Cell culture detection of microvascular cell death in clinical specimens of human neoplasms and peripheral blood


Larry M. Weisenthal, Weisenthal Cancer Group, Huntington Beach, CA 92647, USA.
(fax: +714 596 2110; e-mail: mail@weisenthal.org).


Background.  Angiogenesis studies are limited by the clinical relevance of laboratory model systems. We developed a new method for measuring dead microvascular (MV) cells in clinical tissue, fluid and blood specimens, and applied this system to make several potentially novel observations relating to cancer pharmacology.

Methods.  Dead MV cells tend to have a hyperchromatic, refractile quality, further enhanced during the process of staining with Fast Green and counterstaining with either haematoxylin–eosin or Wright–Giemsa. We used this system to quantify the relative degree of direct antitumour versus anti-MV effects of cisplatin, erlotinib, imatinib, sorafenib, sunitinib, gefitinib and bevacizumab.

Results.  Bevacizumab had striking anti-MV effects and minimal antitumour effects; cisplatin had striking antitumour effects and minimal anti-MV effects. The `nib' drugs had mixed antitumour and anti-MV effects. Anti-MV effects of erlotinib and gefitinib were equal to those of sunitinib and sorafenib. There was no detectable VEGF in culture medium without cells; tumour cells secreted copious VEGF, reduced to undetectable levels by bevacizumab, greatly reduced by cytotoxic levels of cisplatin + anguidine, and variably reduced by DMSO and/or ethanol. We observed anti-MV additivity between bevacizumab and other drugs on an individual patient basis. Peripheral blood specimens had numerous MV cells which were strikingly visualized for quantification with public domain image analysis software using bevacizumab essentially as an imaging reagent.

Conclusions.  This system could be adapted for simple, inexpensive and sensitive/specific detection of tissue and circulating MV cells in a variety of neoplastic and non-neoplastic conditions, and for drug development and individualized cancer treatment.


Angiogenesis is important in embryonic and fetal development, wound healing and tissue regeneration, and numerous disease states, including retinopathy and cancer. Even though many model systems have been developed, all of them have limitations for the study of the perturbation of angiogenesis by pharmaceuticals [1]. A surprisingly important role for angiogenesis in haematologic neoplasms (which are literally bathed in blood) has been more recently recognized and exploited therapeutically [2], yet the nature of the association between haematologic and vascular cells remains murky, and therapeutic exploitation of this relationship (particularly on an individual patient basis) has also been hindered by limitations in the available laboratory systems. Finally, it is increasingly being realized that circulating microvascular (MV) cells may be important markers for a wide variety of non-neoplastic disease states such as cardiac disease, cerebrovascular disease and diabetes [3]. However, there have been important technical limitations in the ability to recognize and quantify these cells as well [4]. We developed a novel system for testing anti-MV drug effects in fresh biopsy specimens of human tissue, cavitary fluids and blood.

Materials and methods

Patients, biopsy specimens, general cell culture methodology and slide preparation

Patients, tumour cell isolation from fresh human biopsy specimens, cell culture conditions, slide preparation, slide staining and scoring of drug effects against tumour cells are all described in detail in http://www.weisenthal.org/w_ovarian_cp.pdf

In brief, three-dimensional microclusters of tumour cells were isolated from fresh tumour biopsy specimens by scissor-mincing, collagenase–DNase digestion and differential sedimenation, with or without Ficoll-diatrizoate centrifugation. Microclusters were cultured for 96 h in anchorage-independent conditions (polypropylene, round-bottomed, 96-well microplates), in the presence and absence of test drugs, with both vehicle controls and 100% cytotoxic positive controls (30 times supramaximal concentration of cisplatin + anguidine). After cell culture, Fast Green (analogous to Trypan Blue, but more tenaciously staining) was added to the microwells. This stained non-viable cells, including tumour cells, inflammatory cells and connective tissue cells, and endothelial and other MV-associated cells. Microclusters were then sedimented onto microscopic slides using a cytospin centrifuge. Slides were counterstained with haematoxylin–eosin or Wright–Giemsa to allow identification of the surviving, viable cells, and to analyse if they were tumour cells and/or of the above-mentioned normal cell types.

