Microenvironment Regulators of Metastasis Favor Lymphangioleiomyomatosis Cell Growth and Disease Progression

To the Editor:

Lymphangioleiomyomatosis (LAM) is a destructive cystic lung disease diagnosed primarily in premenopausal women, affecting the lungs, kidneys (e.g., angiomyolipomas), and lymphatic system (e.g., lymphangioleiomyomas, chylous effusions, adenopathy) (1). Nodules in the walls of LAM lung cysts have a heterogeneous cellular composition with wild-type cells (e.g., type II pneumocytes, mast cells, lymphatic endothelial cells, lymphocytes, fibroblasts) and LAM cells, consisting of smooth muscle–like cells identified as epithelioid cells, expressing the melanosomal protein PMEL17 (premelanosome protein 17), and proliferative cells (2) with mutations or loss of heterozygosity (LOH) of the tuberous sclerosis complex genes TSC2 and/or TSC1. LAM can present sporadically or in association with tuberous sclerosis complex. Lung LAM cells have a benign appearance but are responsible for lung destruction and are capable of metastasis via the lymphatic and blood circulations (3, 4). Because of the cellular heterogeneity of the LAM nodule, we hypothesize that the microenvironment of neoplastic LAM cells plays a role in LAM cell phenotype, growth, and dissemination as well as in disease progression. We examined this hypothesis from four directions.

First, because neoplastic cells can grow in an anchorage-independent manner (5) and because it has been postulated that LAM cells possess neoplastic characteristics (6, 7), and they have been identified as benign but destructive, we seeded primary nontransformed cultured human lung LAM cells in soft agar (5) (see the data supplement) to determine if cultured cells from LAM lungs contain neoplastic cells. Cultured human LAM lung cells from explanted lungs of three different patients with LAM generated colonies in soft agar (Figure 1A), consistent with neoplastic cells (5). The resulting colonies showed genetic heterogeneity (see Figure E1 and Table E1) and were composed of wild-type cells and LAM cells characterized by TSC2 LOH and the presence of the prometastatic marker CD44v6 (7). CD44v6 is a splice variant of the hyaluronic receptor CD44, which is involved in cancer initiation and metastasis (8).

Figure 1.

Microenvironmental factors in lymphangioleiomyomatosis (LAM). (A) Growth of cells in soft agar. A549 lung adenocarcinoma cells (top panels, positive control) and cells from explanted lungs of three patients with LAM (LAM B17, LAM B23, and LAM A30) were tested for growth in soft agar. Cells, grown as described previously (7), were placed in soft agar, where they formed colonies, seen at low (left panels,10×) and high magnification (right, 40×). (B) Immunoreactivity of LAM lung tissue sections with anti-CD66b (neutrophil marker) antibodies, anti-CD68 (macrophage marker), and anti–FSP-1 (fibroblast marker). LAM cells were identified by their reactivity to the monoclonal antibody HMB-45 (red). Scale bars, 10 μM (top panels) and 100 μM (bottom panels). (C). Concentrations of growth factors known to regulate CD44v6 expression. Plasma from patients with LAM and healthy volunteers (HV) was used to determine the concentrations of OPN (osteopontin P ≤ 0.001) (7), HGF (hepatocyte growth factor) (patients with LAM, n = 45; HV, n = 36, P ≤ 0.001), and SDF-1 (stromal cell–derived factor 1) or CXCL12 (patients with LAM, n = 56; HV, n = 20, P ≤ 0.001). Concentrations of these growth factors were significantly higher in patients with LAM. (D) Increased HGF production in cells with loss of TSC2. Cell culture supernatants were collected from tuberous sclerosis complex normal fibroblasts and TSC2-null cells (13 paired samples) (upper panel) and from neonatal foreskin fibroblasts transduced with lentiviral shRNA to TSC2 compared with shNT control (n = 4) (lower panel) for measurement of human HGF using ELISA. Results were normalized to total cellular protein. Data are expressed as mean ± SD. *P < 0.05. (E) Summary of the data showing that the LAM cell in the lung and circulation can increase the production of CD44v6 regulators (OPN, HGF, and SDF-1 or CXCL12). The production of these growth factors may be potentiated by the SNPs of each patient. The fact that CD44v6 is a coreceptor of HGF supports a positive feedback mechanism of action of the microenvironment by increasing the concentrations of HGF and increasing the splicing of CD44 to generate CD44v6. Thus, the effect of the growth factors (HGF, OPN, and SDF-1 or CXCL12) on cells expressing CD44v6 can increase proliferation and growth; LAM cells expressing CD44v6 will proliferate, and TSC2 loss of heterozygosity would characterize those cells. Furthermore, it is postulated that LAM cells as neoplastic cells require other cells of the microenvironment (e.g., macrophages, neutrophils) to form nodules or “microtumors.” FSP-1 = fibroblast-specific protein-1; HMB-45 = human melanoma black 45; shNT = nontarget.

