG immunoreactivity – negative −, weak +, moderate ++, strong +++; FK, foetal kidney; NK, uninvolved/non-neoplastic kidney; WT, Wilms' tumour; NP, not present, non-neoplastic kidneys do not contain metanephric blastema; EL, endothelial lining; EC, endothelial cells; SM, smooth muscle cells; ANG, angiogenin.

^*Skeletal muscle cell differentiation, extensive in stromal-type of WTs, showed strong ANG immunoreactivity.

Figure 3.

(a) Wilms' tumours (WT) shows angiogenin (ANG)-positive vessels (black arrows) in the stroma (S) in a vascular hot spot. Inset shows an ANG-positive vessel on the right and nucleolar staining in blastemal cells on the left. (b) WT shows CD31-positive vessels (black arrows) in the vascular hot spot.

Relationship with hypoxia

Ten of 12 untreated WTs had foci of necrosis. Perinecrotic blastemal ANG up-regulation was seen in seven of 10 WTs. ANG (Figure 2g) and LDHA (Figure 2h) colocalized in six of the seven WTs. Areas of ANG up-regulation also colocalized with HIF-1α and VEGFA when compared with the corresponding areas in serial sections of WTs from our previous study (Dungwa et al. 2011), as shown in the Figure S2.

ANG- and CD31-MVDs are correlated in WTs

Angiogenin- (Figure 3a) and CD31-MVDs (Figure 3b) were decreased in post-CT compared with untreated WTs (Table 4). However, the decrease in MVDs following chemotherapy was not statistically significant. There was a significant correlation between ANG-MVD and CD31-MVD in all WTs (n = 27, r_s = 0.424, P = 0.028).

Table 4.

ANG-MVD and CD31-MVD in prechemotherapy and postchemotherapy Wilms' tumours samples

	Prechemotherapy (n = 12)	Postchemotherapy (n = 15)	P-value
	Median (range)	Median (range)
ANG-MVD	13.38 (0.75–34.75)	4.50 (1.5–25.25)	0.262
CD31-MVD	101.8 (41.40–169.0)	84.25 (31.75–130.8)	0.188

The Mann–Whitney U-test was used to compare MVD in prechemotherapy vs. postchemotherapy samples.

ANG, angiogenin; MVD, microvascular density.

ANG-, hypoxia- and angiogenesis-related gene expression in the NKs and WTs

ANG, CD31, VEGFA and BHLHE40 mRNAs were significantly down-regulated in the WTs relative to their matched-paired NKs (Table 5). Across the combined group of NKs and WTs, significant positive correlations were identified between ANG and both CD31 and BHLHE40 (Table 6). There was also a significant correlation between VEGFA and both CD31 and BHLHE40. CD31 and BHLHE40 levels were also strongly correlated. The relationship between ANG and VEGFA was weaker. In Figure 4a–c, the relationships between ANG and the other gene expression levels are further explored by linking together the pairs of results from the same kidneys. In these, it can be seen that the down-regulation of ANG in WT samples compared with NK samples accompanies similar down-regulations of BHLHE40, VEGFA and CD31.

Table 5.

Relative mRNA expression in FK, WTs and matched-paired NKs

Genes	FK	NKs Mean ± SD	WTs Mean ± SD	P-value*
ANG	3.449	4.190 ± 0.329	3.088 ± 1.335	0.009
CD31	2.586	3.578 ± 0.262	2.476 ± 1.161	0.005
VEGFA	2.744	3.218 ± 1.100	1.699 ± 0.856	<0.001
BHLHE40	2.459	4.285 ± 0.770	2.125 ± 1.195	<0.001

FK – Kidney at 30 weeks of gestation; WTs – Wilms' tumours (n = 15); NKs – matched-paired non-neoplastic kidneys (NKs).

^*Paired Student's t-test (NKs vs. WTs).

The mRNA data were log₂-transformed (fold-change). Values in bold indicate P < 0.05.

ANG, angiogenin; WT, Wilms' tumours.

Table 6.

Correlation between ANG mRNA and hypoxia-induced and angiogenesis-related genes in the combined sample of WTs and NKs (n = 30)

	CD31	VEGFA	BHLHE40
ANG
r	0.534	0.345	0.581
P	0.002	0.062	<0.001
VEGFA
r	0.447		0.685
P	0.0131	–	<0.001
BHLHE40
r	0.762
P	<0.001	–	–

r, Pearson's correlation coefficient. Values in bold indicate P < 0.05.