Tumour cell-associated MV cells – scoring MV death

The senior author (LMW) developed and has been using a previously described novel dye exclusion assay [5, 6] to study the direct cytotoxic effects of anti-cancer pharmaceuticals since 1978 and has personally scored over 500,000 slides prepared from over 10,000 fresh human tumour cultures. In scoring several dozen slides from a highly drug-refractory carcinoid tumour, the author observed viable (haematoxylin–eosin pink staining) tumour cells, indicating that the neoplasm was refractory to the effects of more than 20 different drugs. However, one slide, prepared from cells cultured with bevacizumab, was found to have a strikingly different appearance (Fig. 1b, c). All non-bevicizumab slides contained viable tumour clusters featuring fairly uniform pink-staining membranes with a ‘pink pancake’ appearance. In contrast, viable tumour clusters on the bevacizumab slides displayed small, darker staining concentrations with a ‘blueberry pancake’ appearance (hereafter BB). Careful examination of cells in non-bevacizumab-exposed cultures showed the cell clusters to contain small, often angulated pink-red staining ‘features’ within the interstices of the cell aggregates. It was quite obvious that these were MV features, most probably endothelial cells and/or pericytes.

Figure 1.

 Appearance of pancreatic carcinoid tumour, following 96 h in suspension culture, in the absence and presence of bevacizumab, 2.5 mg mL−1. Left to right: panel (a) from vehicle (0.9% NaCl) control cells, stained with Fast Green, counterstained with hematoxylin-eosion, and again counterstained immunocytochemically with anti-CD31. Brown-orange staining microvascular complexes are evident. Panel (b) again from vehicle control cells, stained only with Fast Green/H&E. Both tumour cells and microvascular complexes (circled) are viable (pink staining, with the cytospin ‘flattened’ cell clusters having a ‘pink pancake’ appearance. Panel (c) Bevacizumab-exposed culture. Tumour cells remain perfectly viable (pink staining, whilst non-viable microvascular complexes first take up the blue-green non-vital stain (Fast Green), and then partially extrude the dye to the surrounding (viable tumour) cells during the alcohol-based H&E counterstaining. This produces the characteristic ‘blueberry pancake’ appearance of a tumour cell cluster exposed to an active anti-microvascular agent.

With the insights obtained from this one case, we re-examined 19 consecutive fresh tumour assays in which bevacizumab had been present in the culture medium. There was no indication of a reduction in tumour cell viability by bevacizumab (the observational end-point when the cultures were first scored) but, in a number of assays, there was an unmistakable ‘BB’ effect, albeit not as striking as in the index (carcinoid tumour) case, because there was confounding material (dead cells, debris, etc.) on these slides. The specific anti-MV effect of bevacizumab was recognized in the carcinoid specimen only because of the absence of debris and a complete absence of effects of any of the other drugs tested. The peculiar effects of bevacizumab were, in this context, striking and led us to realize that these were specific anti-MV effects. With these insights and with experience, the specific anti-MV ‘BB’ staining could readily be distinguished from other types of blue staining, such as that associated with dead tumour cells, dead normal (but non-MV) cells and debris.

Figure 1(a) shows the appearance of the control (saline vehicle alone) culture stained immunocytochemically for CD31, an endothelial cell antigen which has been reported not to be expressed by tumour cells themselves [7]. The distribution of CD31 staining (which we performed on a systematic basis, in more than 50 individual cases: additional examples shown at http://www.weisenthal.org/Weisenthal_ESCI.html, pp. 82–90) is quite similar to the distribution of the ‘BB’ effect within the same tumour cultures, although the ‘BB’ staining is often more extensive than that predicted from corresponding CD31 staining. It is probable that the ‘BB’ staining may allow more sensitive detection of MV cells than that obtained by CD31 staining in clinical specimens highly sensitive to bevacizumab and similar agents. It should be noted that in cases of striking anti-MV activity, it is possible to observe ‘free-floating’ BB features, ostensibly shed from the cell clusters. These free-floating, MV BBs may be readily distinguished from dead, non-MV cells and debris by the refractile and hyperchromatic staining qualities and sharply angulated features of the MV BBs.

