Germline VHL gene variant in patients with von Hippel-Lindau disease does not predict renal tumor growth

Abstract

Purpose:

To determine whether the type of VHL gene pathogenic variant influences the growth rate or CT enhancement values of renal lesions in VHL patients.

Materials and methods:

Thirty-two VHL patients (19 male) were selected from a prospectively maintained imaging database for patients that underwent surgical tumor resection between 2014 and 2016. One hundred and eleven VHL lesions were marked for resection and pathology analysis. Whole lesion volumetric segmentation was performed on nephrographic phase of the two most recent contrast-enhanced CT scans before surgery. Intensity distribution curves were obtained from segmentations. A linear mixed model, accounting for within-patient correlations, was used to compare the growth and enhancement differences between different germline pathogenic variant types.

Results:

There was no significant difference for the lesions’ total growth between different germline pathogenic variants (P value = 0.78). The median growth rate for all lesions was 1.7 cc/year (IQR 0.5, 3.9) with a baseline median size of 4.1 cm³ (IQR 1.7, 11.7). In complex lesions, the solid portion of the tumor demonstrated a higher growth rate (1.6 cc/year) than cystic portions (0.02 cc/year) which stayed relatively unchanged. Only one pathogenic variant (Splice donor) showed some levels of difference in its relative enhancement from other subtypes.

Conclusion:

The type of germline pathogenic variant on the VHL gene does not affect the growth rate or CT enhancement values of renal lesions in patients with VHL. The absolute growth rate of these tumors may be used in the scheduling of follow-up studies.

Von Hippel-Lindau (VHL) is a multisystem neoplastic syndrome that is defined as susceptibility to developing tumors in different parts of the body: most frequently hemangioblastomas in the central nervous system (CNS), angiomas in the retina, pancreatic cysts, neuroendocrine tumors, pheochromocytomas, renal cysts, and clear cell renal cell carcinomas (ccRCC) [1, 2]. VHL has an autosomal dominant pattern of inheritance, and is associated with pathogenic variants of the VHL tumor suppressor gene that encodes for the VHL protein. Several previous studies have proposed genotype–phenotype classifications for VHL disease [2–6]. The advent of genomics has allowed scientists to clone the VHL gene, which has made it possible to investigate the pathogenic variants associated with this syndrome. It has been shown that different classes of VHL pathogenic variants predispose individuals to distinct spectra of prognostic implications [5–7]. Therefore, it has been suggested that VHL pathogenic variations have the potential to be looked at as prognostic and predictive markers of the disease [7].

Renal cell carcinomas, with clear cell pathology, are the most frequent cause of mortality among VHL patients [8, 9]. Current management for VHL-associated renal tumors may include active surveillance and/or tumor removal by parenchymal sparing surgery [9, 10]. Duffey et al. suggested a 3-cm threshold for parenchymal sparing surgery in VHL patients to preserve renal function, as metastases are exceedingly rare with lesions < 3 cm [9]. Pre-operative understanding of growth kinetics of VHL renal lesions may aid in developing thresholds to trigger intervention for better patients counseling and individualized follow-up regimens [11].

Computed Tomography (CT) is the most commonly used imaging modality in the follow-up of renal lesions in VHL patients [12]. With expansion of diagnostic standard of VHL toward genomics, potentials of diagnostic imaging in understanding genomic properties of disease can be used in improving radiogenomic understanding of clear cell renal cell lesions [13]. Previous studies have shown associations between imaging features and underlying molecular and pathogenic variations of ccRCC by looking at tumoral Karyotypic alterations [14] and pathogenic variation of different genes [15]. However, to our knowledge, no previous study has investigated the effects of underlying germline pathogenic variations of VHL gene on imaging features of clear cell renal lesions associated with this syndrome. The aim of the present study was to assess the effect of germline pathogenic variant type on growth rate and CT enhancement values of VHL renal lesions and to examine the absolute growth rate of these lesions in a large cohort of VHL patients.

