Characterizing the ADPKD-IFT140 Phenotypic Signature With Deep Learning and Advanced Imaging Biomarkers

Ahmad Ghanem; Fadi George Munairdjy Debeh; Abdul Hamid Borghol; Nikola Zagorec; Amanda L Tapia; Byron Smith; Stefan Paul; Abdul Basit; Bassel AlKhatib; Nay Nader; Marie Therese Bou Antoun; Adriana V Gregory; Hana Yang; Rachel S Schauer; Neera K Dahl; Christian Hanna; Vicente E Torres; Timothy L Kline; Peter C Harris; Emilie Cornec-Le Gall; Fouad T Chebib

doi:10.1016/j.ekir.2025.04.062

. 2025 May 7;10(8):2690–2707. doi: 10.1016/j.ekir.2025.04.062

Characterizing the ADPKD-IFT140 Phenotypic Signature With Deep Learning and Advanced Imaging Biomarkers

Ahmad Ghanem ¹, Fadi George Munairdjy Debeh ¹, Abdul Hamid Borghol ¹, Nikola Zagorec ^2,³, Amanda L Tapia ⁴, Byron Smith ⁴, Stefan Paul ¹, Abdul Basit ¹, Bassel AlKhatib ¹, Nay Nader ¹, Marie Therese Bou Antoun ¹, Adriana V Gregory ^5,^6,⁷, Hana Yang ^5,⁷, Rachel S Schauer ^5,⁷, Neera K Dahl ^5,⁷, Christian Hanna ^5,^7,⁸, Vicente E Torres ^5,⁷, Timothy L Kline ^5,^6,⁷, Peter C Harris ^5,⁷, Emilie Cornec-Le Gall ², Fouad T Chebib ^1,^∗

PMCID: PMC12348249 PMID: 40814641

Abstract

Introduction

ADPKD-IFT140 is the third most common disease-causing variant in autosomal dominant polycystic kidney disease (ADPKD) after ADPKD-PKD1 and ADPKD-PKD2. This study aimed to characterize the clinical presentation, progression, and distinctive imaging phenotype of ADPKD-IFT140.

Methods

This retrospective cohort study included patients with disease-causing variants in IFT140, nontruncating PKD1 (PKD1NT), or PKD2. Patients were matched by sex (48.1% male), age (mean [SD]: 57.7 ± 13.3 years), and height-adjusted total kidney volume (TKV; htTKV) (median [Q1–Q3]: 572.9 [314.1–1137.9] ml/m). Two predictive models were developed in the development cohort (n = 81): a deep-learning model incorporating cyst-parenchymal surface area (CPSA) and cystic index, and a practical model using percentage of TKV_ellipsoid occupied by the 2 largest cysts, with cyst volumes estimated from cyst diameters using the formula $V = \frac{π}{6} (d_{1}^{3} + d_{2}^{3})$ . Models were validated in an internal specificity cohort (n = 569) and an external sensitivity cohort (n = 36).

Results

Patients with ADPKD-IFT140 exhibited fewer (median cyst number: 42) but larger cysts (average cyst volume: 12.1 ml), with 88.9% having no liver cysts, compared with ADPKD-PKD1NT and ADPKD-PKD2. The estimated glomerular filtration rate (eGFR) of decline was slower in ADPKD-IFT140 (−0.69 ml/min per 1.73 m²/yr) than in ADPKD-PKD1NT (−1.62, P = 0.006) and in ADPKD-PKD2 (−0.90, P = 0.737). The deep-learning model demonstrated an area-under-the-curve (AUC) of 0.949 for distinguishing ADPKD-IFT140 patients in the development cohort, and 88.9% specificity in the internal cohort. A volume-to-TKV ratio ≥ 18.6% identified ADPKD-IFT140 with an AUC of 0.814 and demonstrated 72.2% sensitivity in the external cohort.

Conclusion

We provide a detailed characterization of the ADPKD-IFT140 phenotype that can be distinguished using a practical or deep-learning segmentation model applicable in diverse clinical settings.

Keywords: ADPKD, atypical, CKD, cyst, cyst segmentation, IFT140, imaging biomarkers, PKD, polycystic kidney disease, predictive model, total kidney volume

Graphical abstract

ADPKD is the most common inherited kidney disorder, often leading to kidney failure by the fifth decade of life.1, 2, 3 Disease-causing variants in PKD1 and PKD2 account for approximately 78% and 15% of cases, respectively.⁴ PKD1 variants are divided into truncating (PKD1T) and PKD1NT variants, with PKD1NT generally associated with a milder disease course and later onset of kidney failure.4, 5, 6 However, not all patients with ADPKD have detectable PKD1 or PKD2 variants. A recent study with whole-exome sequencing identified PKD1 or PKD2 variants in 161 of 235 patients (68.5%) with ADPKD, whereas 8.1% carried other minor variants associated with cystic kidney diseases, and 23.4% had undetected mutations, although the latter may be an overestimation because not all minor genes were analyzed.⁷ Identified minor loci include IFT140, GANAB, ALG5, ALG8, ALG9, DNAJB11, and NEK8.7, 8, 9, 10, 11 IFT140, a core component of the intraflagellar transport-complex A, is essential for ciliary function.12, 13, 14 Although biallelic IFT140 variants cause multisystemic ciliopathies,15, 16, 17 monoallelic loss of function (LoF) are associated with ADPKD and termed ADPKD-IFT140.¹⁸^,¹⁹ Accounting for approximately 2% of ADPKD families,¹⁸^,²⁰ ADPKD-IFT140 is the third most common cystic kidney disease variant identified in the Genomics England 100K and the UK Biobank cohorts.¹⁸

ADPKD-IFT140 is characterized by large cysts, mild kidney dysfunction, and minimal liver cystic involvement.¹⁸ Its cyst development mechanism is not completely understood and is thought to be related to increased cell proliferation and signaling pathway alterations, including Wnt, Hedgehog, and Hippo pathways.¹⁸^,²¹ Previous studies by Senum et al.¹⁸ and Zagorec et al.²² revealed a broad clinical spectrum in ADPKD-IFT140 ranging from mild to advanced chronic kidney disease and small to large TKV. Our study aimed to clinically characterize and define a phenotypic distinctive signature of patients with ADPKD-IFT140, assess their disease progression, and compare them with patients with ADPKD-PKD1NT and those with ADPKD-PKD2.

Methods

Study Population

Development Cohort

A cohort of patients with ADPKD, genetic testing, and either a computed tomography scan with contrast or magnetic resonance imaging (MRI) was identified from the Mayo Clinic databases (N = 1020).²³^,²⁴ The inclusion criteria were as follows: (i) aged ≥ 15 years, (ii) monoallelic LoF variant in IFT140 or disease-causing variants in PKD1NT, or PKD2, and (iii) imaging available before clinical interventions, including nephrectomy, preemptive kidney transplantation, kidney replacement therapy, tolvaptan therapy, or any cyst volume reduction interventions. The included cohort (n = 329) was divided into 3 groups (ADPKD-IFT140, ADPKD-PKD1NT, and ADPKD-PKD2) and manually matched based on sex and age at imaging (± 3 years), followed by htTKV (selecting the htTKV most comparable to the ADPKD-IFT140 cases).

Internal Specificity and External Sensitivity Cohorts

An internal specificity cohort was developed by including unmatched patients with ADPKD-PKD1NT and those with ADPKD-PKD2 in addition to a cohort of patients with ADPKD-PKD1T who met the inclusion criteria (n = 569). An external sensitivity cohort included 36 patients with ADPKD, available MRI, and monoallelic LoF variant in IFT140 from a cohort of 1359 individuals recruited in the Genkyst cohort, and in different French centers.²²

Genetic Analysis and the Identification of the Disease-Causing Variants

Monoallelic LoF variants in IFT140 and disease-causing variants in PKD1 and PKD2 were identified via targeted next-generation sequencing using 137, 357, or 385 gene panels, whole-exome sequencing, or whole-genome sequencing as described previously.⁵^,¹⁸^,25, 26, 27, 28, 29

Segmentation and Imaging Biomarkers

A deep-learning planimetry-based approach, known for its accuracy,³⁰ was used to segment kidney images from the patient cohort. Segmentation outputs were quality-checked by a blinded medical imaging expert using an in-house–developed tool (i.e., PKD-GUI). TKV was programmatically calculated by multiplying the total number of labeled voxels by voxel volume. Patients were then classified as typical (Mayo Imaging class 1) or atypical (class 2) based on predefined imaging criteria.³¹ Class 1 were further divided into 5 subclasses by htTKV and age, whereas class 2 was subdivided into 2A (nonatrophic) or 2B (atrophic), based on predefined criteria.³¹ TKV_ellipsoid was additionally obtained by measuring kidney width, depth, and length and applying the ellipsoid equation.³¹ For the cyst biomarker analysis, the segmented images were then processed through a cyst segmentation model and quality checked by a blinded medical imaging expert using PKD-GUI.³⁰ Advanced imaging biomarkers, including total cyst volume (TCV), total cyst number (TCN), renal parenchymal volume, CPSA, and cystic index (TCV/TKV) were calculated using predefined Python formulas.³⁰ Cysts < 0.03 ml were excluded to minimize noise. Observed individual cyst volumes were also programmatically measured by multiplying the total number of cyst voxels by the voxel volume. Observed cyst diameters were manually measured, and calculated individual cyst volumes were then calculated using the formula $V = \frac{π}{6} d^{3}$ , with the assumption that each cyst is a sphere, and that the diameter, d (in cm) is the length of the cyst measured on coronal section. Several biomarkers (TKV, TCV, and renal parenchymal volume) were adjusted for height (m).

Data Collection

Data were collected through a comprehensive review of electronic medical records by 2 medical doctors. Demographic details were extracted electronically, whereas manual data collection included the history of hypertension, diabetes mellitus, urinary tract infections (diagnosed as cystitis or pyelonephritis), and urinary stone disease (symptomatic or asymptomatic kidney stones detected through chart review or imaging). The date of onset or first episode for these conditions, if present, was also obtained. Patient weight and height were collected within 1 year of the imaging date, defined as the date at baseline. For the external cohort, patients with missing heights (n = 9), a height of 1.75 m was assumed for males and 1.65 m for females. Adjusted body weight was calculated by subtracting kidney weight (assuming tissue density = 1 g/cm³), and adjusted body mass index was derived using adjusted weight divided by height squared.³² Baseline laboratory parameters, includingserum hemoglobin, bicarbonate, blood urea nitrogen, albumin, calcium, uric acid, fasting blood glucose, and urinary white blood cell count, were recorded. Serum creatinine levels within 1 year of imaging were identified as baseline values, with subsequent tracking until the last available value before reaching the predefined events described above, if present. The follow-up duration was measured from baseline to the last available creatinine value. The eGFR was calculated using the Chronic Kidney Disease Epidemiology Collaboration 2021 equation.³³ This retrospective cohort study was conducted following the guidelines of the Mayo Clinic Institutional Review Board.

Study Outcomes

The study's primary outcome was to define a phenotypic signature for ADPKD-IFT140, using clinical and imaging biomarkers. Secondary outcomes included comparing eGFR rate of decline between patients with ADPKD-IFT140 and those with other PKD variants.

