Authors’ reply

Dr Medford laudably highlights the interesting findings of other authors regarding levels of vascular endothelial growth factor (VEGF) in idiopathic interstitial pneumonia (IIP). Indeed, our data reflect the findings of Meyer et al who studied 11 patients with IIP.¹ We extend their observations in a larger cohort of patients (n = 49) and specifically relate plasma VEGF levels to disease progression and extent of fibrosis on HRCT scanning.² Indeed, HRCT scanning is perhaps the most reliable surrogate for the extent of disease.³ Like Meyer et al, we observed reduced median bronchoalveolar lavage (BAL) fluid levels of VEGF in patients with IIP (91 pg/ml) compared with controls (204 pg/ml). The reduction in the BAL fluid level of VEGF may reflect the absence of angiogenesis in that specific part of the lung, with the plasma VEGF level identifying a secondary phenomenon of compensatory angiogenesis in alternative areas of the lung. Alternatively, VEGF levels appear to be higher in epithelial surface fluid than in the serum, suggesting vectorial intraluminal secretion and the existence of a concentration gradient from air spaces to intravascular spaces.¹

Thickett et al⁴ are correct in their quotation of the normal range for VEGF in plasma (36–76 pg/ml) as measured by the R&D Systems Quantikine ELISA kit. They point out that this range is quoted by the kit manufacturers and is consistent with their own data and, indeed, with our data in other studies.^5–7 We did not use the Quantikine kit in this study but stated clearly that: “The ELISA capture and detection antibodies for assaying IL-8 and VEGF were selected paired reagents optimised for ELISA performance from R&D Systems”. R&D currently sell these paired reagents under the name “Duoset”.

Different ELISA formats for VEGF quantitation using recombinant VEGF165 as standard are available. Capture reagent: (1) rabbit polyclonal antiVEGF (in house); (2) soluble flt-1 (Sflt-1); (3) Quantikine kit, mouse anti-VEGF; (4) Duoset, mouse anti-VEGF. Detection reagent: (1) mouse anti-VEGF (Genetech 4.6.1);⁸ (2) rabbit anti-VEGF;⁵ (3) mouse anti-VEGF;^5,7,9 (4) mouse anti-VEGF.² Not surprisingly, each assay reports a different normal range. In our experience the Quantikine kit measures low (with up to a third of samples having undetectable levels) and the Duoset combination measures high, as we reported in the article (648 pg/ml). The other assays report intermediate values.

A number of possibilities exist as to why these assays read differently. It is unlikely to be due to platelet release as suggested.⁵ The greatest difference in VEGF levels that is detected between paired serum (complete platelet release of VEGF) and standard plasma samples (low platelet VEGF release) is at most only three to four fold. Similarly, the difference in VEGF between paired platelet poor and platelet rich plasma samples is of the same order of magnitude. To demonstrate this one has to spin plasma samples at 2700g for 15 min to prepare platelet poor samples. In our experience it is important to use plasma rather than serum samples to quantitate VEGF and to treat each study sample similarly in terms of centrifugation—whether it be 300g for 12 min, 1000g for 15 min, or 2700g for 15 min. This will minimise variation in the study samples due to the “platelet release effect”.

There are possible explanations for the VEGF immunoassays reading differently in plasma. It may be a combination of at least two effects—the nature of the epitope detected and the presence of other competing ligands in the sample. Different antibodies raised to VEGF will react variably with available epitopes on the ligand. This can be quite striking with monoclonal antibodies to different epitopes of a ligand when they are used as ELISA capture reagents. Assuming antibody A reacts with an epitope on VEGF that is close to or part of the flt-1 receptor binding site and the other antibody B reacts with an epitope well removed from this site, with an identical absolute amount of VEGF in the plasma sample, antibody A would read low or negative and antibody B would read high in relation to the amount of sflt-1 present in the plasma sample. We have in fact shown that the capture antibody used in the Quantikine VEGF ELISA is, indeed, sensitive to the presence of sflt-1.⁵

In addition to this potential variation in the level of free VEGF and VEGF-sflt-1 complexes in plasma samples, a further confounding species is the amount of placenta growth factor (PLGF). VEGF is a natural homodimer but it does form heterodimers with PLGF and we have detected such complexes.¹⁰ Antibodies detecting epitopes that are variably modulated by the binding of PLGF to VEGF will read low or high depending on the PLGF concentration. This focuses on the nature of the R&D Systems monoclonal antibodies to VEGF—one in the Quantikine kit and the other part of the Duoset. Following the recognition of a difference in performance of these two assays, we contacted R&D Systems for information concerning the nature of these antibodies. We were interested to know if the same antibody or different antibodies were used and what information was available on their specificity. The response from R&D Systems was that the capture antibodies were different, so the scenario outlined above is a possible explanation.

It is important to appreciate the difficulties in interpreting absolute levels of VEGF in complex media such as plasma. To do this rigorously one ought to quantitate not only free VEGF but also VEGF complexes with sflt-1, sKDR, and PLGF to understand how one immunoassay measures against another, and currently this is not possible. It would be simplistic to think that the Quantikine kit values are the “true” VEGF values and the Duoset assay values artefactual. It might simply reflect the fact that the Quantikine values are free VEGF and the Duoset values total VEGF (free VEGF plus VEGF complexes).

The important observation in our study is not the absolute VEGF plasma values but the relative differences in VEGF levels between patients and control samples over time, where the sampling issues have been fully appreciated and rigorously controlled to allow clinical interpretation of the results. We have emphasised the prognostic value of plasma VEGF in idiopathic pulmonary fibrosis and have shown a significant positive relationship between the HRCT fibrosis score and the plasma concentration of VEGF. A comparison with acute respiratory distress syndrome (ARDS) is not useful and perpetuates the concept that ARDS equates with chronic idiopathic pulmonary fibrosis, which is not the case.

In relation to the point about quantitating local VEGF concentrations in the lung where the influence of epithelial sflt-1 might be greater, it is to be expected that different assays could lead to a variation in reported VEGF levels for the reasons already discussed.

In conclusion, we have complete confidence in the validity and reproducibility of the VEGF data presented. In a situation of excess VEGF production which could potentially be driving an angiogenic fibrotic pathology in the lung, we suggest it is entirely appropriate to contemplate antagonising excess VEGF in order to control the disease pathophysiology. Overdosing on VEGF antagonist is clearly counterproductive and, as stated, could lead to the apoptosis of both epithelial and endothelial cells. Clearly, it is a case of restoring homeostasis.