Levels of circulating endothelial cells are low in idiopathic pulmonary fibrosis and are further reduced by anti-fibrotic treatments

The cytometric approach that we used in this investigation was different from that of previous studies, which, mainly for technical reasons, could only analyse a relatively low number of events. The high number of cells that we could acquire and analyse, along with the use of Poisson statistics, allowed a correct interpretation of the data [23]. As shown in Fig. 1, which reports a representative analysis of CEC and EPC, cells were first selected according to physical parameters; debris and aggregates were then removed according to the forward scatter (FSC)-A versus FSC-H dot plot. In this population, dead cells and monocytes were removed using a ‘dump’ channel. The parameter ‘time’ shown in the middle upper panel was used to monitor the stability of the flow cytometric high-speed acquisition of events. CEC and EPC were identified on the basis of the expression of CD34, CD45 and CD133: CEC were defined as CD45dim, CD34+ and CD133− while EPC were defined as CD45−, CD34+ and CD133+ [24]. The parental population was represented by peripheral blood mononuclear cells that were alive (i.e. negative to LIVE/DEAD staining) and negative for CD14. Expression of CD309 (i.e. the type II receptor for the vascular endothelial growth factor, VEGFR-2, also named KDR) was detected among EPC and CEC.

The gating strategy used for the identification of circulating fibrocytes involved the exclusion of aggregates (by using an FSC-A versus FSH-H dot plot). In this population, T lymphocytes, B lymphocytes and dead cells were excluded. In this pure population, circulating fibrocytes were defined as CD34+, CD45+ and collagen I+ cells (Fig. 2). The FMO approach was also used to detect positive cells. Moreover, we could search for the presence of CXCR4 circulating fibrocytes (see below).

By using a sophisticated multiparametric analysis on a very sensitive flow cytometer, we could study and precisely quantify the presence of fibrocytes in patients with IPF. It is of note that, in order to avoid a possible unspecific staining due to a secondary antibody, we used a directly conjugated mAb recognizing collagen I and freshly isolated peripheral blood cells (Fig. 2).

In patients with IPF, the proportion of fibrocytes was below 1 % in almost all samples (see a representative example in Fig. 2, middle right panel). This was also the case in healthy controls (not shown). This result is in contrast with previous studies performed on freeze-thawed, fixed, permeabilized peripheral blood, which claim that the percentage of fibrocytes in acute IPF can be as high as 20 % [25]. This observation has been argued by others, because samples were not optimally used [26], and the number of events was likely too low to reach statistical significance. Moreover, no functional analysis have been performed on purified populations of fibrocytes isolated from blood to demonstrate their lineage. In our study, we have not confirmed the previously reported high levels of fibrocytes in blood from patients with IPF. Moreover, because it is almost impossible to sort and perform functional analysis of such cells (i.e., cells that are not viable, because their cytometric identification through the identification of collagen I requires permeabilization of plasma membrane and cell fixation), further studies are needed to clarify the meaning of collagen I+ cells, which are at present defined as fibrocytes, in peripheral blood.

Patients with IPF displayed low levels of CEC (Fig. 3a), along with a significantly lower amount of CEC expressing CD309 (Fig. 3b), as compared to controls. They also displayed a slightly higher percentage of EPC (Fig. 3c), which, however displayed a lower expression of CD309 (Fig. 3c) than those of healthy participants. It is noteworthy that six out of seven of the patients with the highest levels of EPC were untreated.

The decrease in CD309 expression could be due to different factors, ranging from the progression of the disease per se in untreated patients to the pharmacological effect of pirfenidone and nintedanib, which can decrease the expression of CD309/VEGF-R, altering the VEGF–VEGFR axis [27]. For instance, nintedanib binds to the intracellular ATP-binding pocket of fibroblast growth factor (FGF) receptors, platelet-derived growth factor (PDGF) receptors and VEGFRs, blocking the autophosphorylation of these receptors and the downstream signalling cascades (reviewed in [28]). Alternatively, it could be hypothesized that CEC are able to home into injured tissue to participate in lung re-endothelization, and this phenomenon diminishes their number in peripheral blood.

We then compared the levels of endothelial cell populations and collagen I+ cells in untreated and treated patients with IPF. Untreated patients displayed lower levels of CEC than patients treated with nintedanib or pirfenidone (Fig. 4a); treated patients also displayed lower levels of CEC expressing CD309 (Fig. 4b). Among these three groups of patients with IPF, we did not find statistically significant differences in EPC population (Fig. 4c), although the percentage of EPC expressing CD309 was lower in patients treated with nintedanib (Fig. 4d). The percentage of circulating collagen I+ cells, defined as fibrocytes (Fig. 4e), and of fibrocytes expressing CXCR4 (Fig. 4f) was similar between untreated and treated patients.

We analysed the percentage and the phenotype of CEC, EPC and circulating fibrocytes in 12 patients before and after 6 months of anti-fibrotic treatment. It has to be underlined that the detection and quantification of circulating fibrocytes, that is, cells that express collagen I, is quite problematic for several reasons, starting with their extremely low number. Furthermore, to be extremely rigorous, we cannot exclude the possibility that some CD14+ cells that express CD34 (or that just bind unspecifically to the anti-CD34 mAb by Fc receptors) could express collagen I, and thus this population could become an artefact of the analysis. In any case, given that this hypothesis is quite unlikely and we performed all possible quality control measures, we clearly show that collagen I+ cells decrease significantly after therapy.

The percentages of CEC and CEC expressing CD309 were significantly decreased after 6 months of treatment (Fig. 5a, b). After 6 months, patients with IPF displayed no differences in the percentage of EPC (Fig. 5c), nor in the percentage of EPC expressing CD309 (Fig. 5d). Moreover, after 6 months of nintedanib and pirfenidone treatments, circulating fibrocytes were nearly undetectable in most patients (Fig. 5e), and circulating fibrocytes expressing CXCR4 showed a significant decrease (Fig. 5f). Likely because of the relatively low number of patients we were able to analyse, we could not find any correlation between CEC or CD309 expression, or with any clinical parameter (data not shown).

Treating patients with IPF is a major medical problem [29]. Pirfenidone attenuated the fibrocyte pool size in bleomycin-treated mouse lungs via attenuation of CCL2 and CCL12 production in vivo, and fibrocyte migration was inhibited by pirfenidone in vitro [30]. Inhibition of these cells is considered a mechanism of the anti-fibrotic action of the drug [30], and indeed pirfenidone first showed clinical amelioration in patients with IPF [31].

Recently, nintedanib has shown beneficial effects in patients with IPF (clinical trials TOMORROW, INPULSIS 1 and INPULSIS 2) [32, 33]. Nintedanib was originally developed as an angiostatic factor for cancer treatments, and was approved to treat patients with lung cancer with advanced adenocarcinoma after first-line chemotherapy. Inhibition by nintedanib ultimately results in the reduced proliferation, migration and survival of fibroblasts, and potentially attenuates angiogenesis in the lung [34, 35]. Nintedanib has shown consistent anti-fibrotic and anti-inflammatory activities in bleomycin-induced pulmonary fibrosis in rodents [28, 36] and in human fibroblasts isolated from the lungs of patients with IPF, and inhibits FGF-induced, PDGF-induced, VEGF-induced profibrotic effects in human lung fibroblasts from patients with IPF [36‐39]. Accordingly, in eight patients taking nintedanib, we found significant changes in CEC levels and in numbers of CEC expressing CD309, as well as in collagen I+ cells (Fig. 6). The number of patients being treated with pirfenidone was too low to allow any statistical analysis, although a similar trend was found concerning CD309 expression (data not shown).