Figure 2 illustrates a qualitative (0–5+) scoring system devised by us to gauge anti-MV effects. A ‘0’ effect would indicate the complete absence of detectable BBs, either within the cell clusters or as ‘free-floating’ BB features, ostensibly shed from the cell clusters. A ‘1+’ effect would indicate only rare BBs, usually confirmable only at higher (200×/400×) magnification. A ‘2+’ effect would indicate more frequent BBs, albeit usually only in small, loose cell aggregates and usually confirmable only at higher magnification. A ‘3+’ effect would indicate very frequent BBs, unmistakable even at lower (40×/100×) magnifications, but still observable mainly in the smaller aggregates and only occasionally at the periphery of the larger, tighter aggregates. A ‘4+’ effect would indicate the appearance of BBs in virtually all cell clusters, including at least at the periphery of the larger clusters. A ‘5+’ effect would indicate a veritable ‘blizzard’ of BB material, in the large clusters, in small clusters and diffusely showering the entire slide background with ‘free-floating’ BB features. With only a little experience and practice, this scoring system proved to be remarkably reproducible: we randomly pulled 15 pairs of slides from control and bevacizumab-exposed cultures originally scored between 1 and 5 months previously, and had 100% agreement between the results of original scoring and ‘blinded’ re-scoring in all 15 cases. We are certain that most workers experienced in scoring immunohistochemical slides would find scoring of these anti-MV slides to be more objective, less tedious and more reproducible. We also think that these BB features would readily lend themselves to objective scoring through automated image analysis systems.

Figure 2.

 Panels from a mesothelioma specimen exposed to several putative anti-microvascular agents. In all cases, there is no direct toxicity to the tumour cells, which remain red-pink staining (i.e. viable), whilst the associated microvascular cells are variably ‘sensitive’ to the effects of the different drugs tested. As described in the accompanying text, the scoring system for quantifying anti- microvascular effects depends on surveying the entire slide, but, in this figure, an attempt was made to illustrate the scoring system with only a single field view for each drug. As shown in this figure, anti-microvascular drug effects were scored on a 0+ to 5+ scale, with each individual score multiplied by 10 for purposes of the data analyses, resulting in a 0–50 scoring scale.

Measuring vascular endothelial growth factor (VEGF) in cell culture medium supernatant

After 4 days of culture in 96-well polypropylene plates, the plates were centrifuged and the supernatants withdrawn and tested with a commercial VEGF [enzyme-linked immunosorbent assay (ELISA)] kit (Quantikine,http://www.RnDSystems.com), according to the manufacturer’s instructions.


Relative anti-MV versus anti-tumour activities of cisplatin, bevacizumab, erlotinib, sorafenib, imatinib, gefitinib and sunitinib

Two hundred and fifty four short-term cultures derived from fresh human tumour specimens were scored simultaneously for direct antitumour and anti-MV effects of at least one drug. These represented a wide range of tumour histologies and anatomical sites. The diagnoses most represented were those pertaining to breast (50), non-small cell lung (45), ovary (26) and colon (19). The anti-vascular effects were scored as described above. The anti-tumour effects were scored as referenced in Materials and methods. As shown in Table 1, cisplatin had potent direct antitumour cell effects at concentrations which were almost entirely devoid of anti-MV affects. It is quite likely that the minimal anti-MV effects of cisplatin may have been secondary to reduced production of VEGF and/or other MV viability factors by tumour cells which were themselves being directly destroyed by cisplatin, based upon the observation that ‘positive control’ drugs which killed all the tumour cells markedly reduced VEGF secretion (Fig. 3). In contrast bevacizumab had potent anti-MV effects at concentrations which were largely lacking in antitumour cell effects. Reduced tumour cell viability was most often seen in situations in which larger tumour cell clusters appeared to fracture into small clusters in the presence of bevacizumab. It is possible that the tumour-infiltrating microvasculature may provide a scaffolding which plays a role in binding the tumour cell aggregates together. When this scaffolding is destroyed, cell–cell interactions may be disrupted which would have otherwise contributed to maintaining the viability of the tumour cells within the clusters. Further studies will be required to address the above speculations relating to the putative mechanisms of (comparatively minor) anti-MV effect of cisplatin and (comparatively minor) anti-tumour cell effects of bevacizumab. There was a steep dose–response relationship with bevacizumab: in 111 paired comparisons of bevacizumab tested at 2.5 and 1.25 mg mL−1, the mean BB score at 2.5 mg mL−1 was 2.3 (95% CI 2.0–2.6), whilst the mean BB score at 1.25 mg mL−1 was 1.1 (95% CI 0.89–1.4), despite that fact that both 2.5 and 1.25 mg mL−1 reduced VEGF to undetectable levels in the cell culture supernatant (Fig. 3).

