The Reducing Effect of Sirolimus on Angiomyolipoma is Determined by Decrease of Its Fat-Poor Compartments and Is Associated with Striking Reduction of Vascular Structures

Background: Renal angiomyolipomas hemorrhage is strongly associated with their size and vascular constitution. Since previous studies showed that mTOR inhibitors can differentially act in distinct cell lines, we analyzed the effects of sirolimus on the volume of different components of angiomyolipomas in patients with tuberous sclerosis complex, including vascular structures. We retrospectively analyzed 23 angiomyolipomas from 10 patients treated with sirolimus. An approach based on a Hounsfield-unit threshold was used to classify angiomyolipomas in fat-rich, fat-poor and intermediate-fat tumors, and to categorize tumor compartments in fat rich, fat poor, intermediate fat and highly vascularized. Diameter variations were measured to assess the effects on aneurysmatic/ectatic vascular formations. Results: Volume reduction following treatment with sirolimus was higher in fat-poor than fat-rich angiomyolipomas. Tumor reduction was mainly determined by decrease of the fat-poor and highly-vascularized compartments, with a lesser contribution of the intermediate-fat component, while the volume of the fat-rich compartment increased. Broad liposubstitution was observed in some tumors. Massive reduction of aneurysmatic/ectatic vascular structures was observed, with disappearance of such lesions in most cases. Conclusions: Our study showed that sirolimus reduces the size of angiomyolipomas by decreasing primarily their highly-vascularized and fat-poor compartments. This effect is associated with a remarkable reduction of tumoral aneurysms/ectatic vessels and other vascular structures. Our findings revealed, therefore, the likely mechanism responsible for the reported risk-decreasing effect of mTOR inhibitors on angiomyolipoma bleeding. Our findings also expand the understanding the biology of this tumor, supporting a role for mTOR in maintenance, and maybe assembling, of blood vessels in angiomyolipomas.


Background
Renal angiomyolipomas (AMLs) are neoplasms originated from the perivascular epithelium, containing adipocyte-like, muscle-like and epithelioid cells as well as dysmorphic blood vessels (1,2). AMLs affect up to 2.2% of general adult population (3) and are usually sporadic, however approximately 10% of the cases are associated with tuberous sclerosis complex (TSC).(4) TSC is characterized by the development of neoplasm in various organs and tissues, particularly skin, central nervous system, kidneys, lungs and heart (5,6).
Notably, AMLs have been reported in 49-60% of TSC patients evaluated by renal imaging (7,8). Renal AMLs have also been associated with pulmonary lymphangioleiomyomatosis (LAM) both in the scope of TSC and in sporadic cases (9,10). Interestingly, LAM affects almost exclusively females, is present in ~ 30% of the TSC patients, and 47-60% of its sporadic cases develop AMLs (11,12).
The effects of mTORi on the different AML components, however, are not well characterized. A recent study reported a heterogeneous volume-reducing effect of the mTORi everolimus on AMLs. In that study a more efficient reduction was observed in fatpoor tumors (27), however, the effects on the different tumor compartments have not been assessed. Since hemorrhage is strongly associated with tumor vascularization and presence of intra-tumoral aneurysms larger than 0.5 cm (28,29), the response of this compartment to mTORis is critical to understand the role of mTOR in AML vascular structure. Moreover, given that bleeding is the most relevant clinical aspect associated with AMLs, it is essential to evaluate the effect of these drugs on the patient's bleeding risk. In this study we unraveled different effects of sirolimus on the distinct AML compartments as well as its remarkable reduction effect on the vascular tumor component.

Study Population and Radiologic Analyses
We retrospectively identified nine patients followed at the University of São Paulo Medical Center between April/2010 and March/2018 treated with sirolimus for at least three months due to large AMLs. All patients were submitted to computed tomography scan (CT) prior and after the mentioned period of sirolimus therapy and had pre-and post-contrast imaging.
In non-contrasted sequences we classified the lesions in three categories based on the average tumor attenuation, expressed in Hounsfield units (HU): fat-rich (<-30 HU; FRT), intermediate-fat (≥-30 and ≤ 30 HU; IFT) and fat-poor (> 30 HU; FPT). All patients were screened for FRT, IFT and FPT, having one representative lesion of each profile been selected for analysis whenever identified. Total volume response to sirolimus was quantified for all included tumors. This evaluation was followed by corticomedullary-phase analysis, which allowed the measurement of such a response in specific AML compartments.
The pixel distribution associated with each tumor was initially plotted according to their attenuation pattern, giving rise a specific histogram ( Figure S1). Pixel densities below − 30 HU corresponded to fat-rich compartments (FRC), densities ≥-30 and < 30 HU were associated with intermediate-fat compartments (IFC), attenuation ≥ 30 and < 100 HU indicated a fat-poor compartment (FPC), and pixel densities ≥ 100 HU identified highlyvascularized compartments (HVC). This approach allowed the quantification of volume response to sirolimus for each of the AML compartments, characterizing its potentially differential effects upon each of the tumor components. To improve the assessment of sirolimus actions on the AML vascular structures, each of the analyzed tumors had the diameter of its largest aneurysmatic/ectatic vessel determined following the CT corticomedullary-phase.

