Origin of lymphatic endothelial cells (LECs)

Vascular malformations (VM) can affect any type of vessel; blood vessels as well as lymphatics, isolated or in combined forms. A number of mutations, mostly loss of function mutations, have been identified to cause VM [17]. The combination of venous and lymphatic malformations is obviously rooted in the common origin of the two types of endothelial cells. Precursors for LECs reside in the lining of specific embryonic veins, but also in various mesenchymal compartments of vertebrate embryos [18, 4, 19]. Lymphangioblasts are still present in adult mouse lung [20], and also in the human, where they contribute to inflammation-induced lymphangiogenesis [21]. Smooth muscle cells in the tunica media of vessels originate from many different embryonic compartments [22]. In lymphatic collectors, they are in intimate contact with the luminal LECs, and, at transmission electron microscopic level, present as heterogenous population of cells [23]. Among the indispensable regulators of lymphangiogenesis are Vascular Endothelial Growth Factor-C (VEGF-C) and its high affinity receptor VEGFR-3 [24, 25, 4]. Loss of function mutations of VEGFR-3 are the main cause of congenital (primary) lymphedema, due to hypoplasia of the lymphatics [26].

Activating PIK3CA mutations in LECs of lymphatic malformations

The reasons for hyperplasia of lymphatics have long been unknown, however, recently it has been shown that lymphatic malformations (LM), isolated and in combination with other overgrowth syndromes, are mostly caused by activating mutations in the PIK3CA gene. Mutations have been found in 74% of 31 individuals with isolated LM in a cohort from the Seattle Children’s Hospital [7], and in a cohort from the Boston Children’s Hospital even >90% (16 of 17) LM were somatic mosaic for specific PIK3CA mutations. However, no significant correlation between type of mutation, macro-cystic or micro-cystic LM, or clinical severity could be observed [7]. Similarly, when LM is part of a complex syndrome, PIK3CA is of major importance. It has been shown that 19 of 21 individuals with Klippel-Trenaunay syndrome (KTS), 5 of 8 with fibro-adipose vascular anomaly (FAVA), and 31 of 33 with congenital lipomatous overgrowth with vascular, epidermal and skeletal anomalies (CLOVES) were mosaic for specific PIK3CA mutations [7]. Thereby, 5 recurrent mutations have been found in the majority (~80%) of cases, resulting in amino acid exchange p.C420R, p.E542K, p.E545K, p.H1047R, and p.H1047L [7]. However, the authors also noted that the number of mutated cells in the LM tissue was less than 10%.

Therefore, we isolated both LECs and fibroblasts from LM tissues of 6 patients, and screened the commonly affected exons 8, 10, and 21 of the PIK3CA gene. Of note, a change in exon numbering of PIK3CA has taken place. Since exon 1 is entirely UTR, numerous previous studies have numbered the exons according to the coding exons. As a result, these studies refer to E545 mutations as being in exon 9 and H1047 mutations in exon 20 [27]. However, according to the recommendations of the Human Genome Variation Society (HGVS) exon numbering should always start with exon 1 as it appears, including also non-coding exons (www.hgvs.org/mutnomen/refseq.html#exinnr).

For separation of LM-derived LECs and fibroblasts we performed a 'classical' mechanical isolation procedure, the cell sweeping procedure, originally described by Dr. Judah Folkman [28]. The reason for not performing immuno-isolation techniques with surface molecules such as Podoplanin (PDPN) and LYVE-1 is an obvious one. LM are often derived from lymphatic collectors. Importantly, collector LECs, but also numerous lymph node LECs [29] are negative for the two markers [2, 23], which also applies to 1/3 of LM [6]. The most reliable marker for LECs from all parts of the lymphovascular system, including the lymph nodes, is double-staining for CD31 and Prox1 [30, 29, 23]. This, together with Dapi, enables counting of positive vs. negative cells, and exact quantification. Remaining cells are mostly fibroblasts, which escaped the cell sweeping. Previous studies who have used CD31 plus PDPN sorting of LM-LECs [9, 10] have obviously selected for lymphatic capillary-derived LECs and missed collector-derived LECs.