Table 1.   Anti-vascular drug effects and anti-tumour drug effects in a variety of fresh human tumours (largely adenocarcinomas; see text, n = 26–51, for the range of agents tested)
 No. paired comparisonsAnti-Vascular Score (AVS) (avg)Vehicle Anti-Vascular Score (VAVS) (avg)P2 (paired comparison AVS vs. VAVS)AVS minus VAVS (avg)Anti-Tumour (AT) Score (avg) (AVS-VAVS)/ AT Ratio (avg)
  1. Slides were scored for anti-microvascular (MV) drug effects and the same slides were also scored for direct anti-tumour effects. Shown in columns from left to right for each drug are: (i) the number of paired comparisons between anti-MV drug activity and anti-tumour activity; (ii) the average (mean) anti-MV drug effect score (multiplied by 10) for the combination of the drug + drug vehicle; (iii) the average (mean) anti-MV score for the vehicle alone {either 0.9% NaCl [for bevacizumab (2.5 mg mL−1), cisplatin (3.3 μg mL−1), sunitinib (8.3 μg mL−1) and gefitinib (22.4 μg mL−1)] or 0.5% DMSO + 0.5% ethanol [for bevacizumab (which was tested both with and without DMSO/ethanol), erlotinib (89.4 μg mL−1), sorafenib (6.25 μg mL−1), and imatinib (12.5 μg mL−1)]}. In the case of DMSO/ethanol tested by itself, the anti-vascular score (AVS) for DMSO/ethanol was compared the 0.9% NaCl vehicle results; (iv) the two-sided significance level (paired t-test) for differences between drug plus vehicle versus vehicle alone; (v) the anti-vascular score attributable to the drug alone, obtained by subtracting the vehicle alone score from the drug + vehicle score; (vi) the anti-tumour score, which is the average (mean) per cent tumour cell death in the drug exposed cultures, determined as described in the above web reference; and (vii) the mean ratio of anti-vascular drug effect divided by direct antitumour drug effect, which provides an index of the relative direct antitumour activity compared to anti-vascular activity for each drugs. For expample, DMSO/ethanol is virtually ‘pure anti-vascular’ agent, whilst cisplatin is virtually a ‘pure antitumour’ agent, with the other drugs having both antitumour and anti-vascular activities, bevacizumab heavily skewed toward anti-vascular, and the remaining drugs more heavily skewed towards antitumour.

Bevacizumab 2.5 mg mL−15124.75<0.000119.712.91.53
DMSO 0.5% + EtOH 0.5%4012.73.5<0.00019.233.67
Bevacizumab + DMSO/EtOH3732.811.5<0.000121.3Not scoredN/A
Cisplatin 3.3 μg mL−1374.53.90.370.659.20.01
Sunitinib 8.35 μg mL−1269.54.60.0824.925.80.19
Gefitinib 22.35 μg mL−1328.33.80.00024.528.50.16
Erlotinib 89.4 μg mL−14516.2120.00144.222.50.19
Sorafenib 12.5 μg mL−13512.
Imatinib 12.5 μg mL−13613.911.10.032.823.60.12
Figure 3.

 Effect of drugs and solvents on concentration of VEGF in the culture medium supernatant following 4 days of continuous drug exposure. Following 4 days in culture in 96-well polypropylene plates, the plates were centrifuged and the supernatants withdrawn and tested with a commercial VEGF (ELISA) assay kit, as referenced in the Materials and methods. There was no detectable above background signal for 0.9% NaCl alone, for culture medium (RPMI-1640) alone, or for complete culture medium (including 5% horse serum and 15% fetal bovine serum) without cells. There was a small, but clearly detectable signal for cells cryopreserved in 10% DMSO and then thawed and washed for culture. This signal was abrogated by bevacizumab and by a cytotoxic ‘positive control’ (cisplatin 100 μg mL−1 and anguidine 1 μg mL−1) which killed 100% of the tumour cells over the 4 days in culture. In the case of four fresh (non-DMSO cryopreserved) solid tumour cultures (patients 3, 5 and 6) there was a robust VEGF signal, which was, again, completely abrogated by bevacizumab and markedly reduced by the cisplatin/anguidine cytotoxic ‘positive control’. In a single NHL specimen, there was no detectable above-background VEGF signal in control cultures of cells without drug (data not shown in this figure; but other data from this specimen are shown in Figs 5–7). In this figure, error bars (95% CI) are shown for triplicates; results without error bars are in most cases duplicates (including the cervical squamous case).