Statistical Analysis

Continuous variables were summarized as mean ± SD or as median with interquartile range (Q1–Q3). Categorical variables were expressed as counts (n) and percentages (%). Comparisons between groups were conducted using t tests for continuous variables and chi-square tests for categorical variables, and effect sizes were calculated using Cohen’s d or Cramer’s V. A linear mixed effects model with random intercepts was used to analyze eGFR decline, estimating genotype-specific slopes with 95% confidence intervals (CIs). Correlations between imaging biomarkers were assessed using the Spearman method. The performance of advanced imaging biomarkers for differentiating ADPKD-IFT140 from ADPKD-PKD1NT or ADPKD-PKD2 was assessed using univariate and multivariate logistic regression, with significance evaluated by Wald statistics. AUC (95% CI) and both sensitivity and specificity at Youden’s index were calculated. A multivariable model was then developed (detailed variable selection process is described in the Supplementary Statistical Analysis). The cumulative volume of the largest cysts relative to TKV was assessed as a biomarker for distinguishing ADPKD-IFT140, evaluating AUC across the largest cyst to the top 10. The best-performing biomarker obtained via deep-learning segmentation was then compared with manual measurements using Spearman correlation and a Bland-Altman plot (mean ± 2 SD) for agreement and bias assessment. Threshold specificity was tested in the internal cohort, and sensitivity of the manually cumulative volume of the 2 largest cysts relative to TKV was tested in the external cohort. Statistical significance was based on a 2-tailed P-value < 0.05. Analyses were conducted using R statistical software v.4.2.2 (R Project, Vienna, Austria), JMP software v.18.0 (SAS Institute, Inc., Cary, NC), and Graphpad prism v.10.0.0 (GraphPad Software, San Diego, CA).

Results

Baseline Characteristics

Among the 326 patients with available imaging and confirmed variants in IFT140, PKD1NT, or PKD2 genes, 27 from each group were included after matching based on age, sex, and htTKV (Supplementary Figure S1). Baseline demographics, and kidney characteristics are shown in Table 1. Among the included patients, 48.1% were male, with a mean age at imaging of 57.7 (± 13.3) years. Patients with ADPKD-IFT140 were significantly less likely to have a family history of ADPKD (48.1%) than those with ADPKD-PKD2 (96.3%, P < 0.001) and those with ADPKD-PKD1NT (81.5%, P = 0.010). More patients with ADPKD-IFT140 underwent cyst volume reduction interventions (18.5%) than those with ADPKD-PKD2 and those with ADPKD-PKD1NT (3.7%); whereas none of the patients with ADPKD-IFT140 received kidney replacement, preemptive kidney transplant, nephrectomy, or tolvaptan (Supplementary Figure S2). Hypertension prevalence was significantly lower in ADPKD-IFT140 (40.7%) than in ADPKD-PKD2 (81.5%, P = 0.002) and in ADPKD-PKD1NT (88.9%, P < 0.001), with a significantly higher age at hypertension diagnosis in patients with ADPKD-IFT140 (55.2 ± 6.6 years) compared with patients with ADPKD-PKD2 (46.2 ± 9.6, P = 0.010) and those with ADPKD-PKD1NT (45.5 ± 14.9, P = 0.049). Adjusted body mass index and follow-up length were similar across groups (Table 1).

Table 1.

Baseline demographics of the development cohort

Clinical characteristics	ADPKD variants			IFT140 vs. PKD2		IFT140 vs. PKD1NT
Clinical characteristics	IFT140	PKD2	PKD1NT	P-value^a	ES^b (CI)	P-value^a	ES^b (CI)
n	27	27	27
Sex				1.000	0 (0–1)	1.000	0 (0–1)
Males, n (%)	13 (48.1%)	13 (48.1%)	13 (48.1%)
Females, n (%)	14 (51.9%)	14 (51.9%)	14 (51.9%)
Race				0.582	0 (0–1)	0.582	0 (0–1)
White, n (%)	23 (85.2%)	25 (92.6%)	25 (92.6%)
Non-White, n (%)	1 (3.7%)	1 (3.7%)	1 (3.7%)
Unknown, n (%)	3 (11.1%)	1 (3.7%)	1 (3.7%)
Age at baseline, yrs, mean (SD)	58.0 (13.7)	57.8 (12.9)	57.1 (13.7)	0.951	0.02 (-0.52–0.55)	0.812	0.06 (-0.47–0.6)
Relative with ADPKD, n (%)	13 (48.1%)	26 (96.3%)	22 (81.5%)	< 0.001	0.52 (0.28–1)	0.010	0.32 (0–1)
Adjusted BMI^c at baseline, kg/m²
n	24	26	27
Mean (SD)	26.9 (4.5)	26.9 (5.0)	26.4 (4.8)	0.983	0.01 (−0.55 to 0.56)	0.708	0.11 (−0.44 to 0.65)
Diabetes mellitus, n (%)	2 (7.4%)	5 (18.5%)	3 (11.1%)	0.224	0.09 (0–1)	0.639	0 (0–1)
Age at diabetes mellitus diagnosis, yrs
n	2	4	3
Mean (SD)	74.5 (3.5)	49.8 (26.9)	57.3 (18.6)	0.289	1.29 (−0.46 to 2.92)	0.307	1.28 (−0.76 to 3.16)
Hypertension, n (%)	11 (40.7%)	22 (81.5%)	24 (88.9%)	0.002	0.40 (0.14–1)	< 0.001	0.49 (0.25–1)
Age at hypertension diagnosis, yrs
n	11	21	24
Mean (SD)	55.2 (6.6)	46.2 (9.6)	45.5 (14.9)	0.010	1.09 (0.34–1.83)	0.049	0.84 (0.18–1.48)
Urinary tract infection, n (%)	10 (37.0%)	5 (18.5%)	10 (37.0%)	0.129	0.16 (0–1)	1.000	0 (0–1)
Age at first UTI episode, yrs, mean (SD)
n	10	5	9
Mean (SD)	55.4 (22.5)	50.6 (13.6)	61.9 (11.1)	0.671	0.26 (−0.74 to 1.25)	0.444	−0.37 (−1.26 to 0.54)
Kidney stone detection, n (%)	6 (22.2%)	6 (22.2%)	8 (29.6%)	1.000	0 (0–1)	0.535	0 (0–1)
Age at first kidney stone detected, yrs
n	6	6	8
Mean (SD)	47.8 (11.9)	44.0 (16.1)	50.0 (17.5)	0.649	0.27 (−0.87 to 1.4)	0.799	−0.14 (−1.17 to 0.89)
Length of follow-up, yrs
n	20	25	24
Mean (SD)	5.7 (6.1)	5.6 (4.2)	4.5 (3.9)	0.967	0.01 (−0.59 to 0.61)	0.443	0.23 (−0.38 to 0.84)
eGFR at baseline, ml/min per 1.73 m²
n	24	27	27
Mean (SD)	70.7 (19.7)	56.4 (22.9)	56.5 (22.9)	0.021	0.67 (0.1–1.23)	0.022	0.67 (0.1–1.23)
eGFR at last follow-up, ml/min per 1.73 m²
n	21	25	24
Mean (SD)	66.9 (21.1)	49.2 (29.3)	39.0 (23.2)	0.025	0.7 (0.1–1.28)	< 0.001	1.26 (0.61–1.9)
CKD stage at last follow-up, n	20	25	24	0.046	0.34 (0–1)	0.017	0.41 (0–1)
Stage 1–2, n (%)	11 (55.0%)	8 (32.0%)	6 (25.0%)
Stage 3, n (%)	9 (45.0%)	9 (36.0%)	9 (37.5%)
Stage 4, n (%)	0 (0.0%)	7 (28.0%)	3 (12.5%)
Stage 5, n (%)	0 (0.0%)	1 (4.0%)	6 (25.0%)

Open in a new tab

ADPKD, autosomal dominant polycystic kidney disease; BMI, body mass index; CI, confidence interval; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; ES, effect size; UTI, urinary tract infection.

P-values are derived from t tests for continuous variables and chi-square tests for categorical variables.

Effect sizes (ES) are derived from Cohen’s d test for continuous variables and Cramer’s V test for categorical variables.

Adjusted BMI was derived from the adjusted weight divided by the height squared. The adjusted weight was calculated by subtracting the kidney weight from the body weight, assuming a kidney tissue density equal to that of water.

Clinical Features and eGFR Rate of Decline of ADPKD-IFT140

Kidney function and laboratory biomarkers are shown in Table 1 and Supplementary Table S1. Despite a similar mean age at baseline, patients with ADPKD-IFT140 had a significantly higher baseline eGFR (70.7 ± 19.7 ml/min per 1.73 m²) than those with ADPKD-PKD2 (56.4 ± 22.9, P = 0.021) and those with ADPKD-PKD1NT (56.5 ± 22.9, P = 0.022). After a similar average follow-up period, the eGFR at last follow-up remained significantly higher in patients with ADPKD-IFT140 (66.9 ± 21.1 ml/min per 1.73 m²) than in those with ADPKD-PKD2 (49.2 ± 29.3, P = 0.025) and in those with ADPKD-PKD1NT (39.0 ± 23.2, P < 0.001). None of the patients with ADPKD-IFT140 reached chronic kidney disease stages 4 and 5, whereas 32% of those with ADPKD-PKD2 and 37.5% of those with ADPKD-PKD1NT reached chronic kidney disease stages 4 and 5 (Table 1 and Supplementary Figure S3A). In addition, the eGFR rate of decline was lower in ADPKD-IFT140 (−0.69 ml/min per 1.73 m²/yr, 95% CI: [−1.09 to −0.30]) than in ADPKD-PKD1NT (−1.62, 95% CI: [−2.05 to −1.18], P = 0.006) and in ADPKD-PKD2 (−0.90, 95% CI: [−1.26 to −0.53], P = 0.737) (Figure 1).

Trajectory analysis of eGFR decline across ADPKD disease-causing variant groups. Trajectory analysis of eGFR decline for the disease-causing variant groups using mixed-model linear regression. (a) Individual patient trajectories of eGFR decline over time for each of the 3 disease-causing variant groups: ADPKD-*IFT140*, ADPKD-*PKD2*, and ADPKD-*PKD1NT*. Each line represents longitudinal eGFR measurements for a single patient, plotted against age. (b) The linear mixed-effects model used to summarize eGFR decline trajectories across the 3 groups. The model shows the predicted eGFR as a function of age, with the red, blue, and green lines corresponding to ADPKD-*IFT140*, ADPKD-*PKD2*, and ADPKD-*PKD1NT*, respectively. The shaded areas in the plot indicate regions with insufficient data, reflecting increased uncertainty in the eGFR trajectory at these ages. (c) The slope at the average age for each genotypic group is: −0.69, −0.90, and −1.62 ml/min per 1.73 m²/yr for ADPKD-*IFT140*, ADPKD-*PKD2*, and ADPKD-*PKD1NT*, respectively. ADPKD, autosomal dominant polycystic kidney disease; eGFR, estimated glomerular filtration rate.