Second, because of the heterogeneity of cells obtained in soft agar, we postulated intercellular communication within LAM nodules. Cells reactive to anti–FSP-1 (fibroblast-specific protein-1), anti-CD68 (macrophages), and anti-CD66b (neutrophils) antibodies were located near human melanoma black 45–positive cells (HMB-45, LAM cells) (Figure 1B) and are part of the LAM nodule. Similarly, cells from the lungs of patients with LAM and colonies grown in soft agar showed immunoreactivity to anti-CD66b and anti-CD68 antibodies (see Figure E2). A human kidney TSC2-null cell line (angiomyolipoma 621) formed aggregates with wild-type fibroblasts from LAM lung in the presence of SDF-1 (stromal cell–derived factor 1) or CXCL12 released by fibroblasts (9), similar to the intercellular communication we postulate for the LAM nodule. The cellular heterogeneity of soft-agar colonies mirrors the cellular composition of LAM lung nodules containing LAM cells expressing CD44v6 in addition to fibroblasts, macrophages, and neutrophils, as seen in other cancers (10).

Third, CD44v6 is an inherent marker in LAM cells, as it is expressed in cells with TSC2 LOH from blood (see Table E2), bronchoalveolar lavage fluid (BALF) (see Table E3), chylous effusions (see Table E4), and urine (11) and in cells of the lung nodule (7), supporting the hypothesis that metastatic LAM cells express this prometastatic and procancer initiating receptor (12).

Fourth, to obtain insight into regulators of CD44v6 expression in patients with LAM (13), we measured the concentrations of known growth factor regulators of CD44v6 expression (e.g., OPN [osteopontin], HGF [hepatocyte growth factor], SDF-1 or CXCL12). We found that OPN, HGF, and SDF-1 or CXCL12 were elevated in patients with LAM compared with healthy volunteers (Figure 1C). Furthermore, polymorphisms in HGF, OPN, and SDF-1 or CXCL12 are associated with interstitial lung disease, colorectal cancer, and metastasis. Patients with LAM and healthy volunteers were genotyped for different polymorphisms (see Table E5). The frequencies of three SNPs were found to be significantly different between patients with LAM and healthy volunteers (HGF A43839T [P < 0.001], HGF C57488A [P = 0.003], and OPN66 rs28357094 [P = 0.006]) (see Table E6).

The effects of the SNPs on the yearly decline in percent predicted forced expiratory volume in 1 second (FEV₁) and diffusing capacity of the lungs for carbon monoxide (Dl_CO) were examined, adjusting for initial FEV₁ or Dl_CO and the time of visit (Table 1). All data used were obtained from patients who had not been treated with rapamycin, an mTOR inhibitor. Patients with LAM homozygous for the A allele of HGF A43839T had a significantly faster rate of decline in FEV₁ (−1.81 ± 0.06% predicted per year) than patients heterozygous for AT (−1.37 ± 0.13% predicted per year) (P = 0.002). Polymorphisms in HGF did not affect the rate of decline of Dl_CO. Interestingly, three SNPs in OPN were strongly correlated with the rate of decline in Dl_CO, including OPN282, OPN750, and OPN1239. The polymorphism OPN1083 was correlated with the rate of FEV₁ decline. The polymorphism of SDF-1 or CXCL12 correlated with the rates of decline in both FEV₁ and Dl_CO.

Table 1.

Association of SNPs in Hepatocyte Growth Factor, Osteopontin, and Stromal Cell–derived Factor 1 or CXCL12 Genes with Yearly Rate of Change in Percent Predicted Forced Expiratory Volume in 1 Second or Dl_CO