The (kinase-inhibiting) ‘nib’ drugs were of interest because of their mixed direct antitumour and anti-MV effects. Sunitinib and gefitinib had lesser anti-MV effects relative to antitumour effects than did erlotinib, sorafenib and imatinib. However, the latter three drugs were dissolved and dispensed in dimethylsulfoxide (DMSO)/ethanol, with a final concentration in the culture medium of 0.5% DMSO/0.5% ethanol. Cisplatin, bevacizumab, sunitinib and gefitinib were all dispensed in isotonic (0.9%) sodium chloride. DMSO/ethanol was found to have anti-MV effects in some specimens and to contribute to anti-MV effects of other agents. After subtracting out the effects of the respective drug vehicles (0.9% NaCl for bevacizumab, gefitinib and sunitinib; 0.5% DMSO/0.5% ethanol for bevacizumab, erlotinib, sorafenib and imatinib), it was notable that erlotinib and gefitinib had relative antivascular activities at least equal to those of sorafenib and sunitinib (these latter agents being so-called multi-targeted kinase inhibitors with putatively greater antivascular activity). In terms of anti-MV activity relative to anti-tumour activity, bevacizumab had a ratio approximately 10-fold greater than that of the ‘nib’ drugs which, in turn, had ratios approximately 10-fold greater than cisplatin (Table 1).

Additivity and synergy between anti-MV agents

Figure 4 shows an example of a striking additive or synergistic effect on the tumour microvasculature between bevacizumab and imatinib. Additivity between agents was highly variable between individual specimens, suggesting a potential advantage to individualizing drug selection as opposed to treating all patients with a fixed drug combination.

Figure 4.

 Illustration of antimicrovascular synergy in a specimen of poorly differentiated breast cancer. Erlotinib and imatinib had negligible activity when tested as single agents, whilst bevacizumab had only 1+ activity. However, both erlotiib/bevacizumab and imatinib/bevacizumab were at least additive (and probably synergistic in the case of imatinib). Shown are low-power (40×), intermediate-power (100×), and high-power (400×) views.

Anti-MV effects of DMSO and/or ethanol

When tested without drugs as a vehicle-alone control, the DMSO/ethanol combination (final concentration in the culture of 0.5% for each of the two solvents, although it is probable that ethanol largely evaporated, whilst DMSO persisted over the 4 days in culture) occasionally had surprisingly high anti-MV activity, although DMSO/ethanol never had clearly detectable direct antitumour activity in more than 150 clinical specimens tested. Perhaps more importantly, DMSO/ethanol clearly (and sometimes dramatically) added to the activity of bevacizumab in individual cases and in aggregate (Table 1). In further exploratory studies, we have found that this effect of zenhancing bevacizumab may extend down to concentrations which are clearly within the clinically achievable range (i.e. 0.125%).

Effects of bevacizumab and a 100% cytotoxic drug combination (high-dose cisplatin + anguidine) on VEGF levels in the culture medium

Figure 3 shows the raw data from a study wherein we examined the concentration of VEGF in the medium supernatants of cell cultures from a number of different clinical specimens. It can be seen that: (i) cell culture medium alone (with or without 15% fetal calf serum and 5% horse serum) contained no detectable VEGF; (ii) supernatant medium from negative control (saline vehicle + fresh tumour cells) cultures contained detectable VEGF; (iii) supernatant medium from a 100% cytotoxic drug combination (high concentration cisplatin + anguidine) + tumour cells contained markedly reduced levels of VEGF; and (iv) supernatant medium from tumour cells cultured with bevacizumab for 96 h contained no detectable VEGF. Supernatant medium from three solid tumour specimens thawed after cryopreservation in 10% DMSO had very low but detectable levels of VEGF, and DMSO and/or ethanol in concentrations as low as 0.125% often reduced or abrogated VEGF in the supernatant (data not shown). These preliminary studies suggest inhibition of VEGF secretion from tumour cells as one possible mechanism of antivascular action for DMSO/ethanol. A single specimen of non-Hodgkin’s lymphoma (NHL) did not produce detectable VEGF in the supernatant, despite the fact that bevacizumab had a marked anti-MV effect in this specimen (Figs 5–7). We did not attempt to measure possible drug effects on VEGF receptor proteins.