Statistical Analyses
Data on continuous variables were tested for normality using the Shapiro-Wilk test. Since this analysis revealed non-parametric distributions, these data are presented as median and 25 and 75 percentiles. Categorical data are expressed as absolute values and percentages. Non-parametric data were compared using the Mann-Whitney U test for two independent samples or the Wilcoxon test for two-time measures. Non-parametric data associated with multiple groups were compared using the Kruskal-Wallis test, while multiple comparisons were corrected applying the Bonferroni method.
Given the limited number of cases, multivariable analyses for binary endpoints were performed using logistic regression with Firth´s penalization (30). Statistical significance was considered for asymptotic P < 0.05. The analyses were performed using SPSS 24.0, GraphPad Prism 8.0 and Stata 16.0.

Baseline Patient and Angiomyolipoma Features
We selected 10 patients that fulfilled all inclusion criteria. Twenty-three AMLs from these patients were selected for analyses, since in seven cases one of the AML fat profiles was not identified. The mean AML pre-treatment size was 4.5 cm (2.3-5.1). The baseline features of analyzed patients and respective tumors are depicted in Table 1. Table 1 Characterization of patients and angiomyolipomas

AML Fat Poorness Predicts a Better Response to Sirolimus
Using the Firth's logistic regression for an at-least-50% response of volume reduction, we evaluated fat profile, gender, age, serum sirolimus level, treatment length and tumor size at baseline as potential predictors of volume response to sirolimus. This analysis identified the FPT pattern as the sole predictor of an at-least-50% decrease in tumor volume following this treatment (Table 2).

Sirolimus Induces Massive Reduction of AML Vascular Structures and Is
Essentially Not Effective in the Fat-Rich Compartment Volume analyses of the different AML compartments following sirolimus revealed that all components, except for FRC, responded to treatment with significant volume reduction ( Fig. 3a and 4). Notably, HVC presented a remarkable volume decrease following this  (Fig. 3d), a response of 77.8% (25.0-91.9) volume decrease among the tumors (Fig. 3a). The AML compartment with attenuation of 30 HU or more was, in fact, responsible for 76.7% (25.6-100.0) of total tumor reduction.
Consistently with the aforementioned findings, IFC volume responded to treatment with sirolimus with an intermediate behavior between FRC and FPC. Indeed, the IFC volume decreased from 2.5 cm 3 (0.8-10.0) to 1.5 cm 3 (0.3-4.9) after the mentioned therapy, P = 0.014 (Fig. 3f), a response also expressed as 61.5% (-11.7-75.0) volume reduction among the analyzed AMLs (Fig. 3a). It must be noted, however, that the IFC volume change was significantly different from the FRC one but did not significantly differ from the FPC volume variation (Fig. 3a).
As a result of these differential effects of sirolimus on the AML compartments, this treatment leads to critical tumor structural changes. Within the scope of visible transformations of the tumor fat profile, some FPTs respond to sirolimus becoming FRTs (Fig. 6).