ANG, angiogenin; NKs, non-tumoral kidneys; WT, Wilms' tumours.

Figure 4.

Scatter diagrams showing the relationships between the relative mRNA expression levels in paired NK and Wilms' tumours (WT) samples; the pairs are linked (open circles – NK, solid circles – WT). (a) ANG and CD31, (b) ANG and VEGFA, (c) ANG and BHLHE40.

Unsupervised hierarchical clustering

Unsupervised algorithms seek out similarity between samples to determine whether they can be characterized as forming groups, called clusters. The dendrogram shows clusters of data according to how strongly they are correlated. The 15 WT and 15 NK plus 1 FK sample were included in analysis. The expression of the analysed genes varied in such a way that all of the samples were separated into two main branches by hierarchical clustering (HC) (Figure 5). One of the two clusters included all the 15 NKs and three WTs and was characterized by relatively high levels of CD31, ANG, VEGFA and BHLHE40. The remaining cluster contained 12 WTs (80% of WTs) clustered into the second branch with the FK sample. The second cluster had low expression levels of ANG, VEGFA, BHLHE40 and CD31. The HC of the genes indicated a strong relationship of ANG and BHLHE40 to CD31 and with each other.

Figure 5.

Heat map generated by unsupervised hierarchical clustering analysis with 15 Wilms' tumours (WTs), matched-paired uninvolved kidneys (NKs) and fetal kidneys (FKs) located on the vertical axis. Expression is relative to the FK value among the samples. Red corresponds to high, brown is intermediate and green is low expression (scale on left). The associated dendrogram is shown on the top. Clustering segregates the samples into two main subgroups: the first consists of NKs with relatively high expression, and the second consists of WTs with relatively low expression. WTs cluster with the FKs, indicating similarity in gene expression. The genes ANG, CD31, BHLHE40 and VEGFA are shown on the horizontal axis. ANG exhibits a closer relationship to CD31 and BHLHE40 than to VEGFA. ANG, angiogenin; FK, fetal kidney.

The canonical discriminant analysis determined the best weighted composite of all four mRNA expression levels (discriminant function scores) and confirmed good separation between the NK and WT samples (Figure S3).

Discussion

This study is the first to analyse ANG mRNA and protein in human samples of FKs, childhood NKs, WTs and their precursors, NRs. The main strength of our study is the comparison of ANG expression in WTs and matched-paired NKs to reduce the confounding variation inherent in biological samples.

We showed that ANG mRNA and protein levels were significantly higher in the NKs than in FKs. A previous report analysed ANG expression by Northern blot in the normal adult rat kidney but not in the neonatal rat kidney (Weiner et al. 1987). This study also showed that ANG transcript levels were low in the fetal rat liver, increased during development and reached maximal levels in the adult rat liver. This increase in ANG is consistent with its physiological role in organogenesis (Vallee & Riordan 1997). In our study, particularly strong expression was noted in glomeruli. Further studies utilizing double staining with a podocyte marker and CD31 are essential to identify the precise cellular location of ANG in the glomeruli. We also detected nuclear and nucleolar ANG immunoreactivity in the proximal tubules and vessels of NKs. These findings indicate that ANG plays a role in normal kidney function.

Wilms' tumours also showed nuclear and nucleolar ANG that was most marked in the blastema, tubules and vessels of WTs, the components that display the highest proliferative activity. These subcellular locations of ANG are consistent with its vital role in the induction of cellular proliferation in endothelial cells and tumours (Moroianu & Riordan 1994a; Tsuji et al. 2005; Yoshioka et al. 2006). ANG is translocated to the nucleus from the cell, vascular smooth muscle and endothelial cell surfaces in a cell-density-dependent manner (Moroianu & Riordan 1994b; Tsuji et al. 2005; Yoshioka et al. 2006). It accumulates in the nucleolus where it binds to the ribosomal DNA and stimulates ribosomal RNA (rRNA) transcription (Moroianu & Riordan 1994b; Xu et al. 2002). This is an important step for ribosome biogenesis, protein translation and cell growth. Inhibition of nuclear translocation by neomycin, or through site-directed mutagenesis, abolishes ANG-induced proliferation (Moroianu & Riordan 1994a). ANG-induced angiogenesis also involves intracellular signalling transduction pathways, such as protein kinase B/Akt that are independent of nuclear translocation (Kim et al. 2007).