Figure 5.

 Use of public domain ImageJ software (http://www.rsb.info.nih.gov/ij/) to quantify non-viable endothelial cells in non-Hodgkin’s lymphoma cells obtained from a lymph node biopsy. The top panel shows a 200× view of cells cultured 4 days with bevacizumab. Numerous densely refractile putative dead endothelial cells are present. Bottom panel, appearance of the identical field of cells after the image had been converted to 8-bit black and white and threshold gated to ‘paint’ only the dead putative endothelial cells. Only the painted features within the yellow circle were counted and measured by the software. Our (50 year old) American Optical microscope provided uniform illlumination only in a portion of each field, which resulted in detection artefacts, unless the area to be measured consisted only of the portion of each field in which uniform illumination was present.

Figure 6.

 Drug effects on the viability of microvascular (putative endothelial) and tumour cells persent in the same NHL specimen shown in Fig. 5. Shown are objective, automated ImageJ scores for both individual gated features and also for total area (originally in pixels squared). The raw scores for the gated features (putative dead endothelial cells) are normalized for uneven cell distributions by image analysing each area twice: first with the detection threshold set to gate only the features of interest (putative dead endothelial cells) and then again to detect all features on the slide (this would be analogous to normalizing gene copy to total DNA in cell extracts). Shown additionally are the subjective (0+ to 5+) manual anti-vascular scores which were obtained one month before we installed the ImageJ software. Shown additionally are the direct antitumour cell effects of each of the drugs and solvents tested. Three traditional cytotoxic drugs did not produce endothelial cell death above background levels, in contradistinction to bevacizumab, DMSO/ethanol and several ‘nib’ drugs. The units and magnitude of each of the endpoints are different, i.e. (1) number of individual features, (2) pixels squared, (3) 0 to 5 + BB effect, and (4) 0–100% of control tumour cell survival. In order to compare relative drug effects with these differing units of measurement, all results were normalized to a relative scale of 0–100, where zero is no effect and 100 it the maximal effect observed with any of the drugs tested. For example, 0 to 5+ scale would have units of 0+, 1+, 2+, 3+, 4+, 5+, and these would be converted for purposes of normalization to 0, 20, 40, 60, 80 and 100.

Figure 7.

 Dose–response relationship for effects of DMSO and/or ethanol on endothelial cell viability in the same NHL specimen shown in Figs 5 and 6, as determined by ImageJ software. Both ethanol and DMSO had anti-endothelial effects, which appeared to extend into the clinically achievable range (0.125%).

Effect of bevacizumab on neoplastic cells and normal cells derived from lymph nodes, peripheral blood and cavitary effusion fluid