Imaging Features of ADPKD-IFT140

Descriptive analysis of the imaging biomarkers is shown in Table 2. The distribution of Mayo imaging classification (MIC) is shown in Figure 2a and Supplementary Figure S3B. Patients with ADPKD-IFT140 were predominantly classified as MIC1A (33.3%) compared with MIC1C in ADPKD-PKD2 (44.4%) and MIC1B in ADPKD-PKD1NT (33.3%). In addition, 25.9% of patients with ADPKD-IFT140 were classified as MIC2A compared with 0% of those with ADPKD-PKD2 and 3.7% of those with ADPKD-PKD1NT. Although no significant difference was observed in height-adjusted TCV among the 3 groups, patients with ADPKD-IFT140 were characterized by a significantly fewer number of cysts (median [Q1–Q3]: 42 cysts [21–56]) than those with ADPKD-PKD2 (277 [119–528], P < 0.001), and those with ADPKD-PKD1NT (217 [146.5–460.5], P < 0.001) (Figure 2b–d). Patients with ADPKD-IFT140 exhibited significantly lower height-adjusted renal parenchymal volume, lower CPSA, and higher average cyst volume than those with PKD2 and those with PKD1NT (Table 2). In addition, the majority of those with ADPKD-IFT140 did not have any liver cysts (88.9%), whereas 44.4% of those with ADPKD-PKD2 and 63.0% of those with ADPKD-PKD1NT had > 100 liver cysts, respectively (Table 2 and Supplementary Figure S4). These results highlight that those patients with ADPKD-IFT140 develop enlarged kidneys with few enlarged cysts, and typically lack liver cysts. In Figure 3, we show 4 examples of MRIs with cyst rendering of ADPKD-IFT140 alongside matched ADPKD-PKD2 and ADPKD-PKD1NT images.

Table 2.

Descriptive analysis of the imaging biomarkers in the development cohort

Imaging biomarkers	ADPKD variants			IT140 vs. PKD2		IFT140 vs. PKD1NT
Imaging biomarkers	IFT140	PKD2	PKD1NT	P-value^a	ES^b (CI)	P-value^a	ES^b (CI)
n	27	27	27
Total kidney volume, ml, median (Q1–Q3)	744.8 (477.9–1404.8)	1279.7 (558.9–2475.3)	743.7 (528.0–1910.1)	0.048	−0.55 (−1.1 to 0)	0.191	−0.36 (−0.9 to 0.18)
Height-adjusted total kidney volume, ml/m, median (Q1–Q3)	424.8 (291.0–815.8)	763.5 (330.6–1399.4)	448.0 (322.5–1036.4)	0.051	−0.54 (−1.09 to 0.01)	0.223	−0.34 (−0.87 to 0.2)
Mayo Imaging Classification (MIC)				0.008	0.43 (0–1)	0.096	0.27 (0–1)
1A, n (%)	9 (33.3%)	7 (25.9%)	8 (29.6%)
1B, n (%)	8 (29.6%)	7 (25.9%)	9 (33.3%)
1C, n (%)	3 (11.1%)	12 (44.4%)	8 (29.6%)
1D, n (%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
1E, n (%)	0 (0.0%)	1 (3.7%)	1 (3.7%)
2A, n (%)	7 (25.9%)	0 (0.0%)	1 (3.7%)
Height-adjusted total cyst volume, ml/m, median (Q1–Q3)	218.4 (74.0–528.6)	380.6 (107.2–787.3)	124.3 (60.8–447.9)	0.234	−0.33 (−0.86 to 0.21)	0.962	−0.01 (−0.55 to 0.52)
Height-adjusted renal parenchymal volume, ml/m, median (Q1–Q3)	220.1 (177.0–295.8)	342.0 (227.8–571.0)	338.8 (250.2–585.6)	0.001	−0.94 (−1.51 to −0.34)	< 0.001	−1.08 (−1.67 to −0.47)
Cyst-parenchyma surface area, cm², median (Q1–Q3)	182.3 (107.7–369.7)	1138.1 (412.7–1870.7)	678.3 (370.2–1685.4)	< 0.001	−1.2 (−1.81 to −0.58)	< 0.001	−1.05 (−1.64 to −0.44)
Total cyst number, n, median (Q1–Q3)	42.0 (21.0–56.0)	277.0 (119.0–528.0)	217.0 (146.5–460.5)	< 0.001	−1.37 (−2.01 to −0.72)	< 0.001	−1.51 (−2.16 to −0.84)
Cyst index, median (Q1–Q3)	0.530 (0.292–0.671)	0.455 (0.310–0.612)	0.295 (0.175–0.478)	0.426	0.22 (−0.32 to 0.75)	0.017	0.67 (0.12–1.22)
Average cyst volume, ml/cyst, median (Q1–Q3)	12.1 (4.5–16.9)	2.1 (0.9–3.3)	1.0 (0.6–1.9)	0.004	0.82 (0.24–1.39)	0.005	0.8 (0.22–1.36)
Cumulative volume of the 2 largest cysts/TKV (Deep-learning),^c %, median (Q1–Q3)	29.0 (18.1–38.3)	10.4 (7.5–15.8)	5.3 (3.1–7.2)	< 0.001	1.21 (0.61–1.81)	< 0.001	1.58 (0.93–2.22)
Total number of liver cysts				< 0.001	0.75 (0.48–1)	< 0.001	0.73 (0.45–1)
0 liver cysts, n (%)	24 (88.9%)	3 (11.1%)	4 (14.8%)
1–50 liver cysts, n (%)	2 (7.4%)	11 (40.7%)	4 (14.8%)
51–100 liver cysts, n (%)	0 (0.0%)	1 (3.7%)	2 (7.4%)
> 100 liver cysts, n (%)	1 (3.7%)	12 (44.4%)	17 (63.0%)

Open in a new tab

ADPKD: autosomal dominant polycystic kidney disease; CI: confidence interval; ES: effect size; TKV: total kidney volume.

P-values are derived from t-tests for continuous variables and chi-squared tests for categorical variables.

Effect sizes (ES) are derived from Cohen’s d test for continuous variables and Cramer’s V test for categorical variables.

Cumulative volume of the 2 largest cysts/TKV (Deep-learning) represents the proportion of total kidney volume occupied by the 2 largest cysts. Both TKV and individual cyst volumes were measured using the deep-learning segmentation models. It is expressed as a percentage.

ADPKD-*IFT140* imaging phenotype and advanced imaging biomarker analysis. Advanced imaging biomarker analysis showing that patients with ADPKD-*IFT140* develop enlarged kidneys with fewer but significantly larger cysts than those with ADPKD-*PKD2*, and those with ADPKD-*PKD1NT*. In all plots, ADPKD-*IFT140* are shown in red, ADPKD-*PKD2* are shown in blue, and ADPKD-*PKD1NT* are shown in green. In plots (b) and (d), males are represented by a triangle whereas females are represented by a circle. (a) Plot of height-adjusted TKV (htTKV) versus age for patients with ADPKD-*IFT140* with typical and atypical Mayo imaging classification (MIC) differentiated compared with ADPKD-*PKD2* and ADPKD-*PKD1NT.* (b) Plot of total cyst number (TCN) versus total cyst volume (TCV) for patients with ADPKD-*IFT140* compared with ADPKD-*PKD2* and ADPKD-*PKD1NT* with males shown as triangles and females as circles. This plot shows that patients with ADPKD-*IFT140* have a smaller number of cysts compared with patients with ADPKD-*PKD2* and those with ADPKD-*PKD1NT* despite having similar TCV. (c) Box plot comparing TCN across the 3 different groups with ADPKD-IFT140 showing the lowest median TCN (42 cysts [21–56]) compared with ADPKD-*PKD2* (277 cysts [119 – 528]) and ADPKD-*PKD1NT* (217 cysts [146.5–460.5]). (d) Plot of TCN versus age at imaging, with patients with ADPKD-*IFT140* consistently showing fewer cysts across all age groups than in ADPKD-*PKD1NT* and ADPKD-*PKD2.* ADPKD, autosomal dominant polycystic kidney disease; htTKV, height-adjusted total kidney volume; MIC, Mayo imaging classification.

Granular cyst morphology via deep-learning segmentation on MRI in patients with ADPKD. Abdominal MRI with deep-learning cyst segmentation for 4 pairs of patients across ADPKD-*IFT140*, ADPKD-*PKD2*, and ADPKD-*PKD1NT* groups matched by age, sex, and htTKV. Each image shows segmented cysts in different colors overlaid on the kidney MRI. Patients with ADPKD-*IFT140* exhibit fewer, larger cysts; whereas ADPKD-*PKD2* and ADPKD-*PKD1NT* show a greater number of smaller cysts. ADPKD, autosomal dominant polycystic kidney disease; eGFR, estimated glomerular filtration rate; htTKV, height-adjusted total kidney volume; MIC, Mayo imaging classification; MRI, magnetic resonance imaging; N/A, not available; TCN, total cyst number.

Using volumes generated by deep-learning segmentation, the cumulative volume of the 2 largest cysts relative to TKV exhibited the highest AUC (0.822, 95% CI: [0.707–0.937]) for distinguishing ADPKD-IFT140 (Supplementary Figure S5). It was significantly higher in patients with ADPKD-IFT140 (median [Q1–Q3]: 29.0% (18.1%–38.3%)] than those with ADPKD-PKD2 (10.4% [7.5%–15.8%], P < 0.001) and those with ADPKD-PKD1NT (5.3% [3.1%–7.2%], P < 0.001) (Table 2, Supplementary Figure 6a and b). A strong correlation was observed between the cumulative volume of the 2 largest cysts relative to TKV obtained via deep-learning segmentation and those obtained manually (R = 0.936, P < 0.001) (Supplemental Figure S7A). Bland-Altman analysis revealed an overestimation of the volumes with the manual method compared with the deep-learning segmentation method, with a mean percentage bias of 26.3% (95% CI: −27.2% to 79.9%) (Supplementary Figure 7B).

Distinguishing ADPKD-IFT140 Phenotype Using Advanced Imaging Biomarkers

Advanced imaging biomarkers showed strong predictive power for distinguishing ADPKD-IFT140 (Table 3) with TCN, average cyst volume, and CPSA exhibiting an AUC > 0.8. TCN had the highest AUC (0.907, 95% CI: [0.840–0.974]), where having < 65 cysts achieved a sensitivity of 81.5% and a specificity of 90.7% (Figure 4a). In addition, the cumulative volume of the 2 largest cysts relative to TKV obtained via the deep-learning segmentation model showed an AUC of 0.822 (95% CI: 0.707–0.937) (Table 3, Figure 4a, and Supplementary Figure S8), whereas the manually measured cumulative volume of the 2 largest cysts relative to TKV showed an AUC of 0.814 (95% CI: 0.698–0.930). A value ≥ a threshold of 15.8% in the deep-learning approach and 18.6% in the manual approach had a sensitivity of 81.5% and 84.6%, respectively; and a specificity of 83.3% and 76.5%, respectively (Figure 4a and c, and Supplementary Figure S8). A multiple logistic regression model including CPSA and cystic index performed very well at distinguishing ADPKD-IFT140 with an AUC of 0.949 (95% CI: 0.901–0.991). A predicted probability ≥ 0.234 had a sensitivity of 92.6% and specificity of 81.5% (Table 3, Table 4, and Figure 4a and d). An alternative multivariable model including TCN and cystic index performed very well at distinguishing ADPKD-IFT140 with an AUC of 0.962 (95% CI: 0.917–0.992) (Supplementary Tables S2 and S3). In Figure 5, we present a step-by-step visualization of our approach, demonstrating how these models can be applied in clinical practice.