We identified 1 new and 4 typical activating mutations in these exons, as well as a new in-frame Glu109 deletion in Ly-LECs, but never in fibroblasts. This clearly shows that the mutations are cell-type-specific. In two Ly-LEC lines we found new mutations. Thereby, the cell line Ly-LEC2 showed a Glu109 deletion, which results in the loss of an amino acid in the region linking PI3K-ABD and PI3K-RBD domains (https://ckb.jax.org/geneVariant/show?geneVariantId=3785). Such deletion has been found, among other multiple mutations, in several carcinomas [12, 13], but its effect on protein function has remained unknown. Here, we have shown AKT-Ser473 hyperphosphorylation, but absence of ERK hyperphosphorylation, of Ly-LEC2, indicating that Glu109del serves as an activating mutation, which however differs in its signaling cascade from the other activating mutations. In Ly-LEC-10 we found a mutation in exon 10 (c.1636C>A; p.Gln546Lys = p.Q546K), which has also not been detected in Ly-LECs before. Again, this mutation has previously been found in tumors such as ovarian tumors, glioblastoma, uterine carcinosarcoma, and squamous cell carcinoma [14, 15]; also see: https://www.ncbi.nlm.nih.gov/pubmed/26619011). Both AKT-Ser473 and ERK hyperphosphorylation in Ly-LEC10 shows that this mutation represents an activating mutation as well.

Previously, the p.H1047L mutation in the PIK3CA gene, in addition to a premature stop codon in the PI3K regulatory subunit-3 (PIK3R3), has been found in a LM-derived LEC line isolated at the Boston Children’s Hospital [9]. The authors also observed high angiogenic potential of this cell line in vitro, as well as hyperphosphorylation of AKT-Thr308. Correspondingly, in another four LM-derived LEC lines isolated at the Cincinnati Children’s Hospital, p.E542K and p.E545A mutations, lymphangiogenic phenotype and AKT activation were identified [10]. We found hyperphosphorylation of AKT-Ser473 in LM-derived LECs, which supports the well-known observation of the activation of angiogenesis via the PI3K—AKT—mTOR pathway [16]. Activating mutations in the PIK3CA gene have very recently also been found in patients with venous malformations [31, 32], underlining the need for a causal therapy of vascular malformations, e.g. with kinase inhibitors.

We have therefore tested a panel of 7 kinase inhibitors on LM-derived Ly-LEC in comparison to normal HD-LECs. We studied BKM-120, Wortmannin, Ly294002, which are inhibitors of PIK3CA, CAL-101 (Idelalisib), an inhibitor of PIK3CD, MK-2206, an AKT inhibitor, Sorafenib, a multi-kinase inhibitor, and rapamycin, which inhibits mTOR. We observed significant dose-dependent growth inhibition with all inhibitors used, however, normal LECs responded similarly. Of note, the PIK3CD inhibitor CAL-101 [33] showed no effects, or effects only at very high dosage, which underlines the fact that PIC3CA, but not PIC3CD, is involved in the disease.

In contrast to a previous study, which reported on specific rapamycin effects on LM-derived LECs (a single cell line was used), but not healthy LECs [9], we did not detect an immediately obvious therapeutic window in vitro, neither for rapamycin, nor for the other kinase inhibitors tested, with the exception that, however, MK-2206 might be worth to be further tested with 'typical' lymphatic malformations. After 48h, we observed reduction of cell numbers below the initial value in 2 out of 3 LM-derived Ly-LEC lines with 25μM MK-2206, but not in healthy HD-LECs. Thereby, the LM-derived LEC line that did not respond so well was Ly-LEC-2, which possesses the Glu109 deletion and no ERK hyperphosphorylation. Tests with other Ly-LEC lines will have to determine if MK-2206 might be the inhibitor of choice for patients with PIK3CA mutations accounting for the majority of cases. Thereby, the possibility that a second mutation may accompany the PIK3CA mutation has to be taken into account [34].