The spatial localization and staining intensity of ANG in WTs recapitulated that in the FK. Moreover, the protein and mRNA levels in WTs were in concordance with those seen in the FKs. This result is not surprising given that WTs are developmentally dysregulated tumours that phenotypically and genetically mimic the FK (Rivera & Haber 2005). As a result of disrupted nephrogenesis, residual embryonal cells persist in the mature kidney as NRs. PLNRs mimic later phases of nephrogenesis where the principal development event is differentiation of nephrons from blastema, while ILNRs display the full spectrum of nephrogenesis being indicative of an earlier developmental disturbance (Beckwith et al. 1990). In our study, both types of NRs also revealed a pattern of expression similar to the counterparts in the WTs, in line with their role as the molecular genetic precursor lesions of WTs (Rivera & Haber 2005; Fukuzawa & Reeve 2007).

Hierarchical clustering analysis highlighted three postchemotherapy WTs that expressed relatively high levels of angiogenic factors when compared with other WTs. Two of the three WTs were of the stromal-type that showed extensive skeletal muscle differentiation. The third WT was of mixed type and included skeletal muscle elements in the vascular stroma. Strong ANG immunoreactivity was seen in the skeletal muscle and stromal vessels of all three WTs, and this was reflected in the higher levels observed by immunoassay in two of the three WTs. The stromal-type subtype comprises up to 10% of WTs and has an excellent outcome, which may be related to the low stage and low age of the patients (Verschuur et al. 2010). However, this subtype is resistant to chemotherapy, although it has a favourable prognosis in most cases (Maes et al. 1999).

Microvascular density reflects the net activity of all angiogenic and anti-angiogenic regulators (Bergers & Benjamin 2003). Analyses of MVD, either by CD34 or by CD31 IHC, are frequently employed as surrogate measures of angiogenesis (Vermeulen et al. 2002). As expected, when using the conventional approach of assessing MVD by CD31 staining, there was a decrease in WTs following chemotherapy. Of particular note, ANG-MVD correlated with MVD assessed in CD31-stained sections in WTs. This is in accord with our previous study in neuroblastoma, an embryonal tumour of childhood (Dungwa et al. 2012). We also noted a strong correlation between ANG and CD31 mRNA expression. Together, these data suggest that ANG may contribute to MVD in WTs. ANG functions as a potent angiogenic factor by various mechanisms (Tello-Montoliu et al. 2006; Gao & Xu 2008). It is essential for endothelial cell proliferation induced by VEGFA, α- and β-fibroblast growth factors and epidermal growth factor (Kishimoto et al. 2005). At the mRNA level, we also observed a correlation of borderline significance between ANG and VEGFA. ANG also induces other endothelial cell responses, such as adhesion (Soncin 1992), invasion, migration and tube formation (Tello-Montoliu et al. 2006; Gao & Xu 2008).

Increased microvascular density in WTs predicts relapse in patients with favourable histology (Abramson et al. 2003; Ozluk et al. 2006) and correlates with poorer patient survival (Skoldenberg et al. 2001; Ghanem et al. 2003; Ozluk et al. 2006). Based on our findings, prospective studies including a larger number of samples from patients with WTs may be useful to determine whether MVD or ANG levels have clinical utility in providing prognostic data on the behaviour of WTs.

Perinecrotic up-regulation of ANG was identified in a significant proportion of untreated WTs and colocalized with LDHA, a HIF-1α target gene protein. ANG also colocalized with HIF-1α and VEGFA in untreated WTs that were immunostained in a previous study (Dungwa et al. 2011). Our findings are consistent with its regulation by hypoxia-mediated mechanisms because studies by other groups have shown that ANG levels increase in human kidney proximal tubular epithelial cells, (Nakamura et al. 2006) and human skeletal myoblasts under hypoxic conditions (Perez-Ilzarbe et al. 2008). HIF-1α regulates ANG expression in motor neurons in hypoxia (Sebastia et al. 2009). A recent study has shown that ANG is also up-regulated by hypoxia in oral squamous cell carcinoma (OSCC) cell lines, and ANG correlates with HIF-1α expression in OSCC specimens (Kishimoto et al. 2012).