Table 3.

Discriminatory ability of advanced diagnostic imaging parameters for differentiating ADPKD-IFT140 from PKD1NT and PKD2 variants

Variables	n	P-value^a	Sensitivity	Specificity	Cutoff point^b	AUC (95% CI)
Multivariable model^c	81		92.6%	81.5%	0.234^d	0.949 (0.901–0.991)
Total cyst number	81	<0.001	81.5%	90.7%	65^e	0.907 (0.840–0.974)
Average cyst volume	81	<0.001	74.1%	92.6%	5.1^d	0.839 (0.728–0.949)
Cumulative volume of the 2 largest cysts/TKV (Deep-learning)^f	81	<0.001	81.5%	83.3%	15.8%^d	0.822 (0.707–0.937)
Cyst-parenchymal surface area	81	0.002	85.2%	70.4%	450.1^e	0.820 (0.728–0.911)
Cumulative volume of the 2 largest cysts/TKV (Manual)^g	77	<0.001	84.6%	76.5%	18.6%^d	0.814 (0.698–0.930)
Height-adjusted renal parenchymal volume	81	0.002	92.6%	53.7%	327.5^e	0.767 (0.665–0.869)
Renal parenchymal volume	81	0.002	92.6%	55.6%	550.5^e	0.766 (0.664–0.868)
Cyst index	81	0.062	66.7%	64.8%	0.477^d	0.634 (0.493–0.776)
Height-adjusted total kidney volume	81	0.093	92.6%	33.3%	1137.9^e	0.595 (0.468–0.723)
Total kidney volume	81	0.085	88.9%	35.2%	1836.6^e	0.595 (0.468–0.722)
Total cyst volume	81	0.424	88.9%	24.1%	1241.7^e	0.519 (0.387–0.651)
Height-adjusted total cyst volume	81	0.474	88.9%	25.9%	699.5^e	0.514 (0.382–0.647)

Open in a new tab

AUC, area under the curve; CI, confidence interval; TKV, total kidney volume.

P-value was based on Wald-test statistics of estimated biomarker coefficient.

Cutoff points were determined based on the threshold that maximized the difference between sensitivity and (1 − specificity), optimizing the tradeoff between correctly identifying ADPKD-IFT140 cases while minimizing misclassification of PKD1NT and PKD2.

The first row displays the multivariable logistic regression model CPSA and cystic index with estimates shown in Table 4. All subsequent rows represent univariate associations for each imaging biomarker.

A value above or equal to the threshold is characteristic of IFT140 disease-causing variant.

A value below the threshold is characteristic of IFT140 disease-causing variant.

Cumulative volume of the 2 largest cysts/TKV (Deep-Learning) represents the proportion of total kidney volume occupied by the 2 largest cysts. Both TKV and individual cyst volumes were measured using the deep-learning segmentation models. It is expressed as a percentage.

Cumulative volume of the 2 largest cysts/TKV (manual) represents the proportion of total kidney volume occupied by the 2 largest cysts. Observed cyst diameters were manually measured and calculated individual cyst volumes were then calculated using the formula $V = \frac{π}{6} d^{3}$ , with the assumption that each cyst is a sphere, and that the diameter d is the length of the cyst measured on coronal section. It is expressed as a percentage.

Predictive model performance in the development cohort. Advanced imaging biomarkers show strong predictive power in distinguishing ADPKD-*IFT140* disease-causing variant in the development cohort. The development cohort (n = 81) consisted of included ADPKD-*IFT140*, ADPKD-*PKD2*, and ADPKD-*PKD1NT* matched based on sex, age (± 3 years) and height-adjusted TKV. (a) Receiver operating characteristics curves for various biomarkers, showing their predictive ability in distinguishing ADPKD-*IFT140* disease-causing variants. The multivariable model that includes CPSA and cystic index shows the highest AUC of 0.949. The cumulative volume of the 2 largest cysts relative to TKV obtained via manual measurements achieved an AUC of 0.814. (b) Correlation matrix showing the correlation between the different advanced imaging biomarkers. (c) and (d) demonstrate the classification of ADPKD genotypes using different predictive models based on imaging biomarkers. Blue dots represent correctly classified patients, whereas red dots indicate misclassified patients. (c) shows the distribution of the manually measured cumulative volume of the 2 largest cysts relative to TKV (calculated from cyst diameters using the assumption that cysts are spheres) across ADPKD variants, with the vertical line indicating the threshold of 18.6%. This model achieved a sensitivity of 84.6% and a specificity of 76.5% for distinguishing ADPKD-*IFT140* from the other variants. (d) shows the distribution of the multivariable model, including CPSA and cystic index across ADPKD variants, with the vertical line indicating the threshold of 0.234. This model achieved a sensitivity of 92.6% and a specificity of 81.5% for distinguishing ADPKD-*IFT140* from the other variants. ∗Multivariable model, including CPSA and cystic index. ∗∗Cumulative volume of the 2 largest cysts relative to TKV obtained via deep-learning segmentation. ADPKD, autosomal dominant polycystic kidney disease; AUC, area under the curve; CPSA, cyst-parenchymal surface area; tTKV, height-adjusted total kidney volume; htRPV, height-adjusted renal parenchymal volume; htTCV, height-adjusted total cyst volume; TCN, total cyst number; TKV, total kidney volume.

Table 4.

Multivariable logistic regression model of imaging predictors associated with ADPKD-IFT140 vs PKD1NT and PKD2 variants

Variables	Estimate	Standard error	P-value^a
Intercept	−1.376	0.688	0.045
Cyst-parenchymal surface area	−0.009	0.003	<0.001
Cyst index	12.861	3.423	<0.001

Open in a new tab

The predicted probability of ADPKD-IFT140 disease-causing variant can be calculated using the following formula: $p r o b a b i l i t y (I F T 140) = \frac{1}{1 + e^{- (- 1.376 - 0.009 x C P S A + 12.861 x C y s t i c i n d e x)}}$

P-value was based on Wald-test statistics of estimated biomarker coefficient.

Application of advanced imaging biomarkers for characterizing ADPKD-*IFT140*. The figure illustrates 2 complementary approaches for distinguishing ADPKD-*IFT140* from PKD variants, using imaging biomarkers. (a) The deep-learning segmentation approach (left) utilizes cyst-parenchyma surface area (CPSA) and cystic index. CPSA, which quantifies the total interface between cysts and renal parenchyma, is lower in ADPKD-*IFT140* because of the exophytic nature of its cysts. The cystic index, which measures the proportion of TCV relative to TKV, is higher in ADPKD-*IFT140*, reflecting the predominance of a few disproportionately large cysts. The predictive model incorporates these 2 biomarkers into a logistic regression formula: $\frac{1}{1 + e^{- (- 1.376 - 0.009 x C P S A + 12.861 x C y s t i c i n d e x)}}$ , with a threshold of ≥ 0.234 predicting ADPKD-*IFT140*, achieving a sensitivity of 92.6% and specificity of 81.5% (AUC = 0.949). For instance, if CPSA is 182.3 cm² and the cystic index is 0.53, the probability calculation would be $\frac{1}{1 + e^{- (- 1.376 - 0.009 x 182.3 + 12.861 x 0.53)}}$ ) = 0.978 which is higher than the threshold 0.234, indicating a higher likelihood of ADPKD-*IFT140* variant. (b) The manual measurement approach (right) provides a more accessible alternative by calculating the cumulative volume of the 2 largest cysts relative to TKV, with a threshold of ≥ 18.6% predicting ADPKD-*IFT140*, achieving a sensitivity of 84.6% and specificity of 76.5% (AUC = 0.816). For example, if the 2 largest cysts have diameters of 6 cm and 5 cm, their cumulative volume is calculated as $V = \frac{π}{6} (d_{1}^{3} + d_{2}^{3})$ $= \frac{π}{6} (6^{3} + 5^{3})$ = 178.55 ml. Given a calculated TKV_ellipsoid of 800 ml, the resulting ratio is 178.55 / 800 = 22.3%, which is higher than the threshold 18.6%, indicating a higher likelihood of ADPKD-*IFT140* variant. CPSA, cyst-parenchymal surface area; d, diameter; PKD, polycystic kidney disease; TCV, total cyst volume; TKV: total kidney volume.

Specificity Analysis Using an Internal Cohort

The internal cohort included patients with ADPKD-PKD2 (n = 97, 17%), those with ADPKD-PKD1NT (n = 170, 29.8%), and those with ADPKD-PKD1T (n = 302, 53.0%). Among this internal cohort, 34.1% were males with a mean age at imaging of 41.1 (± 13.2) years and a median htTKV of 694.0 (442.5–1208.8) ml/m. Baseline demographics, and advanced imaging biomarkers are shown in Table 5, Table 6, respectively. Applying the thresholds established in the development cohort, we conducted an analysis to determine the specificity of key biomarkers (Figures 6a and b, and Supplementary Table S4). At a threshold of 15.8%, cumulative volume of the 2 largest cysts relative to TKV obtained via deep-learning segmentation had a specificity of 85.9% at distinguishing ADPKD-IFT140 (Figure 6a and c), and a specificity of 74.3%, 84.1%, and 90.1% when differentiating IFT140 from PKD2, PKD1NT, and PKD1T variants, respectively. At a threshold of 0.234, the established multivariable model, including CPSA and cystic index, had a specificity of 88.9% when distinguishing IFT140 (Figure 6b and d), and a specificity of 73.2%, 91.8%, and 92.4% when differentiating IFT140 from PKD2, PKD1NT, and PKD1T, respectively. The specificity of additional biomarkers is shown in Supplementary Table S4.

Table 5.