Can the in vitro data be directly transferred to the situation in patients, to infants or adults? It is known that blood vascular endothelial cells in adults hardly proliferate [35]. To the best of our knowledge, proliferation of blood vascular and lymphatic endothelial cells has never been studied in children, but there seems to be no doubt that the values are significantly higher. With immunofluorescence we have previously observed expression of VEGFR-2 and -3 in blood vessels and lymphatics of infants [2], which underlines the immaturity of vessels in early years after birth. Nevertheless, the turnover of endothelial cells in infants is likely to be significantly lower than the proliferation of the cells in vitro. Therefore, our in vitro data may not be directly transferrable to the in vivo situation, and a therapeutic window for kinase inhibitors may well be present. It goes without saying that children must be closely monitored under treatment. Even more likely, a therapeutic window may be present for the treatment of adult LM-patients.

In summary, our studies show that activating PIK3CA mutations in LM are specifically localized in LECs. We identified a new mutation in exon 10 and a Glu109 deletion of the gene and show that several kinase inhibitors have the potential to block LEC proliferation. However, since these effects are not restricted to mutated cells, caution is advisable when treating young patients.

Isolation and culture of cells

The Ethics Committee of the University Medical Hospital Göttingen approved this study (application no.: 18/1/18). Studies were performed with the informed and written consent of the patients or their legal representatives. Lymphangioma/LM-derived LECs (Ly-LECs) and LM-derived fibroblasts (Ly-F) were isolated as described [11]. Briefly, we used the cell sweeping procedure, a purely mechanical method which was originally introduced by Judah Folkman [28]. LM tissue was cut into small pieces and cultured in endothelial growth medium (MV2), including supplements, fetal bovine serum (PromoCell, Heidelberg, Germany), and 1% penicillin/streptomycin (Thermo-Fisher, Darmstadt, Germany). Ly-LECs and fibroblasts (Ly-F) that migrated out were separated mechanically, and cells were further cultured at 37°C and 5% CO₂. The lymphendothelial identity was regularly controlled by immunocytology with antibodies against CD31/PECAM-1 (BD, Heidelberg, Germany) and PROX1 (ReliaTech, Wolfenbüttel, Germany). Purity of the cells was at least 80–90%. Pure LM-derived fibroblasts (Ly-F) were cultured in RPMI 1640 (Lonza, Basel, Switzerland), and characterized by their morphology, growth behavior and antibodies against vimentin and α smooth muscle actin (αSMA). Normal foreskin-derived LECs (HD-LEC) were bought from PromoCell (Heidelberg, Germany) and controlled by their CD31/PROX1 expression. Cells were cultured as described above.

Mutation analysis of Ly-LEC and Ly-F

Genomic DNA of LM-derived cell lines (Ly-LEC, Ly-F) and normal lymphatic endothelial cells (HD-LEC) was extracted using DNeasy blood & tissue kit (Qiagen) according to the manufacturer’s protocol. For sequence analysis, exons 8, 10 and 21 (≙ coding exons 7, 9 and 20) of PIK3CA (phosphatidylinositol-4,5-bisphosphate 3-kinase, catalytic subunit alpha; NM_006218.2) were amplified from the genomic DNA using the primer sets depicted in Table 2. For exon 10, it was verified not to amplify parts of the pseudogene. PCR products were purified and used in the sequencing reaction, which included BigDye and sequencing primers, and was performed on ABI 3500 (Applied Biosystems, Thermo-Fisher, Foster City, CA). Data were analyzed using the software SeqPilot (JSI medical systems GmbH, Ettenheim, Germany).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Primers used for mutation analyses of PIK3CA.