Bevacizumab (as well as the ‘nib’ drugs) occasionally had detectable anti-MV effects when tested against cell aggregates derived from cavitary effusions of pleural and peritoneal fluid. On average, however, bevacizumab produced much lesser anti-MV effects in cell aggregates derived from malignant effusions than against cells derived from solid tumours, most probably because the cell clusters derived from malignant effusions contained fewer MV cells than in the case of cell clusters derived from solid tumours. The mean anti-MV (BB) score was 2.3 (95% CI 2.1–2.5, n = 205) in bevacizumab-exposed specimens derived from solid tumours, compared with only 0.5 (95% CI 0.2–0.8, n = 31) in specimens derived from malignant effusions (P2 < 0.0001). In contrast, peripheral blood, bone marrow and lymph node specimens showed striking bevacizumab effects on otherwise occult MV cells in these haematogenous sites (e.g. Fig. 8a,b). When corresponding control cultures were examined carefully, it was found that they contained large numbers of very small, bland, nondescript features in association with the peripheral blood and lymph node cells (Fig. 8a, circled red). Of note, the senior author (LMW) has examined slides of hundreds of leukaemia specimens since the 1970s and had never previously appreciated that these features were anything more than non-specific debris and/or platelets. But by incubating peripheral blood cell cultures with bevacizumab and then staining with Fast Green and counterstaining with Wright–Giemsa, the peripheral blood MV cells (often present in close association with leukaemia cells) ‘light up’ in a fashion analogous to bulbs against the backdrop of a bland Christmas tree in peripheral blood specimens (Fig. 8b, circled black). In addition to non-viable MV BBs, peripheral blood specimens exposed to bevacizumab showed numerous small, dark blue-black staining, refractile donut-shaped structures (not shown), which probably represented platelets, as these latter structures were markedly reduced in number in specimens which had been depleted of platelets through defibrination (data not shown). Platelets are known to contain copious VEGF, and bevacizumab has been shown to bind to VEGF present in platelets [8]. The dark blue-black staining implies disruption of the platelet membranes induced by the bevacizumab/VEGF interaction.

Figure 8.

 (a,b) Appearance of peripheral blood chronic lymphocytic leukaemia cells, following 4 days in culture in the absence (a) and presence (b) of bevacizumab.

Use of public domain image analysis software to quantify drug effects on putative MV cells in lymph node and peripheral blood specimens

Public domain ImageJ image analysis software, downloaded from the National Institutes of Health website, was used to quantify the anti-MV effects of pharmaceuticals on lymph node and blood specimens from patients with haematogenous neoplasms and from normal donors. Peripheral blood and lymph node specimens for discohesive cell types present fewer challenges for image analysis than do solid tumour biopsy specimens, although the authors are confident that more sophisticated image analysis systems could be applied to solid tumour biopsies as well. The imaging principle is that the densely refractile appearance of dead MV cells in our system may selectively and easily be gated by adjusting detection thresholds to highlight only the objects of interest. The ImageJ software is simple and easy to use for this application, and it is easy to program ‘macros’ to entirely automate the batch processing of hundreds of images in a very short period of time.

We tested the effects of several ‘standard’ cytotoxic agents, bevacizumab, and the ‘nib’ drugs, as well as effects of DMSO and ethanol in a lymph node specimen from a patient with follicular cleaved cell NHL (Figs 5–7). Cisplatin, fludarabine and doxorubicin had relatively strong direct cytotoxicity directed towards the NHL cells, but no detectable effects above background against the MV cells. The ‘nib’ drugs had variable direct antineoplastic cell toxicity, and strong anti-MV effects, whilst bevacizumab had no detectable direct antineoplastic cell activity, but a strong anti-MV effect (Fig. 6). DMSO and/or ethanol produced absolutely no direct cytotoxicity against the NHL cells, but produced moderately strong anti-MV activity which appeared to be present down to a clinically achievable range (0.125% for both DMSO and ethanol, Fig. 7). The NHL specimen tested in Figs 5–7 did not produce detectable VEGF in the culture medium (data not shown).


The system described herein appears to offer advantages over previously described systems for examining MV cells from clinical solid tumour, malignant effusion and blood specimens. In addition to describing the system, we have also reported exploratory results to illustrate the range of problems which could be addressed through the use of this assay system, which we refer to as the microvascular viability (MVV) assay. We believe that the system is sufficiently straightforward and economical to be utilized by a large variety of research laboratories and even in some sophisticated clinical laboratories. The equipment used in this study cost less than $10 000, and included a standard, 50-year-old, double-headed binocular microscope, a $1000 photomicroscopy camera, a standard laptop computer, public domain image analysis software, a cytospin centrifuge with commercially available multi-spot slide cuvettes [TEST Laboratories Ltd's Octospot, Thermo Shandon (http://www.thermo.com/com/cda/product/detail/1,1055,21115,00.html)], and standard tissue culture equipment.