Discussion
Renal complications constitute the main cause of death among TSC patients (31)(32)(33). A significant portion of such complications include AML hemorrhage. Although end-stage renal disease is reported at low rates in TSC patients (34,35), the studied populations are usually young. Notably, up to 40% of the TSC patients develop chronic kidney disease (CKD), exhibiting an estimated glomerular filtration rate equivalent to 30-year-older subjects from the general population (36). While the pathogenesis of TSC-associated CKD remains not completely understood, several factors are known to contribute to renal function decline, including tumor bleeding, tumor encroaching to normal surrounding renal tissue, renal cystic involvement, focal and segmental glomerulosclerosis and tubulointerstitial disease. A molecular mechanism involving TSC1 or TSC2 loss of function in renal parenchyma has also been proposed to play a role in reducing glomerular filtration rate (2,37). Invasive interventions to prevent tumor hemorrhage, moreover, including partial nephrectomy and selective arterial embolization, can also impact on early loss of renal function as a consequence of loss of functional renal tissue (32). It should be noted that these invasive procedures are not uncommon in clinical practice (32,36), given the potential lethality associated with AML hemorrhage. A recent databank study reported that 24.2% of TSC patients had at least one invasive kidney intervention.
(38) Because TSC patients usually present multiple AMLs, and the incidence increases with age, such interventions usually cannot treat all lesions. Invasive procedures, therefore, are often repeated along life, leading to increased risk of CKD (32).
In response to this scenario, clinical studies have shown efficacy of mTORi in reducing AML volume, supporting these drugs as first-line therapy of asymptomatic AMLs larger than 3 cm in TSC patients (37). Such studies, however, had the primary end-point focused on tumor size reduction, not addressing whether the verified AML shrinkage lowers the bleeding rate and preserves renal function. Interestingly, however, no event of AML bleeding was reported in the extension of the phase 3 study "Everolimus for renal angiomyolipoma in patients with tuberous sclerosis complex or sporadic lymphangioleiomyomatosis (EXIST-2)" (39), a trial that followed 112 patients treated with everolimus for a median period of 28.9 weeks. This finding suggests a bleeding protective role of mTORi.
Despite the recognition of this effect of mTORi on AML volume, the response to these drugs has been shown to significantly vary among such tumors (39). In addition, AMLs are neoplasms histologically complex, with components derived from at least three distinct cellular phenotypes. Based on these two observations, we raised the hypothesis that Our findings also revealed that the differential effect of sirolimus on AMLs is essentially based on differential effects on specific tumor compartments. Our data showed a primary reduction of the fat-poor portions of the tumors. Interestingly, the profound effect of sirolimus on the FPCs and HVCs promoted the transformation of some FPTs into FRTs, a finding not yet reported. The liposubstitution observed in such AMLs represents, in fact, the most striking translation of the remarkable differential actions of mTORi in the different AML compartments. Our findings also unraveled a dramatic decrease of vascular aneurysmatic formations. In line with this concept, the volume of the AML fat-rich compartment did not decrease, but instead increased, in response to treatment. Our findings of massive reduction of the vascular tumor components in response to sirolimus, in turn, suggest that this treatment is likely protective against AML bleeding.
Our data are in line with previous observations that in vitro effects of sirolimus differ among distinct cell lineages (40) and TSC patients display high levels of the proangiogenic molecules VEGF-A and VEGF-D (vascular epithelial growth factors A and D) (41). mTOR inhibitors, therefore, could differentially act in the different components of AML, with particularly high efficiency on vascular formations. mTOR complex 1 (mTORC1), in fact, is known to drive HIF-1α (hypoxia-induced factor 1α) and VEGF-A signaling via multiple mechanisms involving 4E-BP1 (eukaryotic translation initiation factor 4E-binding protein 1), S6K1 (p70 ribosomal protein S6 kinase 1) and STAT3 (signal transducer and activator of transcription 3), an angiogenic process potentially attenuated by sirolimus (42). Interestingly, this mechanism is associated with acceleration of endothelial senescence (43). It is possible, therefore, that the molecular basis of the differential mTORi effects on the AML compartments is based on VEGF downregulation induced by sirolimus through inhibition of mTORC1 (44). Reduction of VEGF-D circulating levels was also reported in patients receiving rapamicyn. (24) The remarkable reduction of AML vascular components shown in our study suggests that mTOR activity is fundamental to the maintenance of such structures in AMLs. Moreover, since aneurysms are the main determinant for tumor hemorrhage (28), the inhibition of maintenance and development of aneurysmatic/ectatic formations induced by mTORi is the likely mechanism responsible for the reduction of bleeding occurrence observed in the EXIST-2 study and its extension (22,39).

Conclusions
Our findings suggest, therefore, that image analysis of AML compartments is likely helpful to predict tumor response to mTORi. Our data revealed, moreover, that treatment with sirolimus not only reduces the tumor size but selectively acts on components associated with more often and more severe clinical complications, such as AML bleeding. Our results provide additional support to the recent recommendation of chronic treatment with mTORi of TSC individuals with AMLs > 3 cm, and suggest that the presence of large aneurysms/vascular formations should be an independent criterium to initiate this therapy.

Consent for publication:
Not applicable

Availability of data and materials:
The datasets generated and/or analysed during the current study are not publicly available due to including files and images with personal information of patients, but, are available from the corresponding author on reasonable request.

Competing interests:
All the authors have no conflicts of interest to disclosure.

Funding:
This work was funded by institutional resources from the University of São Paulo Medical Center.

Author´s contributions:
Elieser H Watanabe had a lead role in conceptualization, methodology, formal analysis,