Baseline demographics of the internal specificity cohort

Clinical characteristics	ADPKD variants				IFT140 vs. PKD2		IFT140 vs. PKD1NT		IFT140 vs. PKD1T
Clinical characteristics	IFT140	PKD2	PKD1NT	PKD1T	P-value^a	ES^b (CI)	P-value^a	ES^b (CI)	P-value	ES (CI)
n	27	97	170	302
Age at imaging, mean (SD)	58.0 (13.7)	46.6 (14.1)	42.2 (12.5)	38.6 (12.6)	< 0.001	0.82 (0.36–1.27)	< 0.001	1.2 (0.69–1.71)	< 0.001	1.47 (0.92–2.02)
Sex					0.402	0 (0–1)	0.109	0.09 (0–1)	0.124	0.06 (0–1)
Male, n (%)	13 (48.1%)	38 (39.2%)	55 (32.4%)	101 (33.4%)
Female, n (%)	14 (51.9%)	59 (60.8%)	115 (67.6%)	201 (66.6%)
Race					0.002	0.29 (0.07–1)	< 0.001	0.3 (0.16–1)	< 0.001	0.31 (0.21–1)
White, n (%)	23 (85.2%)	86 (88.7%)	156 (91.8%)	274 (90.7%)
Non-White, n (%)	1 (3.7%)	11 (11.3%)	14 (8.2%)	28 (9.3%)
Unknown, n (%)	3 (11.1%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Known FH of ADPKD, n (%)	13 (48.1%)	68 (70.1%)	115 (67.6%)	211 (69.9%)	0.034	0.17 (0–1)	0.048	0.12 (0–1)	0.020	0.12 (0–1)
eGFR at baseline, ml/min per 1.73 m²
n	24	97	165	293
Mean (SD)	70.7 (19.7)	76.1 (30.2)	75.0 (28.8)	71.9 (33.3)	0.409	−0.21 (−0.6 to 0.18)	0.481	−0.17 (−0.54 to 0.19)	0.860	−0.04 (−0.36 to 0.28)
Hypertension, n (%)	11 (40.7%)	78 (81.2%)	143 (85.6%)	242 (81.5%)	< 0.001	0.37 (0.21–1)	< 0.001	0.38 (0.26–1)	< 0.001	0.27 (0.17–1)
Age at hypertension diagnosis, yrs
n	11	76	140	231
Mean (SD)	55.2 (6.6)	41.7 (12.1)	36.6 (12.6)	32.4 (10.8)	< 0.001	1.38 (0.74–2.01)	< 0.001	1.84 (1.07–2.59)	< 0.001	2.55 (1.46–3.62)
Adjusted BMI^c at baseline, kg/m²
n	24	94	160	282
Mean (SD)	26.9 (4.5)	27.5 (5.5)	26.4 (5.0)	26.3 (7.2)	0.623	−0.12 (−0.54 to 0.31)	0.645	0.1 (−0.31 to 0.52)	0.715	0.09 (−0.24 to 0.42)

Open in a new tab

ADPKD, autosomal dominant polycystic kidney disease; BMI, body mass index; CI, confidence interval; eGFR, estimated glomerular filtration rate; ES: effect size; FH, family history.

P-values are derived from t-tests for continuous variables and chi-squared tests for categorical variables.

Effect sizes (ES) are derived from Cohen’s d test for continuous variables and Cramer’s V test for categorical variables.

Table 6.

Descriptive analysis of the imaging biomarkers in the internal cohort

	ADPKD variants				IFT140 vs. PKD2		IFT140 vs. PKD1NT		IFT140 vs. PKD1T
	IFT140	PKD2	PKD1NT	PKD1T	P-value^a	ES^b (CI)	P-value^a	ES^b (CI)	P-value	ES (CI)
n	27	97	170	302
Age at imaging, mean (SD)	58.0 (13.7)	46.6 (14.1)	42.2 (12.5)	38.6 (12.6)	< 0.001	0.82 (0.36–1.27)	< 0.001	1.2 (0.69–1.71)	< 0.001	1.47 (0.92–2.02)
Total kidney volume, ml, median (Q1–Q3)	744.8 (477.9–1404.8)	1010.7 (550.6–2333.8)	1095.1 (711.2–1902.8)	1299.7 (826.4–2090.8)	0.050	−0.51 (−0.85 to −0.17)	0.058	−0.46 (−0.8 to −0.12)	0.008	−0.61 (−0.96 to −0.26)
Height-adjusted total kidney volume, ml/m, median (Q1–Q3)	424.8 (291.0–815.8)	581.2 (358.5–1381.7)	632.9 (428.9–1135.8)	759.9 (477.6–1230.1)	0.044	−0.52 (−0.86 to −0.17)	0.056	−0.46 (−0.8 to −0.12)	0.007	−0.62 (−0.98 to −0.26)
Height-adjusted total cyst volume, ml/m, median (Q1–Q3)	218.4 (74.0–528.6)	267.0 (86.9–753.2)	221.9 (108.2–540.4)	326.6 (142.1–603.7)	0.245	−0.29 (−0.65 to 0.07)	0.727	−0.08 (−0.44 to 0.29)	0.425	−0.16 (−0.55 to 0.23)
Height-adjusted renal parenchymal volume, ml/m, median (Q1–Q3)	220.1 (177.0–295.8)	348.3 (259.5–535.8)	393.1 (293.5–564.9)	425.2 (312.1–622.3)	< 0.001	−1.01 (−1.34 to −0.69)	< 0.001	−1.29 (−1.57 to −1)	< 0.001	−1.31 (−1.56 to −1.06)
Cyst-parenchyma surface area, cm², median (Q1–Q3)	182.3 (107.7–369.7)	744.4 (329.4–1541.5)	937.1 (488.3–1716.3)	1194.3 (650.7–2125.5)	0.001	−0.90 (−1.22 to −0.59)	< 0.001	−1.23 (−1.5 to −0.95)	< 0.001	−1.38 (−1.62 to −1.14)
Total cyst number, n, median (Q1–Q3)	42.0 (21.0–56.0)	221.0 (124.0–420.0)	324.0 (191.2–517.0)	374.5 (218.0–598.2)	< 0.001	−1.42 (−1.77 to −1.07)	< 0.001	−1.7 (−2 to −1.4)	< 0.001	−1.6 (−1.83 to −1.38)
Cyst index, median (Q1–Q3)	0.530 (0.292–0.671)	0.426 (0.242–0.558)	0.334 (0.239–0.487)	0.411 (0.289–0.493)	0.108	0.33 (−0.12 to 0.79)	0.002	0.58 (0.11–1.05)	0.009	0.43 (−0.05 to 0.89)
Average cyst volume, ml/cyst, median (Q1–Q3)	12.1 (4.5–16.9)	1.9 (0.9–4.0)	1.2 (0.7–2.1)	1.4 (0.8–2.0)	< 0.001	0.8 (0.23–1.37)	< 0.001	0.88 (0.29–1.45)	< 0.001	0.89 (0.3–1.47)
Cumulative volume of the 2 largest cysts/TKV, %, median (Q1–Q3)	29.0 (18.1–38.3)	9.8 (6.2–15.2)	5.7 (3.1–10.4)	6.4 (4.1–9.8)	< 0.001	1.26 (0.66–1.84)	< 0.001	1.55 (0.9–2.19)	< 0.001	1.63 (0.94–2.29)

Open in a new tab

ADPKD, autosomal dominant polycystic kidney disease; CI, confidence interval; ES, effect size; htTKV, height-adjusted TKV; TKV, total kidney volume.

P-values are derived from t tests for continuous variables and chi-squared tests for categorical variables.

Effect sizes (ES) are derived from Cohen’s d test for continuous variables and Cramer’s V test for categorical variables.

Validation of the predictive models in an internal specificity and an external sensitivity cohort. Advanced imaging biomarker specificity analysis in an internal validation cohort and sensitivity analysis in an external validation cohort. The internal cohort consisted of all the unmatched patients with ADPKD-*PKD1NT* and those with ADPKD-*PKD2* in addition to a cohort of patients with ADPKD-*PKD1T* that fit the previously established inclusion criteria. The external cohort consisted of patients with ADPKD who had an MRI available and had a confirmed monoallelic LoF variant in *IFT140* from a cohort of 1359 individuals recruited in the Genkyst cohort, and in different French centers. (a) and (b) include the same patients with ADPKD-*IFT140* from the development cohort. They demonstrate the classification of ADPKD genotypes, using different predictive models based on imaging biomarkers. Blue dots represent correctly classified patients, whereas red dots indicate misclassified patients. (a) and (c) show the distribution of the cumulative volume of the 2 largest cysts relative to TKV derived from the deep-learning segmentation model across ADPKD genotypes, with the vertical line indicating the threshold of 15.8%. This variable achieved a specificity 85.9% for distinguishing ADPKD-*IFT140* from patients with ADPKD and identified disease-causing variants in *PKD1* or *PKD2* variants. (b) and (d) show the distribution of the multivariable model, including CPSA and cystic index across ADPKD genotypes, with the vertical line indicating the threshold of 0.234. This model achieved a specificity 88.9% for distinguishing ADPKD-*IFT140* from patients with ADPKD and identified disease-causing variants in *PKD1* or *PKD2* variants. (e) shows the distribution of the performance manually measured cumulative volume of the largest cysts relative to TKV in distinguishing an external ADPKD-*IFT140* cohort with the horizontal line indicating the threshold of 18.6%. This model achieved a sensitivity of 72.2% for distinguishing patients with ADPKD-*IFT140*. Males are represented as a triangle and females as a circle. ∗Multivariable model including CPSA and cystic index. ADPKD, autosomal dominant polycystic kidney disease; CPSA, cyst-parenchymal surface area; LoF, loss of function; MRI, magnetic resonance imaging.

Sensitivity Analysis Using an External Cohort

The external cohort consisted of 36 patients with ADPKD-IFT140, 50% of whom were male, with a mean imaging age of 55.8 (± 14.6) years and a median htTKV of 458.2 (278.6–865.3) ml/m. The majority (89.0%) were classified as MIC2A, and 86.1% did not have any liver cysts. Baseline demographics and imaging biomarkers are shown in Supplementary Tables S5 and S6, respectively. To validate the sensitivity of our practical model, we assessed the manually calculated cumulative volume of the 2 largest cysts relative to TKV using the previously established threshold of 18.6%. This threshold accurately classified 26 of 36 patients as ADPKD-IFT140 across all ages, yielding a sensitivity of 72.2% (Figure 6e).

Discussion

This study provides the most comprehensive characterization of the ADPKD-IFT140 phenotype to date, highlighting its distinct clinical and imaging features. We found that patients with ADPKD-IFT140 develop enlarged kidneys with significantly lower CPSA, fewer but disproportionally larger cyst volumes, and minimal liver cystic involvement than those with ADPKD-PKD2, and those with ADPKD-PKD1NT. To further delineate the ADPKD-IFT140 phenotype, we applied advanced imaging biomarkers and developed models that effectively characterized ADPKD-IFT140, achieving high differentiation accuracy (AUC = 0.81–0.95), with validation in both internal and external cohorts.

MIC classification, a well-established predictor of future renal function decline in ADPKD,³¹ is less reliable in patients with ADPKD-IFT140 because of atypical cyst morphology and preserved parenchyma.¹⁸^,²⁰^,³⁴^,³⁵ Although atypical features were defined in MIC as class 2A, not all patients with ADPKD-IFT140 fit the lop-sided description (< 5 cysts account for > 50% TKV) and we believe that this atypical feature seen in ADPKD-IFT140 is a continuum. To address these limitations, we introduced advanced imaging biomarkers that better reflect the structural and morphological nuances of ADPKD-IFT140. In addition, to ensure a more precise characterization of these features, we manually matched patients in the development cohort by age, sex, and the closest htTKV, allowing us to control for kidney size and focus on more subtle morphological differences that would not be obscured by large discrepancies in kidney volume. CPSA, a biomarker that accounts for cyst volume, number, and exophytic distribution, is significantly lower in ADPKD-IFT140, suggesting a more exophytic cyst pattern and greater parenchymal preservation. Meanwhile, the cystic index, which measures the proportion of cyst volume relative to TKV, remains high despite a reduced TCN, underlining the presence of a few disproportionately large cysts that substantially contribute to TKV. By capturing these unique structural features, these biomarkers could offer future prognostic value for patients with ADPKD-IFT140, potentially refining risk stratification in patients with borderline or uncertain disease trajectories.