		n	FEV1 Rate (% Predicted/yr)	P Value	DlCO Rate (% Predicted/yr)	P Value
HGF C(−1652)T promoter	CC	45	−1.83 ± 0.10	NS	−1.82 ± 0.10	NS
	CT	82	−1.70 ± 0.07		−1.55 ± 0.07
	TT	39	−1.80 ± 0.13		−1.78 ± 0.13
HGF A43839T intron 8	AA	101	−1.81 ± 0.06	0.002 (AA vs. AT)	−1.58 ± 0.06	NS
	AT	38	−1.37 ± 0.13		−1.70 ± 0.12
	TT	0	N/A		N/A
HGF C44222T intron 9	CC	8	−1.69 ± 0.24	NS	−1.84 ± 0.26	NS
	CT	44	−1.71 ± 0.10		−1.60 ± 0.10
	TT	105	−1.79 ± 0.06		−1.65 ± 0.06
HGF C57488A intron 13	CC	105	−1.80 ± 0.06	NS	−1.64 ± 0.06	NS
	CA	50	−1.70 ± 0.09		−1.62 ± 0.10
	AA	9	−1.50 ± 0.24		−1.87 ± 0.24
OPN66 rs28357094	GG	5	−1.88 ± 0.41	NS	−1.08 ± 0.40	NS
	GT	83	−1.86 ± 0.07		−1.74 ± 0.07
	TT	79	−1.61 ± 0.08		−1.51 ± 0.08
OPN282 rs4754	CC	20	−1.51 ± 0.16	NS	−1.42 ± 0.17	0.0003 (CT vs. TT)
	CT	65	−1.74 ± 0.08		−1.40 ± 0.08
	TT	84	−1.85 ± 0.07		−1.79 ± 0.07
OPN750 rs1126616	CC	89	−1.89 ± 0.07	NS	−1.85 ± 0.07	0.0002 (CC vs. CT)
	CT	56	−1.72 ± 0.09		−1.43 ± 0.09
	TT	21	−1.65 ± 0.17		−1.52 ± 0.17
OPN1083 rs1126772	AA	95	−1.68 ± 0.07	0.0063 (AA vs. AG)	−1.71 ± 0.07	NS
	AG	60	−1.98 ± 0.09		−1.53 ± 0.09
	GG	11	−2.34 ± 0.24		−1.91 ± 0.24
OPN1239 rs9138	AA	78	−1.81 ± 0.08	NS	−1.88 ± 0.08	⩽0.0003 (AA vs. AC or CC)
	AC	42	−1.93 ± 0.10		−1.34 ± 0.10
	CC	34	−1.89 ± 0.13		−1.36 ± 0.12
SDF-1 rs1801157	AA	12	−2.03 ± 0.17	0.0008 (AG vs. GG)	−2.72 ± 0.17	<0.0001 (AA vs. AG or GG); 0.0045 (AG vs. GG)
	AG	71	−1.58 ± 0.07		−1.38 ± 0.07
	GG	82	−1.94 ± 0.08		−1.68 ± 0.08

Definition of abbreviations: Dl_CO = diffusing capacity of the lungs for carbon monoxide; FEV₁ = forced expiratory volume in 1 second; HGF = hepatocyte growth factor; N/A = not applicable; NS = not significant; OPN = osteopontin; SDF-1 = stromal cell–derived factor 1.

Pairwise genotype comparisons were adjusted for the initial value of FEV₁ or Dl_CO and the time of visit. All patients were women, with an initial age of 41.3 ± 0.7 years (mean ± SEM), initial percent predicted FEV₁ of 73.5 ± 1.8, and initial percent predicted Dl_CO of 69.4 ± 2.1. One hundred seventy patients were White, 9 Asian, 6 African American, 1 American Indian or Alaska Native, and 2 of unknown race. None of the patients were treated with sirolimus or everolimus.

The polymorphism HGF A43839T correlated significantly with HGF concentration (see Table E7); patients with AA had higher concentrations of HGF than those with AT. In addition, we found that human TSC2-null (TSC2^−/−) skin cells expressed more HGF than did TSC2^+/− cells from the same patient (Figure 1D). Silencing of TSC2 resulted in greater expression of HGF than was seen in the control cells expressing TSC2 (tuberin) (Figure 1D). These data suggest that LAM cells express the HGF receptor (c-Met) and that lack of TSC2 results in increased HGF concentrations. We found that circulating LAM cells expressing CD44v6 with TSC2 LOH possess c-Met (see Table E8), whose function is modulated by CD44v6 (8, 12, 14).

As TSC1 and TSC2 have not been identified as tumor driver genes in multiple screens (15), the growth advantage for TSC2-null cells is unknown. We propose that the genetic reprogramming of the LAM cell, due to loss of TSC2 function, may promote unregulated growth and mobilization of wild-type and mutated cells, leading to the formation of hamartomas or benign tumors (Figure 1E). LAM cells in the lung and circulation can increase the production of CD44v6 regulators (OPN, HGF, SDF-1 or CXCL12). The production of these growth factors can be potentiated on the basis of the genotype of each patient. The fact that CD44v6 is a coreceptor of HGF supports a positive paracrine and autocrine feedback mechanism of action of the microenvironment by increasing the concentrations of HGF and increasing the splicing of CD44 to generate CD44v6 and potentiating LAM cell growth. Therefore, neoplastic LAM cells appear to interact with and require the cells of the microenvironment (e.g., macrophages, neutrophils) to form nodules or “microtumors.”