Methods

In this HIPAA compliant, Institutional Review Board approved study, a prospectively maintained urology oncology branch imaging database for patients that underwent surgical tumor resection between 2014 and 2016 was searched for patients with confirmed VHL diagnosis that had at least two time points contrast-enhanced CT scans before surgery. In a review of medical records of 66 patients with confirmed VHL diagnosis, 32 patients were identified, 13 females (mean age 50, range 30–77) and 19 males (mean age 39, range 25–61). Thirty-four patients were excluded from the final cohort because they did not have two contrast-enhanced CT scans prior to surgery. All lesions confirmed to be clear cell renal carcinoma by means of surgical pathogenical reports. Germline VHL gene pathogenic variants were identified in all patients. Pathogenic variants were categorized into seven subtypes, according to the location and type of the alteration on VHL gene, including AA deletion, frameshift, missense, nonsense, partial deletion, splice acceptor, and splice donor. Genetic consequences of these pathogenic variant categories are provided in Table 1.

Table 1.

Genetic consequence of pathogenic variant categories

Pathogenic variation	Result
Missense	Cause a change in a single amino acid
Frameshift	Disrupt the reading frame (the grouping of the codons)
Large (partial) deletion	Cause elimination of whole or a large region of chromosome
Nonsense	Results in a shorter protein by generating a chain-terminating codon
Splice acceptor	Caused by point mutations/insertion/deletion at acceptor site
Splice donor	Caused by point mutations/insertion/deletion at donor site
In-frame (amino acid) deletion/insertion	Affect one or few nucleotides

These definitions are according to our institutions methods for detection and classification of VHL pathogenic variants

Imaging

CT scans were obtained using multi-detector row CT scanners (Siemens biograph, Siemens SOMATOM definition flash, and Siemens SOMATOM Force) with a 192 * 0.6 mm collimation, 100 kV, 297 mA or 120 kV, 313 mA, 2–5 mm slice thickness with 1-mm increments and 0.5 gantry rotation time in a spiral mode. CT examination was performed according to NIH’s Clinical Center institutional CT protocol for renal masses, which consisted non-enhanced data acquisition, data acquisition with 50-s delay (corticomedullary phase), 2-min delay (nephrographic phase), and 8-min delay (excretory phase) if requested by referring clinician. To determine the onset of imaging after contrast injection bolus tracking algorithm was used. Iodinated contrast material (iopamidol 300 mg/mL; Isovue-300, Bracco Diagnostics, Melville, NY) was administered intravenously (1.8 mL/kg, up to a maximum of 130 mL, per our routine clinical protocol) at a rate of 3–4 mL/sec.

Image evaluation

The last two contrast-enhanced abdominal CT scans acquired before surgery were investigated for all patients. Tumors that were surgically removed were identified on both time point studies. Commercially available imaging analysis software (PACS, Carestream, version: 12.1) was used for volume and enhancement measurements of the lesions. On the baseline studies, lesions were volumetrically segmented by drawing free hand ROIs around every cross section, on axial view of 2.00-mm slice image series for the nephrographic phase by a trained research fellow (Fig. 1). The same lesion was identified on subsequent scan and segmented with the same procedure. Contours were confirmed or corrected if necessary by a fellowship trained radiologist (with 4 years of experience) for accuracy. Intensity distribution curves were produced from segmentations, providing the number of image pixels for each attenuation value (measured in HU). The volume of lesions was calculated by using the total number of pixels in each lesion and the individual pixel volume specific for each study. According to the accepted upper limit of water density, a 20-HU threshold was considered for differentiating lesion components [16]. Parts of the lesion with individual pixel attenuation similar to that of water ( < 20HU) were considered cystic. Other parts of the lesion were considered solid if they showed attenuation greater than that of water ( > 20 HU). Oval regions of interest (ROI) (10 × 5 mm) were also placed on normal renal cortex in each study for normalization. The ratio of lesion to renal cortex was used to calculate relative enhancement for each lesion.

Fig. 1.