Whether the densely blue-black (BB) features truly represent dead MV structures is a matter of debate. Identification as such is based primarily on morphology at the light microscope level, supplemented by immunocytochemical (anti-CD31) staining, reinforced by the observation that VEGF-depleting concentrations of bevacizumab elicited the appearance of these BB features in the absence of toxicity to other (tumour, inflammatory and non-MV connective tissue) cells, whilst cytotoxic agents (e.g. cisplatin) typically did not elicit these BB features at concentrations which were toxic to non-MV cells. We are confident that our findings will be reproducible and that these BB features will be confirmed to be of MV origin.

One interesting observation was that bevacizumab 2.5 mg mL−1 was typically much more active (with respect to producing non-viable MV BB features) than was bevacizumab 1.25 mg mL−1, despite the fact that both concentrations reduced VEGF concentrations in the supernatant medium to undetectable levels in the ELISA assay. Likewise, bevacizumab produced non-viable MV BB features in an NHL specimen, despite the lack of detectable VEGF in control cultures. Taken together, and in the context of the literature [9], these results strongly suggest that there is a threshold level of VEGF which is required for maintenance of MV viability in vitro, which is below the detection limits of the ELISA assay.

In addition to the findings described above and summarized in the abstract, we would like to suggest two additional, potentially important applications. First, the finding of myriad MV cells in association with leukaemia cells provides a unique model system to study not only the biology of haematologic neoplasms but also the activity of putative MV agents in a convenient and practical human system. One could simply obtain peripheral blood specimens from patients with chronic lymphocytic leukaemia (CLL) and culture them in the presence and absence of bevacizumab and other putative anti-MV and direct antitumour agents and simultaneously look for activity, synergy and antagonism between different agents, both against the MV cells and against leukaemia cells. In this latter regard, we would like to call attention to what is almost certainly an interpretive error in a recent study of the effects of bevacizumab against CLL cells obtained from a clinical specimen [10]. The authors of this study cultured CLL cells against the same concentration range as used in the present study, using annexin as the cell death reagent, with quantification by flow cytometry. These authors claimed to have detected a direct cytotoxic effect of bevacizumab against CLL cells based on this assay, but in our opinion, based on our experience as illustrated by the results presented here (Figs 5–8), it is most likely that what their annexin/flow cytometry assay detected was instead the death of MV cells and not of CLL cells.

Quite different from the many obvious applications of this (MVV assay) technology in cancer research, we think that the MVV assay could be applied advantageously to studying the effect of pharmaceuticals on the microvasculature present in normal tissues as well, including specimens of normal bone marrow, lung, breast, lymph node and retina. Finally, it should be noted that the field of endothelial cells and endothelial microparticles circulating in the peripheral blood has become of increasing importance in cardiovascular disease, neurology (stroke) and diabetes research. However, existing methodologies for identifying and quantifying endothelial cells in the peripheral blood (chiefly anti-CD31 staining, as detected by flow cytometry) have not proven to be ideal or reliable [4] and require sophisticated instrumentation. We think it should be feasible to draw a sample of peripheral blood, isolate the platelet-depleted mononuclear cell fraction using readily available methods, and utilize bevacizumab (and/or related drugs) as a ‘visualization reagent’ for the circulating MV cells. Short-term culture with bevacizumab will deplete the culture medium of VEGF; the MV cells alone will die specifically as a result of VEGF depletion, and these MV cells can then be readily visualized and quantified, following the application of our staining procedure, using public domain image analysis software.

Perhaps the most intriguing findings were the anti-MV effects of DMSO and ethanol, including significant enhancement of the anti-MV activity of bevacizumab and other agents (Table 1), most plausibly through pertubation by DMSO/ethanol of VEGF secretion and/or VEGF receptors. DMSO was previously reported to have anti-proliferative effects against bovine endothelial cells [11]. Ethanol is well known to retard wound healing, perhaps mediated through an antivascular effect [12]. Most intriguing is a case report from the University of Heidelberg, wherein a breast cancer patient with a daily wine consumption of 1.5 L continues to enjoy (M. H. R. Eichbaum, personal communication) a remarkable 5 year plus remission on therapy with trastuzumab, an agent also known to inhibit angiogenesis in vivo [13, 14]. The therapeutic potential of DMSO and/or ethanol as an adjunct to anti-MV cancer treatment should be considered.

In summary, we feel that we have described a novel method which can be successfully applied to the study of the biology and pharmacology of human microvascularity.

Conflict of interest statement

No conflict of interest was declared.