By providing advanced imaging data that was previously lacking, our findings confirm that ADPKD-IFT140 presents a distinct renal phenotype characterized by enlarged kidneys due to a few large cysts, preserved renal parenchyma, and asymmetry, with 26% and 88.9% of patients with ADPKD-IFT140 in the development and external cohorts having an atypical MIC2A, respectively. In addition, patients with ADPKD-IFT140 showed significantly lower renal parenchymal volume than both those with ADPKD-PKD2 and those with ADPKD-PKD1NT despite having a better clinical course. These findings can be explained by the presence of microcysts in patients with ADPKD-PKD1NT and those with ADPKD-PKD2, likely undetected because of MRI resolution limits, thus not captured by the artificial intelligence algorithm.³⁰ Furthermore, ADPKD-IFT140 were found to have minimal liver cystic involvement, with only 1 patient having > 100 liver cysts in our cohort. Our findings align with the previous literature classifying monoallelic LoF IFT140 variants within the ADPKD spectrum.⁷^,¹⁸^,²⁰^,³⁴ A multicohort study by Senum et al.¹⁸ was the first to identify monoallelic LoF IFT140 variants in families diagnosed with ADPKD without previous genetic testing or lacking PKD1 or PKD2 variants, characterizing ADPKD-IFT140 as a milder phenotype with a few large cysts, increased htTKV, minimal liver involvement, and limited progression to kidney failure.¹⁸

Our study observed that patients with ADPKD-IFT140 were significantly less likely to have a family history of ADPKD, consistent with Fujimaru et al.,³⁶ who reported higher prevalence of monoallelic LoF IFT140 variants in sporadic cases (4.5%) than in the general ADPKD population (2%).⁷^,¹⁸ In addition, we observed that patients with ADPKD-IFT140 were significantly less frequently hypertensive, with a later age of onset than those with ADPKD-PKD2 and those with ADPKD-PKD1NT. These findings align with Senum et al.¹⁸ and Zagorec et al.,²² who respectively reported hypertension in 66.1% and 50.7% in patients with ADPKD-IFT140, with an average age of onset in the late 50s.¹⁸^,²² In addition, none of our patients with ADPKD-IFT140 progressed to chronic kidney disease stages 4 and 5, unlike 32% of those with ADPKD-PKD2 and 37.5% of those with ADPKD-PKD1NT. Annual eGFR rate of decline was also slower in patients with ADPKD-IFT140 (−0.69 ml/min per 1.73 m²/yr) than in those with ADPKD-PKD1NT (−1.62) and in those with ADPKD-PKD2 (−0.90), consistent with Senum et al.¹⁸ and Zagorec et al.,²² where only 1 patient in each study reached kidney failure.¹⁸^,²² Interestingly, in our cohort, a higher proportion of patients with ADPKD-IFT140 underwent cyst volume reduction procedures, which may reflect their larger cyst morphology.

Our study proposes 2 models for characterizing ADPKD-IFT140. The first multivariable model, which estimates that a lower CPSA and higher cystic index, effectively characterizes ADPKD-IFT140 (AUC = 0.949, sensitivity: 92.6%, specificity: 81.5%). This model was chosen over the alternative (including TCN and cystic index) because of the greater clinical relevance of CPSA, which better captures the exophytic nature and volume burden of ADPKD-IFT140 cysts. Recognizing that deep-learning segmentation may not be widely accessible, we developed a practical manual model based on the cumulative volume of the 2 largest cysts relative to TKV (AUC = 0.814, sensitivity: 84.6%, specificity: 76.5%), an easily measurable biomarker that highlights the disproportionally large cysts in ADPKD-IFT140. This model involves calculating the cumulative volume of the 2 biggest cysts using the formula $V = \frac{π}{6} (d_{1}^{3} + d_{2}^{3})$ , where d₁ and d₂ are the diameters of the 2 biggest cysts measured from the coronal section in an MRI or computed tomography scan. TKV can be estimated using the ellipsoid equation.²^,³¹ For example, a patient whose cumulative volume of the 2 largest cysts constitutes ≥ 18.6% of their TKV is more likely to carry an IFT140 variant, with a sensitivity of 84.6% and specificity of 76.5%. Although Bland-Altman analysis revealed a systematic overestimation with manual measurements compared with deep-learning segmentation, attributable to the spherical assumption of manual methods, this model remains highly practical. It offers ease of use in clinical practice without requiring advanced imaging software. These predictive models were validated in both the internal and the external cohorts. The multivariable model demonstrated high specificity (88.9%) in the internal cohort, whereas the practical model achieved high sensitivity (72.2%) in the external cohort. The lower sensitivity observed in the external validation likely reflects variations in cyst morphology, kidney volume across populations, and operator-dependent variability in measuring cyst diameters.

This study presents several strengths. It provides the most comprehensive characterization of the ADPKD-IFT140 phenotype to date. In addition, we developed both a practical and an advanced deep-learning segmentation model that granularly characterizes the cystic morphology of ADPKD-IFT140. However, our study has limitations. We manually matched patients in the development cohort by age, sex, and the closest htTKV, optimizing use of our limited ADPKD-IFT140 population without excluding valuable cases. Yet, significant differences in MIC remained between ADPKD-IFT140 and ADPKD-PKD2, likely because of larger kidney volumes in ADPKD-PKD2; this difference was not observed with ADPKD-PKD1NT, possibly because of the higher prevalence of PKD1NT variants.⁶^,³⁷ ADPKD-PKD1NT was included in the development cohort because of its milder disease course,4, 5, 6 aligning more closely with ADPKD-IFT140. Notably, PKD1T variants—known for their more severe presentation—were excluded from the development cohort to focus on PKD1NT and PKD2, whose phenotypes are closer to ADPKD-IFT140. However, our internal specificity analysis did include PKD1T, where applying the same cutoffs established in the development cohort yielded a high specificity of 92.8%. This indicates that, though PKD1T was not part of the training cohort, the model is of value beyond the PKD2/PKD1NT groups; it can serve as a broadly applicable imaging-based tool that can supplement genetic testing. However, the inclusion of milder PKD1NT and PKD2 phenotypes, matched at a mean age of 57 years, may have underestimated the severity differences between IFT140 and the other variants. Clinical presentation bias is also possible, because symptomatic IFT140 patients are more likely to seek medical attention, unlike those with minimal cyst growth. In addition, the development cohort was limited in size and enriched to maximize phenotypic contrast with ADPKD-IFT140; we addressed this by validating our models in the following 2 distinct cohorts: internal specificity and external sensitivity cohorts. Nonetheless, we recognize that these validation datasets may not fully reflect the full spectrum of real-world clinical variability.

Future research could further characterize atypical and minor ADPKD variants, including GANAB and DNAJB11. Validating our models in larger cohorts, including patients with other minor variants, could enhance their performance in distinguishing IFT140 across the full spectrum of atypical ADPKD-related variants. Applying these models in a prospective cohort of patients with ADPKD without previous genetic testing or detected disease-causing variants followed by confirmatory genetic testing could further assess their utility. In addition, future prospective studies with larger cohorts are needed to evaluate whether advanced imaging biomarkers such as CPSA and cystic index can predict renal outcomes over time, further clarifying how they contribute to the prognosis of ADPKD-IFT140.

In conclusion, we describe a distinct phenotype with enlarged kidneys, fewer enlarged kidney cysts, and minimal liver involvement in patients with ADPKD-IFT140. We developed practical and advanced deep-learning segmentation models that are effective in distinguishing patients with ADPKD-IFT140.

Disclosure

NKD reports consulting fees from Vertex, Regulus Therapeutics, and Natera. She also holds leadership roles as Associate Editor of Kidney360 and serves on the PKD Foundation Scientific Advisory Board. VET reports grants and/or consulting fees from Palladio Biosciences, Mironid, Sanofi Genzyme, GSK, Reata, and Regulus Therapeutics. He also serves on the advisory boards of the PKD Foundation and the editorial board of the American Society of Nephrology. PCH discloses grants and consulting fees from Espervita, Navitor, Acceleron, and Regulus. ECLG reports consulting fees from Rhythm Pharmaceuticals, GSK, and Vertex. She is vice-chair of the Genes & Kidney working group of the ERA, cochair of the ClinGen Ciliopathies and Cystic Kidney Diseases group. FC reports consulting fees and/or research funding from Otsuka Pharmaceuticals, Regulus, and Natera. He also serves as a member of the Board of Directors and past co-chair of Center of Excellence advisory committee for the PKD Foundation. All the other authors declared no competing interests.

Acknowledgments

We extend our heartfelt gratitude to all our patients with autosomal dominant polycystic kidney disease (ADPKD), whose daily experiences inspire us to deepen our understanding of the clinical history and distinguishing features of ADPKD. Their stories motivate us to tailor individualized medical care more effectively.

Footnotes

Supplementary File (PDF)

Supplemental Statistical Analysis.

Figure S1. Study flow chart and development cohort selection.

Figure S2. Prevalence of events affecting kidney volume across disease-causing variants.

Figure S3. Prevalence of chronic kidney disease stages and Mayo imaging classification across disease-causing variants.

Figure S4. Prevalence of liver cysts across the 3 disease-causing variant groups.

Figure 5. Predictive performance of the cumulative volume of the biggest cysts relative to TKV.

Figure S6. Comparative analysis of the cumulative volume of the 2 largest cysts relative to total kidney volume (TKV) obtained via deep-learning segmentation.

Figure S7. Correlation and Bland-Altman analysis plots.

Figure S8. Performance of the cumulative volume the 2 biggest cysts relative to TKV obtained via deep-learning in the development cohort.

Table S1. Descriptive analysis of the laboratory biomarkers in the development cohort.

Table S2. Logistic regression classification of ADPKD-IFT140 versus ADPKD-PKD1NT/PKD2 variants using advanced imaging biomarkers.

Table S3. Estimates of multivariable logistic regression classification of IFT140 versus ADPKD-PKD1NT/PKD2 variants.

Table S4. Specificity analysis of the advanced imaging biomarkers in the internal cohort.

Table S5. Baseline demographics of the external cohort.

Table S6. Descriptive analysis of the imaging biomarkers in the external cohort.