Stabilization of HIF-1α is increased in cells experiencing hypoxia, not only in rapidly growing tumours (Rey & Semenza 2010; Semenza 2012), but also during normal organogenesis (Lee et al. 2001). The HIFs (HIF-1α and HIF-2α) display a cell type- and stage-specific expression pattern during nephrogenesis (Bernhardt et al. 2006) and are responsible for the transcriptional activation of several angiogenic factors that play a role in kidney vascular development (Freeburg & Abrahamson 2003). CD31 expression increases with progressive development and maturation of the glomerulus (Takano et al. 2007). VEGA, a target gene of both HIFs (Hu et al. 2003), plays an important role in nephrogenesis and vasculogenesis in the FK (Tufro et al. 1999). Blockade of its receptor VEGFR2 in the first week of life adversely affects proximal tubule growth and leads to post-natal renal cyst formation and impaired glomerulogenesis in mice (Grath-Morrow et al. 2006). Further research to gain insight into ANG function in nephrogenesis and regulation by HIFs is essential.

Wilms' tumours displayed a distinct hypoxia-angiogenesis mRNA expression profile from NKs when compared by HC and canonical discriminant analysis. We identified a strong relationship between BHLHE40 and ANG at the transcriptional level, which is consistent with the previous studies at the protein level in epithelial cancers (Chakrabarti et al. 2004; Campo et al. 2005; Ivanova et al. 2005). BHLHE40 is a potent transcriptional regulator whose expression is regulated by hypoxia, as reviewed by Yamada (Yamada & Miyamoto 2005). Nuclear BHLHE40 protein is present in the glomerular endothelial and tubular cells in normal adult kidneys and has also been reported in two of three WTs studied (Ivanova et al. 2005). We also observed a strong relationship between BHLHE40 and CD31, consistent with a role for BHLHE40 in tumour angiogenesis. In support of our findings, a previous study showed BHLHE40 immunolocalization in the tumour-associated vessels in 78% of breast carcinomas (Chakrabarti et al. 2004). BHLHE40 mRNA expression was higher in the NKs than in the FKs in our study, a feature in agreement with the increasing levels observed during the development of the rat brain by Northern blot analysis (Rossner et al. 1997).

In conclusion, ANG expression (mRNA and protein) in FKs and NKs suggests that it plays a physiological role in the human kidney. ANG expression in WTs and their precursor lesions, NRs, recapitulates what is seen in renal embryogenesis. A notable observation in our study was the relatively high level of ANG in the stromal-type WTs with extensive skeletal muscle (rhabdomyomatous) component. Strong expression in perinecrotic areas and colocalization with LDHA, a HIF-1α target gene, is in line with up-regulation by hypoxia in untreated WTs. Correlation with CD31-MVD suggests that ANG may contribute to angiogenesis in WTs. CD31, ANG, VEGFA and BHLHE40 are down-regulated in WT compared with NK; taken together, they appeared to effectively separate these two groups. WTs display a distinct hypoxia-angiogenesis mRNA profile when compared with the NKs. Our study contributes to the understanding of angiogenesis during kidney development and in WTs.

Conflict of interest

We have no conflict of interest.

Financial support

Above and Beyond Charities of the University Hospitals Bristol NHS Foundation Trust

Supporting information

Additional Supporting Information may be found in the online version of this article:

Figure S1. (a) A low power view of a fetal kidney showing the developing glomeruli. The metanephrogenic (blastemal) is at the periphery (the top of the Figure). The four stages of glomerular development (Takano et al. 2007): vesicle, S-shaped, capillary loop and maturation are seen. Maturation progresses from the periphery and the most mature glomeruli are in the deeper portion (lower part of the Figure). ANG shows reactivity in the glomeruli from comma- and S-shaped stages onwards. (b–d) Glomeruli from the non-neoplastic kidney of an 11-month-old infant showing (b) ANG immunoreactivity (brown) in the lining of the capillary loops (red arrows). The cells at the periphery of the capillary loops (yellow arrows) also show nuclear staining. (c) nuclear WT1 staining of podocytes at the periphery of the capillary loops (yellow arrows). (d) CD31 staining (red) of lining of the capillary loops (red arrow). The yellow arrows show cells unstained for CD31 at the periphery of the capillary loops.

Figure S2. Relationship of ANG, HIF-1α and its target gene products (LDHA and VEGFA) to hypoxia indicated by up-regulation in blastemal cells in the corresponding perinecrotic areas in untreated WTs.

Figure S3. Canonical discriminant analysis (a) the histograms show the separation between the NKs (left) and WTs (right) obtained by using the canonical discriminant score (Can Disc Score) based on BHLHE40, VEGFA, ANG and CD31 mRNA. (b) ROC curve compares the canonical discriminant score derived from the four mRNA expressions above with that of BHLHE40 alone. The score itself was most strongly correlated with BHLHE40 (r = 0.93) and less strongly with the other variables (with VEGFA r = 0.67; CD31 r = 0.57 and ANG r = 0.49).