Axial contrast-enhanced (nephrographic phase) abdominal CT shows freehand ROIs drawn on cross section of a lesion. Prior (A) and follow-up (B) screening before surgery, in a 51-year-old male with VHL disease.

Statistical analysis

R Statistical Software (Foundation for Statistical Computing, v3.3.1, Vienna, Austria) was used for all statistical analysis. Summary statistics were expressed as medians with interquartile ranges. Linear mixed models accounting for within-patient correlations were used to determine whether tumor growth and enhancement are different across genetic pathogenic variants. The fixed effects included categorical variables for genetic pathogenic variants (AA deletion, frameshift, missense, nonsense, partial deletion, splice acceptor, and splice donor). To account for the correlation of measurements within a patient, we included a random effect for patients. All models were calculated using the R “lme4” package with Tukey’s test for post hoc pairwise comparisons [17]. A statistically significant difference was defined as a two-side P value < 0.05. No multiplicity correction was made due to the exploratory nature of the study.

Results

Out of 132 CT scans from 66 patients in two time points before surgery, 132/132 (100%) of the studies included unenhanced data. 79/132 (60%) of the scans had contrast-enhanced phase (including corticomedullary and nephrographic phases). Thirty-two patients (64/132 scans) that had contrast-enhanced study available on both time points before the surgery were selected for analysis. 36/132 (27%) of studies included excretory phase in addition to the latter two contrast-enhanced phases. All final 64 CT studies were performed between 2011 and 2016. The mean difference between two scans was 401 days (range 67–1778 days). A total number of 111 resected lesions were identified. Analyzing lesion attenuation, with 20-HU upper limit threshold for water, showed 7 lesions that were completely solid and 104 complex lesions with less than 15% of cystic volume in all cases. Germline VHL gene pathogenic variants included a total of 7 types with the following distribution among lesions: 6 AA deletion (2 patients), 4 frameshift (3 patients), 52 missense (12 patients), 5 nonsense (3 patients), 37 partial deletion (10 patients), 3 splice acceptor (1 patient), 4 splice donor (1 patient). Lesions at baseline had an average size of 4.1 (IQR 1.7, 11.7) cm³. The median growth rate for lesions was 1.7 cc/year (IQR 0.5, 3.9). This number for the solid portion of lesions was 1.6 cc/year (IQR 0.5, 4) and for the cystic portion was 0.02 cc/year (IQR – 0.003, 0.2). There was no significant difference for the lesions’ total growth between different pathogenic variant types (P value = 0.78) (Fig. 2). Pairwise review of genetic pathogenic variants also did not show any significant difference in their growth rate. Baseline and follow-up tumor volume (cm³) and growth rate (cc/year) for pathogenic variant types are shown, median (interquartile range), in Table 2. Component analysis of lesions showed an average of 94% (range 85%, 100%) solid portion for lesions at the baseline. There was no significant growth difference among pathogenic variant types when cystic (P value = 0.74) or solid (P value = 0.67) lesions were analyzed (Figure 3).

Fig. 2.

Total growth rate vs. pathogenic variant type.

Table 2.

Median (interquartile range) lesion volume (cm³) at baseline and follow-up, and growth (cc/year) for all pathogenic variant types

Volume		AA deletion	Frameshift	Missense	Nonsense	Partial deletion	Splice acceptor	Splice donor
Baseline	Median	4.6	5.8	4.6	5.2	3.2	0.7	6.3
	Interquartile range	2.1, 12.6	3.2, 8.2	1.8, 13.2	1.3, 6.0	2.0, 8.0	0.6, 6.6	4.3, 72.4
Follow-up	Median	8.1	11.1	6.2	5.6	5.6	1.6	10.1
	Interquartile range	2.8, 16.1	9.8, 12.0	2.4, 13.5	5.1, 10.0	3.1, 15.7	1.5, 14.8	7.9, 69.6
Change	Median	2.9	2.3	1.1	0	2.0	1.6	1.0
	Interquartile range	0.8, 6.0	1.9, 2.9	0.1, 3.6	− 0.3, 2.8	1.3, 4.5	1.4, 12.6	− 1.5, 1.6

Fig. 3.

Component specific lesion growth vs. pathogenic variant type. For solid (A) and Cystic (B) growth rates.

Tumors at the baseline did not show a significant group difference in their relative enhancement (P value = 0.054) between pathogenic variant types. Further pairwise comparisons of relative enhancement of tumors showed that partial deletion was 32 HU higher than splice donor (95% CI – 0.9 to 66, P = 0.059). Splice acceptor was 45 HU higher than splice donor (95% CI – 1.4 to 92, 0 = 0.06). Difference in relative enhancement among pathogenic variant types was statistically significant in the follow-up study (P value = 0.03) (Fig. 4). At follow-up study, tumors with splice acceptor pathogenic variation showed 50-HU higher enhancement than splice donor (95% CI 4–96, P = 0.02). Frameshift variant tumors showed 42 HU higher than splice donor (95% CI 3–81, P = 0.03). Partial deletion group was 36 HU higher than splice donor (95% CI 3–70, P = 0.02). Missense variant was 35 HU higher than splice donor (95% CI 2–68, P = 0.03). The average change in lesions’ relative enhancement was – 1.5 HU (range – 25, 36 HU), and there was no significant difference among different pathogenic variant classes regarding change in relative enhancement (P value = 0.89). Relative lesion enhancement at the baseline and follow-up scan, and change between the two time points are reported, median (interquartile range), in Table 3.

Fig. 4.

Lesion relative enhancement (HU) vs. pathogenic variant type. For baseline (A) and follow-up (B).

Table 3.

Median (interquartile range) relative enhancement (HU) for all pathogenic variant types

Growth rate		AA deletion	Frameshift	Missense	Nonsense	Partial deletion	Splice acceptor	Splice donor
Baseline	Median	4.6	5.8	4.6	5.2	3.2	0.7	6.3
	Interquartile range	2.1, 12.6	3.2, 8.2	1.8, 13.2	1.3, 6.0	2.0, 8.0	0.6, 6.6	4.3, 72.4
Follow-up	Median	8.1	11.1	6.2	5.6	5.6	1.6	10.1
	Interquartile range	2.8, 16.1	9.8, 12.0	2.4, 13.5	5.1, 10.0	3.1, 15.7	1.5, 14.8	7.9, 69.6
Change	Median	2.9	2.3	1.1	0	2.0	1.6	1.0
	Interquartile range	0.8, 6.0	1.9, 2.9	0.1, 3.6	− 0.3, 2.8	1.3, 4.5	1.4, 12.6	− 1.5, 1.6

Discussion

Individuals with a pathogenic variant allele of VHL tumor suppressor gene are highly likely to develop VHL manifestations, which include CNS hemangioblastomas, retinal angiomas, pheochromocytomas, clear cell renal carcinomas, and/or pancreatic cysts [5]. Even though one pathogenic variant allele is enough for inheritance, inactivation of both alleles is required for tumor formation. Tumorigenesis in patients that are heterozygotic for this alteration is explained by the two-hit hypothesis proposed by Knudson and Strong [12]. About 200 distinct pathogenic variations have been found among more than 900 families in whom VHL has been identified [18]. Pathogenic variation of this gene, which is located on the short arm of chromosome 3, results in abnormal or absent VHL protein, which results in increased levels of hypoxia-inducible factors (HIF) mediated angiogenesis and cell proliferation, hence the hypervascular nature of VHL lesions [2, 9, 19].

The pathogenic variant spectrum includes missense (52% of patients), frameshift (13%), large (partial) deletions (11%), nonsense (11%), splice site (Splice acceptor/donor) (7%), and in-frame (amino acid) deletions or insertions (6%). These pathogenic variants are identifiable in almost all patients with classic multi-organ involvement of VHL. Deletion, nonsense microdeletion, and insertion pathogenic variants have been associated with a decreased incidence of pheochromocytomas, and patients with missense and point pathogenic variants had a higher incidence [5, 7]. Differences in site-specific symptoms of the disease could rise from differences in biochemical products of these variations [20].