https://doi.org/10.1371/journal.pone.0200343.t002

Immunocytology

Cells were seeded on chamber slides (BD). After 24 hours, immunostaining was performed as follows: after short (1 min) 4% paraformaldehyde (PFA) fixation, cells were washed with PBS and incubated in bovine serum albumin (BSA). Then 0.1% Tween/PBS (for better PROX1 staining results—30 seconds) was applied, followed by immunostaining. Anti-human PROX1 (1:200, ReliaTech, Wolfenbüttel, Germany), anti-human CD31 (1:50; BD, NJ, USA), mouse-anti-human vimentin (1:1, DAKO; clone V9), and mouse-anti-human smooth muscle α-actin (αSMA) (1:500, Sigma, clone 1A4) antibodies were incubated for one hour, rinsed with PBS, and then secondary antibody staining with Alexa594-conjugated goat-anti-rabbit and Alexa488-conjugated goat-anti-mouse (1:200; ThermoFisher Scientific, MA, USA) was performed. Cells were counter-stained with Dapi and analyzed with AxioImager Z.1 (Zeiss, Göttingen, Germany).

Western blot

Cells were washed twice with 5 ml DPBS (Pan Biotech; Aidenbach, Germany) and 1 mM sodium orthovanadate (Sigma-Aldrich; Steinheim, Germany) and subsequently lysed with RIPA buffer for 10 minutes on crushed ice. Samples were transferred to micro tubes (Sarstedt; Nürnberg, Germany) and centrifuged by 20,800x g for 15 minutes. Protein lysates were collected and used for SDS-PAGE according to standard procedures (Bio-Rad Laboratories; München, Germany). Proteins were transferred to PVDF membranes (Roth; Karlsruhe, Germany), which were then washed with TBST (0.1% Tween 20) for 10 minutes and blocked for one hour in blocking buffer (5% BSA in TBST 0.1%). Primary antibodies were: pAKT-Ser473, 1:500 in 5% BSA (rabbit mAb: Cell Signaling Technology, #4060); total AKT (rabbit pAb: Cell Signaling, #9272, 1:1000); phospho-p44/42MAPK (pERK1/2), Thr202/Tyr204, (rabbit mAb: Cell Signaling, #4370); p44/42MAPK (ERK1/2) (rabbit mAb: Cell Signaling, #4695); and peroxidase-conjugated α-tubulin, 1:10000 (Abcam, ab40742). Incubation was at 4 °C overnight. After washing with TBST (0.1% Tween 20) for 10 min., the peroxidase-conjugated secondary anti-rabbit IgG, 1:2000 (Santa Cruz; USA) were incubated for one hours. After washing, antigen-antibody complexes were visualized with ECL and Chemidoc Touch Imaging System (Bio-Rad).

Proliferation assays

Proliferation assays were performed as described [36]. Briefly, 5000 cells were seeded into 96-well cell culture plates (Sarsted, Nümbrecht, Germany), and allowed to adhere overnight. Inhibitors or solvent controls were added at indicated concentrations. Cells were fixed with 4% glutaraldehyde after 24, 48 and 72 hours, stained with crystal violet (0.01% in deionized water), and absorption was measured at 570 nm with a microplate reader. Data are presented as the mean values of n = 3 independent experiments with each 5–8 replicates.

Kinase inhibitors

All kinase inhibitors were bought from Seleckchem (Absource Diagnostics, Munich, Germany). We used Buparlisib/BKM120, Wortmannin, and Ly294002 (to inhibit PIK3CA), CAL-101/Idelalisib (to inhibit PIK3CD), MK2206 (AKT inhibitor), Sorafenib (multi-kinase inhibitor), and rapamycin (mTOR inhibitor). Inhibitors were diluted in DMSO (Sigma-Aldrich, Taufkirchen, Germany). Controls were performed with DMSO, corresponding to the highest concentration used in experimental groups.

Statistical methods

Statistical analyzes were conducted by two-way-ANOVA with STATISTICA software (Statsoft, Tulsa, Oklahoma). Normal distribution was verified before testing. The standard error of the mean (SEM) was calculated and is shown in the graphs. Significant differences were considered as p-value ≤ 0.05 [*], p ≤ 0.01 [**] and p ≤ 0.001 [***].