Supplementary Material

Supplementary File (PDF)

Supplemental Statistical Analysis. Figure S1. Study flow chart and development cohort selection. Figure S2. Prevalence of events affecting kidney volume across disease-causing variants. Figure S3. Prevalence of chronic kidney disease stages and Mayo imaging classification across disease-causing variants. Figure S4. Prevalence of liver cysts across the 3 disease-causing variant groups. Figure 5. Predictive performance of the cumulative volume of the biggest cysts relative to TKV. Figure S6. Comparative analysis of the cumulative volume of the 2 largest cysts relative to total kidney volume (TKV) obtained via deep-learning segmentation. Figure S7. Correlation and Bland-Altman analysis plots. Figure S8. Performance of the cumulative volume the 2 biggest cysts relative to TKV obtained via deep-learning in the development cohort. Table S1. Descriptive analysis of the laboratory biomarkers in the development cohort. Table S2. Logistic regression classification of ADPKD-IFT140 versus ADPKD-PKD1NT/PKD2 variants using advanced imaging biomarkers. Table S3. Estimates of multivariable logistic regression classification of IFT140 versus ADPKD-PKD1NT/PKD2 variants. Table S4. Specificity analysis of the advanced imaging biomarkers in the internal cohort. Table S5. Baseline demographics of the external cohort. Table S6. Descriptive analysis of the imaging biomarkers in the external cohort.

mmc1.pdf^{(996KB, pdf)}

References

1.Chebib F.T., Torres V.E. Autosomal dominant polycystic kidney disease: core curriculum 2016. Am J Kidney Dis. 2016;67:792–810. doi: 10.1053/j.ajkd.2015.07.037. [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Chebib F.T., Torres V.E. Recent advances in the management of autosomal dominant polycystic kidney disease. Clin J Am Soc Nephrol. 2018;13:1765–1776. doi: 10.2215/CJN.03960318. [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Cornec-Le Gall E., Alam A., Perrone R.D. Autosomal dominant polycystic kidney disease. Lancet. 2019;393:919–935. doi: 10.1016/S0140-6736(18)32782-X. [DOI] [PubMed] [Google Scholar]
4.Cornec-Le Gall E., Audrézet M.P., Chen J.M., et al. Type of PKD1 mutation influences renal outcome in ADPKD. J Am Soc Nephrol. 2013;24:1006–1013. doi: 10.1681/ASN.2012070650. [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Heyer C.M., Sundsbak J.L., Abebe K.Z., et al. Predicted mutation strength of nontruncating PKD1 mutations aids genotype-phenotype correlations in autosomal dominant polycystic kidney disease. J Am Soc Nephrol. 2016;27:2872–2884. doi: 10.1681/ASN.2015050583. [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Lavu S., Vaughan L.E., Senum S.R., et al. The value of genotypic and imaging information to predict functional and structural outcomes in ADPKD. JCI Insight. 2020;5 doi: 10.1172/jci.insight.138724. [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Chang A.R., Moore B.S., Luo J.Z., et al. Exome sequencing of a clinical population for autosomal dominant polycystic kidney disease. JAMA. 2022;328:2412–2421. doi: 10.1001/jama.2022.22847. [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Cornec-Le Gall E., Torres V.E., Harris P.C. Genetic complexity of autosomal dominant polycystic kidney and liver diseases. J Am Soc Nephrol. 2018;29:13–23. doi: 10.1681/ASN.2017050483. [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Hwang Y.H., Conklin J., Chan W., et al. Refining genotype-phenotype correlation in autosomal dominant polycystic kidney disease. J Am Soc Nephrol. 2016;27:1861–1868. doi: 10.1681/ASN.2015060648. [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Porath B., Gainullin V.G., Cornec-Le Gall E., et al. Mutations in GANAB, encoding the glucosidase IIα subunit, cause autosomal-dominant polycystic kidney and liver disease. Am J Hum Genet. 2016;98:1193–1207. doi: 10.1016/j.ajhg.2016.05.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Claus L.R., Chen C., Stallworth J., et al. Certain heterozygous variants in the kinase domain of the serine/threonine kinase NEK8 can cause an autosomal dominant form of polycystic kidney disease. Kidney Int. 2023;104:995–1007. doi: 10.1016/j.kint.2023.07.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Hesketh S.J., Mukhopadhyay A.G., Nakamura D., Toropova K., Roberts A.J. IFT-A structure reveals carriages for membrane protein transport into cilia. Cell. 2022;185:4971–4985.e16. doi: 10.1016/j.cell.2022.11.010. [DOI] [PubMed] [Google Scholar]
13.Pedersen L.B., Rosenbaum J.L. Intraflagellar transport (IFT) role in ciliary assembly, resorption and signalling. Curr Top Dev Biol. 2008;85:23–61. doi: 10.1016/S0070-2153(08)00802-8. [DOI] [PubMed] [Google Scholar]
14.Satir P., Christensen S.T. Structure and function of mammalian cilia. Histochem Cell Biol. 2008;129:687–693. doi: 10.1007/s00418-008-0416-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Bayat A., Kerr B., Douzgou S., Study D.D.D. The evolving craniofacial phenotype of a patient with Sensenbrenner syndrome caused by IFT140 compound heterozygous mutations. Clin Dysmorphol. 2017;26:247–251. doi: 10.1097/MCD.0000000000000169. [DOI] [PubMed] [Google Scholar]
16.Perrault I., Perrault I., Saunier S., et al. Mainzer-Saldino syndrome is a ciliopathy caused by IFT140 mutations. Am J Hum Genet. 2012;90:864–870. doi: 10.1016/j.ajhg.2012.03.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Schmidts M., Schmidts M., Schmidts M., et al. Combined NGS approaches identify mutations in the intraflagellar transport gene IFT140 in skeletal ciliopathies with early progressive kidney Disease. Hum Mutat. 2013;34:714–724. doi: 10.1002/humu.22294. [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Senum S.R., Li Y., Benson K.A., et al. Monoallelic IFT140 pathogenic variants are an important cause of the autosomal dominant polycystic kidney-spectrum phenotype. Am J Hum Genet. 2022;109:136–156. doi: 10.1016/j.ajhg.2021.11.016. [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Clark D., Burns R., Bloom M.S., et al. Heterozygous loss of function variants in IFT140 are associated with polycystic kidney disease. Am J Med Genet A. 2024;194 doi: 10.1002/ajmg.a.63841. [DOI] [PubMed] [Google Scholar]
20.Dordoni C., Zeni L., Toso D., et al. Monoallelic pathogenic IFT140 variants are a common cause of autosomal dominant polycystic kidney disease-spectrum phenotype. Clin Kidney J. 2024;17 doi: 10.1093/ckj/sfae026. [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Jonassen J.A., Jonassen J.A., Jovenal T., Pazour G.J. Disruption of IFT complex A causes cystic kidneys without mitotic spindle misorientation. J Am Soc Nephrol. 2012;23:641–651. doi: 10.1681/asn.2011080829. [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Zagorec N., Calamel A., Delaporte M., et al. Clinical spectrum and prognosis of atypical autosomal dominant polycystic kidney disease caused by monoallelic pathogenic variants of IFT140. Am J Kidney Dis. 2025;85:465–476.e1. doi: 10.1053/j.ajkd.2024.10.009. [DOI] [PubMed] [Google Scholar]
23.Pei Y., Obaji J., Dupuis A., et al. Unified criteria for ultrasonographic diagnosis of ADPKD. J Am Soc Nephrol. 2009;20:205–212. doi: 10.1681/asn.2008050507. [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Pei Y., Hwang Y.H., Conklin J., et al. Imaging-based diagnosis of autosomal dominant polycystic kidney disease. J Am Soc Nephrol. 2015;26:746–753. doi: 10.1681/ASN.2014030297. [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Rossetti S., Consugar M.B., Chapman A.B., et al. Comprehensive molecular diagnostics in autosomal dominant polycystic kidney disease. J Am Soc Nephrol. 2007;18:2143–2160. doi: 10.1681/asn.2006121387. [DOI] [PubMed] [Google Scholar]
26.Rossetti S., Kubly V.J., Consugar M.B., et al. Incompletely penetrant PKD1 alleles suggest a role for gene dosage in cyst initiation in polycystic kidney disease. Kidney Int. 2009;75:848–855. doi: 10.1038/ki.2008.686. [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Dahl N.K., Bloom M.S., Chebib F.T., et al. The clinical utility of genetic testing in the diagnosis and management of adults with chronic kidney disease. J Am Soc Nephrol. 2023;34:2039–2050. doi: 10.1681/ASN.0000000000000249. [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Hopp K., Cornec-Le Gall E., Senum S.R., et al. Detection and characterization of mosaicism in autosomal dominant polycystic kidney disease. Kidney Int. 2020;97:370–382. doi: 10.1016/j.kint.2019.08.038. [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Perrone R.D., Abebe K.Z., Watnick T.J., et al. Primary results of the randomized trial of metformin administration in polycystic kidney disease (TAME PKD) Kidney Int. 2021;100:684–696. doi: 10.1016/j.kint.2021.06.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Gregory A.V., Chebib F.T., Poudyal B., et al. Utility of new image-derived biomarkers for autosomal dominant polycystic kidney disease prognosis using automated instance cyst segmentation. Kidney Int. 2023;104:334–342. doi: 10.1016/j.kint.2023.01.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Irazabal M.V., Rangel L.J., Bergstralh E.J., et al. Imaging classification of autosomal dominant polycystic kidney disease: a simple model for selecting patients for clinical trials. J Am Soc Nephrol. 2015;26:160–172. doi: 10.1681/asn.2013101138. [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Nowak K.J., Nowak K.L., You Z., et al. Overweight and obesity are predictors of progression in early autosomal dominant polycystic kidney disease. J Am Soc Nephrol. 2017;29:571–578. doi: 10.1681/asn.2017070819. [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Inker L.A., Eneanya N.D., Coresh J., et al. New creatinine- and cystatin C-based equations to estimate GFR without Race. N Engl J Med. 2021;385:1737–1749. doi: 10.1056/NEJMoa2102953. [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Salhi S., Doreille A., Dancer M.S., et al. Monoallelic loss-of-function IFT140 pathogenic variants cause autosomal dominant polycystic kidney disease: a confirmatory study with suspicion of an additional cardiac phenotype. Am J Kidney Dis. 2024;83:688–691. doi: 10.1053/j.ajkd.2023.08.019. [DOI] [PubMed] [Google Scholar]
35.Ghanem A., Borghol A.H., Munairdjy Debeh F.G., et al. Biomarkers of kidney disease progression in ADPKD. Kidney Int Rep. 2024;9:2860–2882. doi: 10.1016/j.ekir.2024.07.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Fujimaru T., Mori T., Sekine A., et al. Importance of IFT140 in patients with polycystic kidney disease without a family history. Kidney Int Rep. 2024;9:2685–2694. doi: 10.1016/j.ekir.2024.06.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Grantham J.J., Torres V.E., Chapman A.B., et al. Volume progression in polycystic kidney disease. N Engl J Med. 2006;354:2122–2130. doi: 10.1056/nejmoa054341. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary File (PDF)

mmc1.pdf^{(996KB, pdf)}

[bib1] 1.Chebib F.T., Torres V.E. Autosomal dominant polycystic kidney disease: core curriculum 2016. Am J Kidney Dis. 2016;67:792–810. doi: 10.1053/j.ajkd.2015.07.037. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib2] 2.Chebib F.T., Torres V.E. Recent advances in the management of autosomal dominant polycystic kidney disease. Clin J Am Soc Nephrol. 2018;13:1765–1776. doi: 10.2215/CJN.03960318. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib3] 3.Cornec-Le Gall E., Alam A., Perrone R.D. Autosomal dominant polycystic kidney disease. Lancet. 2019;393:919–935. doi: 10.1016/S0140-6736(18)32782-X. [DOI] [PubMed] [Google Scholar]