VHL-associated renal cell carcinomas are uniformly of the clear cell histologic phenotype and are the major cause of mortality among VHL patients. There has been a 70% lifetime risk of RCC reported for most common subtypes of VHL. However, risk of RCC is different among various forms of the disease [21]. While inactivation of VHL gene has a clear role in development of clear cell renal lesions association with VHL, it is yet unknown whether the type or severity of the damage to the gene could provide useful predictive or prognostic information in patients with this syndrome [7]. It is suggested that understanding the behavior of these lesions can help in delaying or preventing surgical treatment, retaining maximal renal function [9, 22], and also could assist clinicians in categorizing patients for active surveillance by prolonged observation or delayed intervention for non-interventional strategy [22].

Previous works on tumor behavior have shown a relationship between tumors’ maximal diameter and malignant features at surgery, including metastases and high Fuhrman grade [22]. Since no other radiographic or histological features have been associated with the malignant potential of VHL lesions, tumor size is currently the most clinically relevant factor that helps guide clinicians to the appropriate time for intervention [22]. We used genetic data available from VHL patients to inspect the effect of pathogenic variant type on renal lesion phenotype, including their growth rate and enhancement characteristics.

Molecular findings have shown that the presence of VHL gene alteration did not show an association with various tumor characteristics, including tumor cell proliferation rate, angiogenesis, tumor stage, and grade [23]. Our results indicate that all VHL renal lesions have the same growth rate across all pathogenic variant types. The finding that renal lesion growth rate does not change, regardless of the severity of the effect of pathogenic variation of the gene, may suggest that the type of pathogenic variation of VHL gene does not affect the progression of these lesions.

Previous reports on growth of VHL renal lesions showed that cystic lesions have a different growth rate compared to solid lesions [1]. It has also been observed in serial CT studies that in complex lesions the solid component enlarges and the cystic portion regresses [12]. By enhancement analysis of each individual pixel, we found out that in complex lesions the solid portion of the lesion grew significantly faster than the cystic portion that stays relatively unchanged. However, change in relative enhancement for the total body of the lesion was not significant over time. This could be due to the smaller contribution that the cystic portion of lesions had in this number at the baseline study. In all of the complex lesions in this study, the cystic portion occupied a very small (less than 15%) part of the lesion.

Relative enhancement of lesions was not different between pathogenic variant types at the baseline. Pairwise comparisons showed that difference in enhancement in follow-up studies was due to only one of the variant types (splice donor). However, because of small sample size (n = 4) in this subtype, it is hard to come to any conclusion regarding this difference. The latter may suggest pathogenic variation types do not differently affect the characteristics of VHL lesions that correspond to CT enhancement values.

There were limitations to this study. Our sample did not include all pathogenic variant types possible for VHL disease. This study did not control for possible nongermline pathogenic variations of VHL gene or other genes that could play a role in growth rate or enhancement of renal lesions. CT scans were obtained across a rather long period of time, resulting in use of multiple types of scanners. Even though attenuation values were normalized using cortex of the kidney in each study, this could have prevented from achieving uniform enhancement across all scans. Another limitation was the small sample size. The number of lesions in some pathogenic variant types (e.g., splice acceptor) was significantly lower than other groups. Latter undermines the power of the statistical test that was performed for this study. Therefore, lack of significant difference in both growth rate and CT enhancement values, in certain subtypes, could be because of their small sample size.

Understanding a growth pattern for VHL renal lesions could help clinicians in categorizing these lesions for active surveillance and predicting the appropriate time for intervention. We found out that VHL lesions grow uniformly regardless of type of germline VHL gene variant. Germline pathogenic variants did not affect the enhancement levels on CT scans. Further study, with larger cohort and more subtypes of pathogenic variations, is necessary to validate our findings. Our findings could be helpful to further our understanding of the effect of pathogenic variations on characteristics of VHL renal lesions.