[bib4] 4.Cornec-Le Gall E., Audrézet M.P., Chen J.M., et al. Type of PKD1 mutation influences renal outcome in ADPKD. J Am Soc Nephrol. 2013;24:1006–1013. doi: 10.1681/ASN.2012070650. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib5] 5.Heyer C.M., Sundsbak J.L., Abebe K.Z., et al. Predicted mutation strength of nontruncating PKD1 mutations aids genotype-phenotype correlations in autosomal dominant polycystic kidney disease. J Am Soc Nephrol. 2016;27:2872–2884. doi: 10.1681/ASN.2015050583. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib6] 6.Lavu S., Vaughan L.E., Senum S.R., et al. The value of genotypic and imaging information to predict functional and structural outcomes in ADPKD. JCI Insight. 2020;5 doi: 10.1172/jci.insight.138724. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib7] 7.Chang A.R., Moore B.S., Luo J.Z., et al. Exome sequencing of a clinical population for autosomal dominant polycystic kidney disease. JAMA. 2022;328:2412–2421. doi: 10.1001/jama.2022.22847. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib8] 8.Cornec-Le Gall E., Torres V.E., Harris P.C. Genetic complexity of autosomal dominant polycystic kidney and liver diseases. J Am Soc Nephrol. 2018;29:13–23. doi: 10.1681/ASN.2017050483. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib9] 9.Hwang Y.H., Conklin J., Chan W., et al. Refining genotype-phenotype correlation in autosomal dominant polycystic kidney disease. J Am Soc Nephrol. 2016;27:1861–1868. doi: 10.1681/ASN.2015060648. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib10] 10.Porath B., Gainullin V.G., Cornec-Le Gall E., et al. Mutations in GANAB, encoding the glucosidase IIα subunit, cause autosomal-dominant polycystic kidney and liver disease. Am J Hum Genet. 2016;98:1193–1207. doi: 10.1016/j.ajhg.2016.05.004. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib11] 11.Claus L.R., Chen C., Stallworth J., et al. Certain heterozygous variants in the kinase domain of the serine/threonine kinase NEK8 can cause an autosomal dominant form of polycystic kidney disease. Kidney Int. 2023;104:995–1007. doi: 10.1016/j.kint.2023.07.021. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib12] 12.Hesketh S.J., Mukhopadhyay A.G., Nakamura D., Toropova K., Roberts A.J. IFT-A structure reveals carriages for membrane protein transport into cilia. Cell. 2022;185:4971–4985.e16. doi: 10.1016/j.cell.2022.11.010. [DOI] [PubMed] [Google Scholar]

[bib13] 13.Pedersen L.B., Rosenbaum J.L. Intraflagellar transport (IFT) role in ciliary assembly, resorption and signalling. Curr Top Dev Biol. 2008;85:23–61. doi: 10.1016/S0070-2153(08)00802-8. [DOI] [PubMed] [Google Scholar]

[bib14] 14.Satir P., Christensen S.T. Structure and function of mammalian cilia. Histochem Cell Biol. 2008;129:687–693. doi: 10.1007/s00418-008-0416-9. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib15] 15.Bayat A., Kerr B., Douzgou S., Study D.D.D. The evolving craniofacial phenotype of a patient with Sensenbrenner syndrome caused by IFT140 compound heterozygous mutations. Clin Dysmorphol. 2017;26:247–251. doi: 10.1097/MCD.0000000000000169. [DOI] [PubMed] [Google Scholar]

[bib16] 16.Perrault I., Perrault I., Saunier S., et al. Mainzer-Saldino syndrome is a ciliopathy caused by IFT140 mutations. Am J Hum Genet. 2012;90:864–870. doi: 10.1016/j.ajhg.2012.03.006. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib17] 17.Schmidts M., Schmidts M., Schmidts M., et al. Combined NGS approaches identify mutations in the intraflagellar transport gene IFT140 in skeletal ciliopathies with early progressive kidney Disease. Hum Mutat. 2013;34:714–724. doi: 10.1002/humu.22294. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib18] 18.Senum S.R., Li Y., Benson K.A., et al. Monoallelic IFT140 pathogenic variants are an important cause of the autosomal dominant polycystic kidney-spectrum phenotype. Am J Hum Genet. 2022;109:136–156. doi: 10.1016/j.ajhg.2021.11.016. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib19] 19.Clark D., Burns R., Bloom M.S., et al. Heterozygous loss of function variants in IFT140 are associated with polycystic kidney disease. Am J Med Genet A. 2024;194 doi: 10.1002/ajmg.a.63841. [DOI] [PubMed] [Google Scholar]

[bib20] 20.Dordoni C., Zeni L., Toso D., et al. Monoallelic pathogenic IFT140 variants are a common cause of autosomal dominant polycystic kidney disease-spectrum phenotype. Clin Kidney J. 2024;17 doi: 10.1093/ckj/sfae026. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib21] 21.Jonassen J.A., Jonassen J.A., Jovenal T., Pazour G.J. Disruption of IFT complex A causes cystic kidneys without mitotic spindle misorientation. J Am Soc Nephrol. 2012;23:641–651. doi: 10.1681/asn.2011080829. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib22] 22.Zagorec N., Calamel A., Delaporte M., et al. Clinical spectrum and prognosis of atypical autosomal dominant polycystic kidney disease caused by monoallelic pathogenic variants of IFT140. Am J Kidney Dis. 2025;85:465–476.e1. doi: 10.1053/j.ajkd.2024.10.009. [DOI] [PubMed] [Google Scholar]

[bib23] 23.Pei Y., Obaji J., Dupuis A., et al. Unified criteria for ultrasonographic diagnosis of ADPKD. J Am Soc Nephrol. 2009;20:205–212. doi: 10.1681/asn.2008050507. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib24] 24.Pei Y., Hwang Y.H., Conklin J., et al. Imaging-based diagnosis of autosomal dominant polycystic kidney disease. J Am Soc Nephrol. 2015;26:746–753. doi: 10.1681/ASN.2014030297. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib25] 25.Rossetti S., Consugar M.B., Chapman A.B., et al. Comprehensive molecular diagnostics in autosomal dominant polycystic kidney disease. J Am Soc Nephrol. 2007;18:2143–2160. doi: 10.1681/asn.2006121387. [DOI] [PubMed] [Google Scholar]

[bib26] 26.Rossetti S., Kubly V.J., Consugar M.B., et al. Incompletely penetrant PKD1 alleles suggest a role for gene dosage in cyst initiation in polycystic kidney disease. Kidney Int. 2009;75:848–855. doi: 10.1038/ki.2008.686. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib27] 27.Dahl N.K., Bloom M.S., Chebib F.T., et al. The clinical utility of genetic testing in the diagnosis and management of adults with chronic kidney disease. J Am Soc Nephrol. 2023;34:2039–2050. doi: 10.1681/ASN.0000000000000249. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib28] 28.Hopp K., Cornec-Le Gall E., Senum S.R., et al. Detection and characterization of mosaicism in autosomal dominant polycystic kidney disease. Kidney Int. 2020;97:370–382. doi: 10.1016/j.kint.2019.08.038. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib29] 29.Perrone R.D., Abebe K.Z., Watnick T.J., et al. Primary results of the randomized trial of metformin administration in polycystic kidney disease (TAME PKD) Kidney Int. 2021;100:684–696. doi: 10.1016/j.kint.2021.06.013. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib30] 30.Gregory A.V., Chebib F.T., Poudyal B., et al. Utility of new image-derived biomarkers for autosomal dominant polycystic kidney disease prognosis using automated instance cyst segmentation. Kidney Int. 2023;104:334–342. doi: 10.1016/j.kint.2023.01.010. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib31] 31.Irazabal M.V., Rangel L.J., Bergstralh E.J., et al. Imaging classification of autosomal dominant polycystic kidney disease: a simple model for selecting patients for clinical trials. J Am Soc Nephrol. 2015;26:160–172. doi: 10.1681/asn.2013101138. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib32] 32.Nowak K.J., Nowak K.L., You Z., et al. Overweight and obesity are predictors of progression in early autosomal dominant polycystic kidney disease. J Am Soc Nephrol. 2017;29:571–578. doi: 10.1681/asn.2017070819. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib33] 33.Inker L.A., Eneanya N.D., Coresh J., et al. New creatinine- and cystatin C-based equations to estimate GFR without Race. N Engl J Med. 2021;385:1737–1749. doi: 10.1056/NEJMoa2102953. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib34] 34.Salhi S., Doreille A., Dancer M.S., et al. Monoallelic loss-of-function IFT140 pathogenic variants cause autosomal dominant polycystic kidney disease: a confirmatory study with suspicion of an additional cardiac phenotype. Am J Kidney Dis. 2024;83:688–691. doi: 10.1053/j.ajkd.2023.08.019. [DOI] [PubMed] [Google Scholar]

[bib35] 35.Ghanem A., Borghol A.H., Munairdjy Debeh F.G., et al. Biomarkers of kidney disease progression in ADPKD. Kidney Int Rep. 2024;9:2860–2882. doi: 10.1016/j.ekir.2024.07.012. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib36] 36.Fujimaru T., Mori T., Sekine A., et al. Importance of IFT140 in patients with polycystic kidney disease without a family history. Kidney Int Rep. 2024;9:2685–2694. doi: 10.1016/j.ekir.2024.06.021. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib37] 37.Grantham J.J., Torres V.E., Chapman A.B., et al. Volume progression in polycystic kidney disease. N Engl J Med. 2006;354:2122–2130. doi: 10.1056/nejmoa054341. [DOI] [PubMed] [Google Scholar]

PERMALINK

Characterizing the ADPKD-IFT140 Phenotypic Signature With Deep Learning and Advanced Imaging Biomarkers

Ahmad Ghanem

Fadi George Munairdjy Debeh

Abdul Hamid Borghol

Nikola Zagorec

Amanda L Tapia

Byron Smith

Stefan Paul

Abdul Basit

Bassel AlKhatib

Nay Nader

Marie Therese Bou Antoun

Adriana V Gregory

Hana Yang

Rachel S Schauer

Neera K Dahl

Christian Hanna

Vicente E Torres

Timothy L Kline

Peter C Harris

Emilie Cornec-Le Gall

Fouad T Chebib

Abstract

Introduction

Methods

Results

Conclusion

Graphical abstract

Methods

Study Population

Development Cohort

Internal Specificity and External Sensitivity Cohorts

Genetic Analysis and the Identification of the Disease-Causing Variants

Segmentation and Imaging Biomarkers

Data Collection

Study Outcomes

Statistical Analysis

Results

Baseline Characteristics

Table 1.

Clinical Features and eGFR Rate of Decline of ADPKD-IFT140

Figure 1.

Imaging Features of ADPKD-IFT140

Table 2.

Figure 2.

Figure 3.

Distinguishing ADPKD-IFT140 Phenotype Using Advanced Imaging Biomarkers

Table 3.

Figure 4.

Table 4.

Figure 5.

Specificity Analysis Using an Internal Cohort

Table 5.

Table 6.

Figure 6.

Sensitivity Analysis Using an External Cohort

Discussion

Disclosure

Acknowledgments

Footnotes

Supplementary Material

References

Associated Data

Supplementary Materials

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases