Expansion of Microvascular Bed and Increased Solute Flux in Human Basal Cell Carcinoma in Vivo, Measured by Fluorescein Video Angiography¹ Free

Studies of microvascular perfusion in cancers in vivo are necessarily based chiefly on animal models (e.g., Refs. 1, 2, 3). Such studies, supported by tissue histology, evidence of VEGF³ up-regulation, and the reduction of tumor growth by angiostatic agents, support the view that angiogenesis is a crucial requirement for tumor nutrition, growth, and in some cases metastasis (Refs. 4, 5, 6). Endogenous antiangiogenic factors such as angiostatins are overwhelmed by proangiogenic secretions such as VEGF, platelet-derived growth factors, and others, leading to a chaotic overgrowth of tortuous, dilated, abnormal exchange vessels (7, 8, 9). The new vessels are essential for the delivery of O₂ and nutrients to the growing tumor and waste product removal. They may also exert a paracrine influence on the tumor (10) and facilitate blood-born metastasis (11). From a therapeutic viewpoint, the new vessels also provide a gateway for drug permeation into the tumor, so studies of endothelial solute permeation in tumors provide important insights into drug access (2, 9). Despite advances in therapeutic antibodies, gene therapy and liposomal drug delivery systems, most current chemotherapeutic drugs are small, rapidly diffusing solutes of <2000 daltons.

Microvascular changes in human tumors are, for obvious reasons, difficult to study directly in vivo, and BCC of skin offers a unique opportunity in this respect (12, 13). Human BCC accounts for ∼70% of skin malignancies and its incidence is increasing (14). Because BCC is only slowly invasive and rarely metastasizes, there is a temporal window for microvascular investigation before treatment. The tumor commonly presents as an opalescent plaque (but can be pigmented, cystic, or ulcerative), often in sun-exposed sites such as the head, neck, and forearms in fair-skinned, elderly subjects. The tumor is composed of islands of mitotic epithelioid cells in a proliferating connective tissue stroma with lymphocytes, chronic inflammatory components, and many microvessels. Expression of the vascular permeability factor VEGF is relatively weak in BCCs compared with, for example, squamous cell carcinomas (15, 16, 17). A high rate of apoptosis may contribute to the slow growth of the tumor. The stromal microvasculature confers a telangiectatic appearance, increases blood flow ∼2-fold (18), and may influence tumor aggressiveness (19).

Biopsy studies show that tumor microvessels are abnormal, with increased diameters (dilated), a fragmented basal lamina, and sometimes fenestrations (20). Tumor microvessels in animal models have an increased permeability to macromolecules (e.g., Refs. 1, 4, 9, 21), due partly to the action of VEGF. VEGF also increases endothelial hydraulic conductance (22), contributing to the high tumor interstitial fluid pressures, which in turn limits convective delivery of therapeutic macromolecules to tumor cells (2). Raised macromolecular permeability results in a fibrin-rich interstitial matrix that supports angiogenesis (23).

A minimally invasive, low-cost method to quantify angiogenesis and solute exchange in an easily accessible human tumor in vivo would provide pathologists, pharmacologists, and clinicians with a useful investigative tool, e.g., to evaluate the efficiency of angiogenesis inhibitors in human cancer. The only current alternative is high technology and very expensive, namely magnetic resonance scans for tagged macromolecular complexes, with attendant questions of signal interpretation and quantification (9). The aim of this study was to assess the use of FA in human BCC to quantify angiogenesis and small solute exchange kinetics.

Human FA is a relatively simple procedure pioneered by Bollinger et al. (24, 25). FA can be used either with a small detection window to study single capillary permeability (e.g., Refs. 1, 26, 27) or with a large window that records the averaged solute flux from whole network of microvessels (25, 28, 29). This study uses large window FA to test the hypothesis that angiogenesis in human BCC is associated with increased small solute dynamics. FA has been used to demonstrate cutaneous angiogenesis in human breast cancer-related lymphoedema (30). This article also assesses the view that measurements of large window fluorescence over several minutes is a marker of vascular permeability (29, 31). We show here that large window FA enables microvascular density and exchange rate constants to be evaluated but that numerous factors confound total field fluorescence as a permeability index for a small solute.

Images of the BCC and control skin at an equivalent site on the opposite side of the body were recorded by fluorescence videomicroscopy for up to 30 min after a rapid i.v. injection of sodium fluorescein. Local red cell flux was recorded by laser Doppler fluximetry. The initial images of filled microvessels, before obscuration by fluorescein diffusion into the interstitium, were analyzed for area fraction, length density, and width. Accumulation of fluorescein in the tissue and its subsequent washout were charted in later fields.

Fifteen Caucasian patients (10 male and 5 female; mean age, 71 ± 16 years) were recruited from the Dermatology Outpatient Clinic, St. George’s Hospital (London, United Kingdom). The study was approved by St. George’s Hospital Ethics Committee, and written informed consent was obtained from each subject. The diagnosis of BCC was confirmed later by histology after therapeutic excision of the lesion. The BCC was located on the forehead or temple in 11 cases and on the upper limb in the remainder. BCC diameter was 7–20 mm, and reported duration was 3–36 months (mean, 19.6 ± 11.6 months). In addition, control studies were performed on 2 healthy male subjects ages 24 and 41 years.

A region of interest of area 4.25 mm² was examined through a Wild-Leitz video-modified epi-illuminated DM LB microscope (Leica U. K. Ltd., Milton Keynes, United Kingdom) using a 100-W mercury vapor lamp and a plan fluotar 2.5/0.08 objective. Illumination intensity in lux was checked before each study using a Minolta TL-1 Illuminance Meter (Minolta Camera Co., Japan). The microscope image was recorded by an analogue monochrome Hitachi KP161 videocamera with automatic gain switched off and stored on Konica SE120 super-VHS videotape by a Panasonic AG 7350 videorecorder (Siel Imaging, Aldermaston, United Kingdom). The images were displayed continuously on monochrome Sony monitors during the study. Images for offline morphometric analysis were printed at a final magnification of ×77.3 from the videotape using a Sony Multiscan UP-930 videoprinter.

Skin temperature was monitored by a YSI telethermometer (Henleys Medical, Welwyn Garden City, United Kingdom). Local red cell flux was measured in AU with a MBF3D laser Doppler flowmeter and P1 probe (Moor Instruments, Axminster, United Kingdom; 100 AU ≡ 2.5 V of flowmeter output).

The BCC and control site were studied on separate days. Patients habituated to the temperature-regulated laboratory (25.0°C ± 0.5°C) on the examination couch for 30 min. The head or limb was stabilized using a vacuum pillow support and the microscope positioned vertically above the site. Laser Doppler red cell flux was measured in four positions (1 mm apart, total sampling area ∼4 mm²) by rotating the probe head in its circular holder to obtain an averaged signal. The skin temperature probe was taped adjacent to the site. BCCs on the head were located 14.5 ± 7.6 cm and those on the arm 2.4 cm above the manubriosternal angle.

The study site was trimmed of hairs and coated lightly with mineral oil to improve epidermal transparency and reduce reflective glare. A 1-mm² orange cardboard square on the edge of the microscope field provided a reference point. Adjacent fields were first inspected and recorded using white light epi-illumination (native capillaroscopy), with a green filter to enhance vessel contrast. A 4.25-mm² field for FA was then selected randomly for study (25). Fluorescein, a hydrophilic solute of 376 daltons (disodium salt) and diffusion radius of 0.45 nm, is a safe agent for repeated use in all age groups (32). Fluorescein at the plasma concentrations used here (∼7 × 10⁻⁶ g/ml) is 15–55% free (33, 34, 35) with the remainder bound loosely and reversibly to plasma proteins. The unbound fraction is almost constant at 10⁻⁶–10⁻⁴ g/ml because the bound fluorescein readily dissociates [dissociation constant ∼6 × 10⁻⁴ m, three to four binding sites/albumin molecule (35, 36)]. Fluorescein is excreted into the urine via glomerular filtration (32, 33). FA brightly highlights all of the plasma-perfused microvessels and detects 19–27% more vessels than native capillaroscopy (37). The fluorescein bolus comprised 0.2 ml of sodium fluorescein solution (20 g/100 ml, 0.53 m; Martindale Pharmaceuticals, Romford, United Kingdom)/liter of blood. Blood volume was estimated from the patient’s weight, age, and gender using a nomogram (38). The fluorescein was injected rapidly (<1 s) through the contralateral antecubital vein at a recorded time. The field was then observed continuously for 30 min, using a blue excitation filter (wavelength 450–500 nm) and recording the emitted green fluorescence (wavelength 515–600 nm). The laser Doppler signal did not change after the injection indicating that the fluorescein had no vasoactive effect, and there were no adverse reactions (nausea, vomiting, or allergic reactions).

Vessel density was evaluated using videoprints of early images before fluorescein had flooded the tissue or in some cases from negative images from the clearance phase where plasma-filled vessels from dark images against a fluorescent background. Control studies showed that the morphometric results were not significantly different for positive and negative images from the same field.

The fraction of the field area occupied by vessels was determined by point counting (39). A square array of 1064 dots at 5-mm intervals on an acetate sheet was randomly placed over the 186 × 138 mm videoprint. A_A was calculated as number of dots falling on vessels divided by total dots on the image. The mean result from three random placements was calculated. With ∼15% dots falling on vessels, this method has a theoretical SE of 4% (39).

The line intersection method of Weibel (40) was used to determine the total length of blood vessels/unit area of tissue, L_A. An acetate sheet ruled with a grid of parallel lines at 5-mm intervals was placed randomly over the videoprint. The total intersections between blood vessels and lines were divided by cumulative line length in centimeters. The result was divided by the magnification factor (77.3) and multiplied by the stereological correction factor π/2 to estimate L_A in units of cm vessel/cm² field (cm⁻¹).

A line was placed randomly on the videoprint 10 times to select 10 vessels. The random vessel width was measured by a ruler (30). Also, the width of the widest (W_max) and narrowest vessel (W_min) in each videoprint was measured.

Microscope light intensity was 332 ± 33 lux during the BCC study and 319 ± 18 lux during the control study. As the fluorescein bolus passed through the microcirculation, fluorescein rapidly diffused into the interstitium and FI of the field increased progressively. Later, as plasma fluorescein concentration declined, interstitial fluorescein diffused back into the circulation, and FI declined progressively. FI per unit area of the field of interest, inclusive of tissue and blood vessels, was quantified by digitizing selected videotape images onto computer using an ELVIS PCI card (Vision Dynamics, Hemel Hempstead, United Kingdom) and analyzing the resulting 256 × 256 pixel ASCII images using the computer program Caproc developed in-house. The position of the orange card was used to align images and adjust for cardiorespiratory movements. Such movements reduced the area of overlap of the sequence of images, and analysis was performed finally over a total area of 3.6 mm², made up of those pixels common to every image. Each pixel was 7.4 × 7.4 μm. The pixel GLs, on a scale of 0–255, were summed and divided by the total number of pixels to give FI in units of GL. FI was determined every second for ∼120 s after injection, then at wider intervals for 30 min. Preinjection background FI, usually 44–46 GL, was subtracted from the result to give the reported FI values.

After a rapid bolus injection of fluorescein solution into an antecubital vein, arterial blood was sampled serially from the opposite brachial artery of a healthy man at ∼1-s intervals for 100 s using a fraction collector. The heparinized samples were centrifuged, and the plasma analyzed for fluorescence level using a spectrofluorimeter [mpf-44a; Perkin-Elmer (U. K.) Ltd., Wokingham, United Kingdom].

To assess the long-term decay of plasma concentration, venous blood was sampled from the antecubital vein of a healthy man at 10-s–5-min intervals over 30 min after an i.v. fluorescein injection into the opposite arm. Plasma fluorescein concentration was measured using a Fluor-S multi-imager and Quantity One Quantitation Software (Bio-Rad Labs Ltd., Hemel Hempstead, United Kingdom).

Several technical issues were investigated as follows.

(a) The linearity of the relation between FI and the number of fluorescein molecules in the field of view (mass) was examined in vitro. A 100-μm thick layer of fluorescein solution of various concentrations in a hemocytometer chamber overlying a black card was excited by blue incident light for 30 s. FI was measured using the videomicroscope and Caproc. Fluorescein mass in the field was also varied at a fixed concentration by progressively occluding part of the field with an opaque black card.

(b) The effect of focus on FI was assessed because differences in vessel depth and slight tissue movement in vivo made it impossible to keep all vessels sharply in focus throughout the recording. The microscope was first focused on the underside of the hemocytometer coverslip and FI measured. The microscope was then raised by 1–7 mm or lowered 1 mm, without refocusing, and FI remeasured.

(c) Fluorescence attenuation during continuous illumination (photobleaching) was evaluated in vitro. Fluorescein in a hemocytometer chamber was exposed for 30 min to blue excitation light at the intensity used in vivo. Images captured at 5-min intervals were analyzed using Caproc.

(d) Fluorescence quenching by red cells was measured in a hemocytometer chamber containing fluorescein in diluted, heparinized human blood. Hemoglobin absorbs strongly at 480 and 515 nm, the excitation and emission maxima of fluorescein. The hematocrits studied were 24.5% (approximate dynamic hematocrit in microcirculation), 3.9% (dynamic hematocrit × typical vessel area fraction A_A in BCC), and 0.7% (dynamic hematocrit × vessel area fraction A_A in control site).

Results are presented as mean ± SD (or SE where specified). Student’s t test was used for paired comparisons, or Wilcoxon’s matched pairs signed ranks test where results were skewed (e.g., Fig. 1, B and C). Line fits and correlations were determined by linear regression analysis and regression slopes compared by analysis of covariance as implemented in Graphpad Prism (San Diego, CA). Significance was accepted at P < 0.05.

Skin temperatures adjacent to BCCs, 34.2°C ± 1.4°C, and control sites, 34.2°C ± 0.6°C, were the same. Arterial blood pressure and heart rate were 144 ± 16/86 ± 6 mmHg and 74 ± 7 min⁻¹, respectively, in BCC patients and 144 ± 16/86 ± 9 mmHg and 71 ± 7 min⁻¹ in the controls. Native capillaroscopy confirmed a more chaotic pattern of vessel in BCC than control skin, with the numerous branched, dilated-looking, sometimes tortuous vessels (12).

Mean A_A was 2.6-fold higher in BCCs, namely 0.194 ± 0.058, than in controls, 0.074 ± 0.025 (P < 0.001, Wilcoxon test, n = 14; Fig. 1 A).

Mean L_A was twice as great in BCCs, 66.0 ± 20.4 cm⁻¹, than in controls, 33.0 ± 19.4 cm⁻¹ (P < 0.001, Wilcoxon test, n = 14; Fig. 1 B).

The mean width of randomly selected vessels (10/site) was 2.1 times greater in BCCs, 40.9 ± 16.2 μm, than in controls, 19.5 ± 4.5 μm (P = 0.0003, Wilcoxon test, n = 14; Fig. 1 C). The diameter of the widest vessel image in each site, W_max, was substantially greater in BCCs (97.5 ± 40.6 μm) than controls (38.8 ± 11.2 μm; P = 0.0001). The diameter of the narrowest vessel image in each site, W_min, was also significantly greater in BCCs (16.0 ± 4.6 μm) than controls (11.6 ± 2.9 μm; P = 0.003).

The BCC red cell flux, 268 ± 144 AU, was >3-fold bigger than control, 69 ± 41 AU (P < 0.01, paired t test, n = 10; Fig. 2 A). Control red cell flux in the upper limb (10 and 15 AU, n = 2) was substantially less than on the head (range, 43–116 AU), in keeping with the higher cutaneous microvascular density on the head (41). By contrast, BCC red cell flux in the upper limb (298 and 389 AU) was among the highest observed, being 20- and 39-fold increases over the matched control site. For head BCCs, the mean increase was 3.4-fold.

The interval from i.v. injection to arrival, t_appearance was recorded as the time at which field FI began to increase. The t_appearance was shorter in BCC, 20.0 ± 3.3 s, than controls, 23.7 ± 4.7 s (P < 0.01, paired t test, n = 15; Fig. 2,B). Laser Doppler red cell flux correlated negatively with t_appearance; in other words, fluorescein appeared earliest in the regions with the highest red cell fluxes (correlation coefficient r² = 0.32, P < 0.05; Fig. 2 C). As might be predicted from distance considerations, control t_appearance in the head, 22.2 ± 1.3 s (n = 11), was earlier than in the arm, 28.0 ± 1.9 s (n = 4; P = 0.03, unpaired t test). In the hand, t_appearance is even later, 34.6 ± 7.2 s (24), and in the ankle, it is ∼54 s (29).

Three phases were recognized in the FI versus time plots (Fig. 3, A and B). Phase I, extending from t_appearance to 10–20 s, was the initial, very steep increase in FI. Phase II was the ensuing slower rise in FI to a maximum at ∼4–11 min. In Phase III, FI declined slowly.

Inspection showed that Phase I was due partly to the rapid filling of the plasma compartment. The time from the arrival of the fluorescein wave-front to its peak concentration was at least 9 s [control studies in vivo (arterial results) are below and Fig. 7 A]. Vessels did not all fill at the same instant because their anatomy and transit times varied. Frame-by-frame inspection indicated that all BCC vessels had filled with fluorescein by ∼6 s from t_appearance (time of appearance in first vessel to fluoresce) and that vessels began to clear (i.e., plasma fluorescence fell as the initial concentrated bolus passed) at ∼14 s from t_appearance. Phase I was not due solely, however, to intravascular fluorescence; extravascular accumulation too was visible within a few seconds of fluorescein arrival. Similarly, in the perfused human nailfold, transcapillary fluorescein diffusion occurs within 5–10 s (24). Because extravascular fluorescein increased rapidly, there was generally no dip in FI as the plasma concentration fell to its relatively stable lower level upon recirculation and mixing.

The Phase I slope d(FI)/dt was much steeper in BCC than controls (Fig. 3). The regression slope for the first 10 s in BCC, 4.27 ± 2.52 GL s⁻¹, was 2.9-fold steeper than control, 1.49 ± 1.05 GL s⁻¹ (P < 0.001, paired t test, n = 15). There was a significant correlation between d(FI)/dt and local red cell flux (r² = 0.54, P < 0.001, n = 18; Fig. 4,A) or vascular area fraction A_A (r² = 0.42, P < 0.001, regression analysis, n = 28; Fig. 4 B). Also, the 3.6-fold increase in BCC d(FI)/dt relative to control was almost identical to the 3.9-fold increase in red cell flux.

Inspection showed that fluorescein continued to accumulate and diffuse progressively through the tissue in Phase II. The continuing rise in FI (Fig. 3, A and B) indicated a concentration gradient from plasma to interstitium, although plasma concentration had fallen to its recirculated, mixed level (arterial results are below and Fig. 7,A). Fluorescence was relatively uniform by the time it reached its maximum value FI_max, except for filling defects occupied by hair follicles. FI_max was reached earlier in BCC (180 ± 193 s) than controls (687 ± 324 s; P < 0.001, paired t test, n = 15). The control value was similar to the mean of 704 s in normal ankle skin (29). The interval between t_appearance and FI_max correlated negatively with microvascular area fraction (r² = 0.42, P < 0.001, linear regression analysis, n = 28 BCC and control sites), indicating that a high vessel/tissue volume ratio promotes rapid interstitial equilibration. The magnitude of FI_max was on average no greater in BCC (55.7 ± 19.4 GL) than controls (63.4 ± 26.5 GL; P = 0.29, paired t test, n = 15) despite the greater vascularity of BCC (Fig. 3 B).

Phase II was more variable in BCC than controls. Control sites always had a distinct Phase II (Fig. 3). In 8 of 15 cases, BCC likewise had a clear Phase II (Fig. 3, A and B), but in 7 of 15 cases, the BCC curve reached FI_max so early that a Phase I-Phase II distinction was not possible (Fig. 3 C). Tumors in the latter group had a 2–3-fold higher red cell flux, 372 ± 39 AU, than tumors with a clear Phase II, 165 ± 50 AU (P = 0.01, unpaired t test).

Because BCCs reached FI_max earlier than controls, a point was reached where a rising control curve intersected a declining BCC curve, after which, control FI exceeded tumor FI. A similar crossover has been described between controls and diabetic skin FI curves (29).

Although it is intuitively tempting to equate FI after a given time interval with microvascular permeability, the transport equations show that many additional factors influence FI (“Appendix”). To evaluate a definable exchange parameter such as plasma clearance, it is necessary to consider the decay of slope d(FI)/dt as interstitial concentration rises and the gradient for diffusion falls (equations 6 and 7, “Appendix”). Because slope analysis requires a well-defined Phase II, it was only applicable to 8 of 15 BCCs (Fig. 3 B).

Tangents were fitted to points along the curve by ruler and the measured tangent slopes, d(FI)/dt, were plotted as a function of FI. Linearization was good (Fig. 5,A), in keeping with equation 6a of the appendix. The regression slope represents an accumulation rate constant β. The BCC accumulation rate constant, −0.0149 ± 0.0082 s⁻¹ (mean of 8), was 2.7-fold faster than the control rate constant, −0.00542 ± 0.00124 s⁻¹ (P < 0.01, paired t test; Fig. 5 B).

Back-extrapolation of a plot of Phase II d(FI)/dt versus time to zero time evaluates the intercept α (Fig. 5,A). This represents the fluorescein flux that would occur at zero interstitial concentration and the postbolus arterial plasma concentration. Because solute flux divided by arterial concentration is clearance, the ratio of (d(FI)/dt)_t=0 in BCC to that in the control represents the ratio of plasma clearances, assuming identical plasma levels in a given subject (“Appendix,” equation 7). The advantage of evaluating the plasma clearance at time zero is that, unlike the rate constant, it does not depend on interstitial fluid volume. The intercept (d(FI)/dt)_t=0, was 2.6-fold greater for BCC (0.871 ± 0.418 GL s⁻¹, mean of 8) than for the 8 paired controls (0.329 ± 0.160 GL s⁻¹; P < 0.01, paired t test; Fig. 5 C).

Analysis of the two averaged curves in Fig. 3 B (contrast with individual curves) gave similar results to those above. The BCC rate constant β (−0.0161 ± 0.0011 s⁻¹) was 3.5-fold faster than control (−0.0046 ± 0.0003 s⁻¹; P < 0.0001, analysis of covariance). The intercept (d(FI)/dt)_t=0 was 3.2-fold greater for BCC (0.923 ± 0.052 GL s⁻¹) than control (0.288 ± 0.012 GL s⁻¹). The results thus indicate increased plasma clearance in BCC.

The construction of tangents (above) involves subjective judgement. This can be circumvented by a logarithmic linearization that yields the rate constant but not clearance. From equation 8, “Appendix,” the relation between ln (zC_AV_i − FI) and time approximates to linearity. zC_AV_i represents the interstitial fluorescence upon equilibration to arterial plasma concentration; see “Appendix” for definitions. zC_AV_i could not be determined independently, but good linearization was obtained upon substituting FI_max, for zC_AV_i. FI_max may therefore be close to the equilibrium fluorescence level, zC_AV_i. The slopes of ln (FI_max − FI) versus time plots provided a second estimate of the accumulation rate constant β (equation 8). As with the tangent method, the BCC rate constant of −0.0130 ± 0.0056 s⁻¹ (mean of 8) was much faster (2.5-fold) than the controls, −0.0053 ± 0.0017 s⁻¹ (P < 0.01, paired t test, n = 8). Analysis of ln (FI_max − FI) versus time for the averaged data in Fig. 3 B likewise produced excellent linearization (r² > 0.98) and yielded a BCC rate constant of −0.0187 ± 0.0003 s⁻¹ that was 3.3 times the control rate constant, −0.0056 ± 0.0001 s⁻¹.

The log-linearization results for 16 studies (8 BCCs and 8 control) were not significantly different from the tangent method results (P = 0.3, paired t test) and were distributed about the line of equality (Fig. 5 D). This agreement supports the validity of the tangent method, which provided information about clearance (intercept) as well as rate constant.

The gradual clearance of fluorescein from the tissue in Phase III showed that the plasma concentration had fallen below the interstitial concentration. The decline was approximately exponential, and is well linearized by plots of ln (FI) versus time (Fig. 6,A), in agreement with “Appendix” equations 13 and 14. The slope of the logarithmic plot is the decay rate constant, which represents the fractional removal rate. The mean of the BCC removal rate constants, −(0.533 ± 0.194) × 10⁻³ s⁻¹, was 1.95-fold greater than control, −(0.273 ± 0.219) × 10⁻³ s⁻¹ (P < 0.01, paired t test, n = 11; Fig. 6 B). Phases II and III thus both showed a 2–3-fold increase in solute exchange rate in BCC. Phase III rate constants in BCCs that lacked a clear Phase II (−0.000603 s⁻¹, n = 5) tended to be greater than those of BCCs with an obvious Phase II (−0.000474 s⁻¹, n = 6), but the difference was not significant (P = 0.30, unpaired t test). Clearance was not followed for long enough in four early experiments for Phase III analysis.

Phase III kinetics differed from Phase II in one major respect: the rate constants were an order of magnitude slower in Phase III, as shown in Fig. 6c (P < 0.01, paired t test, n = 16 pairs). This quantifies an earlier report that fluorescein clearance from the nailfold by skin capillaries is slower than fluorescein accumulation (24).

Because of the invasive nature of the experiment, the arterial concentration profile after a rapid i.v. fluorescein injection was studied in only one subject. Concentration rose rapidly for 6 s as the bolus front reached the systemic arteries, then decayed over 14 s as the bolus passed by (Fig. 7 A). After the recirculation hump and mixing, arterial concentration stabilized at a reduced level for >100 s after injection.

Venous plasma from one subject was sampled for 30 min after i.v. injection of fluorescein to determine the changes in plasma concentration during the later stages of Phase II and Phase III (Fig. 7 B). Venous concentration decayed slowly from an initial peak at 80 s to 51% of the peak value at 30 min, an average decay rate of ∼1.8%/min. The plasma fluorescein is cleared by the kidneys (32, 33).

The relation between FI and number of fluorescein molecules in a 3.6 mm² hemocytometer field, whether varied by concentration or partial field occlusion, was linear up to 0.23 nmol and 136 GL after background subtraction. The latter light intensities were not exceeded in vivo (Fig. 3). The slope of the relation z at the magnification used in vivo was 669 GL nmol⁻¹. Linear mass intensity relations, both in vitro and in vivo, have been reported previously (3, 24, 26, 42). Alterations of the plane of focus by ±2 mm had no significant effect on FI; the maximum change was 1 GL. Gross misfocus by 7 mm caused a 35% increase in FI. Excitation of 0.023 nmol fluorescein in a hemocytometer chamber for 30 min reduced the emitted fluorescence from 54 ± 0.5 to 47 ± 0.5 GL (n = 3). For 0.100 nmol in a plastic well, the signal fell from 106 ± 10 to 87 ± 5 GL over 30 min (n = 3). Substantial photobleaching of fluorescein isothiocyanate has been reported (3, 26) or only 5–10% fade over 90 min of excitation (1) and a fade of 2–3% for sodium fluorescein over 20 min (25). Quenching of fluorescence by red cells (3, 26) depended on hematocrit. For a given mass of fluorescein in the field, a hematocrit of 24.5% reduced fluorescence emission by 49%; a hematocrit of 3.9% reduced emission by 19%; and a hematocrit of 0.7% reduced it by 7%.

The principal findings were increased microvascular density, increased blood flow and faster solute exchange rates in BCC compared with a matched control skin site.

The FI levels in vivo were well within the linear range in vitro. The large window averaged the exchange kinetics of numerous, heterogeneous microvessels in contrast to small window permeability measurements on individual vessels in optically favorable tissues (e.g., Refs. 26, 27). Tissue light scattering and light from regions above and below the plane of focus are less of a problem in large window studies than in small window, single vessel permeability studies. FI was not significantly changed by substantial (±2 mm) changes in focus in vitro. The photobleaching observed in vitro presumably caused some background decay in vivo, but it did not obscure the consistent difference between BCC and control rate constants in all three phases. In view of the quenching-hematocrit relation in vitro, it is possible that quenching was greater in the hypervascular BCC than in control sites. Because the rate constant is a fractional change with time, differences in quenching between sites should not affect it. Differences in quenching could, however, obscure the meaning of absolute FI (contrast with Ref. 29).

The ≥2-fold increase in microvessel image area fraction and length density is presumptive evidence for angiogenesis in BCC (Fig. 1). High A_As have also been reported in human BCC by native capillaroscopy (12, 13). The observed increase in vessel width indicates endothelial growth, so the raised vascularity can be attributed to angiogenesis rather than vessel incorporation. The fact that minimum profile width at the control site (11.6 μm) was bigger than normal capillary diameter (5–8 μm) is attributed to light scattering, which blurs the vessel edges. The reported image widths in BCC and controls should therefore be treated as of comparative rather than absolute value. Angiogenesis is crucial for tumor growth, and VEGF is recognized as an important angiogenic factor in many tumors (5, 6, 43, 44). VEGF expression is weaker in BCCs than squamous cell carcinomas (15, 16, 17) but when present is greater than in adjacent skin (17). These findings may indicate that additional growth factors are driving angiogenesis in BCC.

The >3-fold increase in blood flow demonstrates that a large reduction occurred in BCC microvascular resistance (Fig. 2). The pre- to postcapillary vascular resistance ratio in normal skin at room temperature is ≥2.3:1 (45). Therefore, a 3-fold increase in flow is much more than can be explained by a fall in capillary/venular resistance because of increased cross-sectional area, which at the most would raise flow by a factor of 3.3/2.3, i.e., 1.4-fold. It is inferred, therefore, that arteriolar resistance falls in BCC because of mediator/shear-induced vasodilatation and/or remodeling and/or arteriogenesis.

The slope d(FI)/dt in Phase I was ∼3-fold steeper in BCC than control skin (Fig. 3). The slope correlated with microvascular area fraction and red cell flux (Fig. 4). Although significant extravascular accumulation of fluorescein commences within 5–10 s of the arrival of sodium fluorescein in skin capillaries (24), d(FI)/dt was greatly influenced by and, perhaps dominated by, the filling of the intravascular compartment. Differences in the heterogeneous transit times may also influence the slope. Phase I does not, therefore, provide unambiguous, interpretable information about exchange kinetics.

Inspection showed that the slow rise in FI during Phase II was related to the accumulation of extravascular fluorescein. The progressive accumulation indicated that there was a concentration gradient driving diffusion from plasma to interstitium. Transendothelial convective transport is insignificant for small, rapidly diffusing solutes. The relative brightness of vessel and extravascular space was no guide to relative concentrations because brightness depends on fluorescein mass and therefore on the volume of the fluorescein-containing space as well as its concentration. As a result, the extravascular image was brighter than the microvessel image.

Averaged cutaneous FI over a fixed time interval (7 min) has been used as a measure of capillary permeability to fluorescein (29). We do not support this intuitive usage, at least with respect to a small, rapidly diffusing solute such as fluorescein, because FI at times approaching FI_max depends also on interstitial fluid volume, plasma volume, microvessel surface area, and blood flow (equation 9, “Appendix”). We therefore sought an alternative way of characterizing the exchange of a rapidly diffusing small solute by using information inherent in the curvature of the FI versus time relation.

Plots of d(FI)/dt versus FI (tangent analysis, Fig. 5,A) or ln (FI_max − FI) versus time successfully linearized the Phase II (Fig. 5,A) and yielded reassuringly similar rate constants (Fig. 5 D). The success of the linear transformation indicates that the simplifying assumptions needed to reach equations 6 and 7, specified in full in the “Appendix,” are acceptable approximations for this system.

The BCC microcirculation had a 2.6-fold greater initial plasma clearance and 2.5–3.5-fold greater rate constant than control. For a single, uniform exchange vessel, the rate constant depends on the ratio of permeability-area product to blood flow, PS/Q̇, and on blood flow per unit interstitial fluid volume, Q̇/V_i (equation 6c). Plasma clearance ratio at time 0 depends on PS/Q̇ but not on V_i because interstitial concentration is zero at time 0 (equation 7). In a heterogeneous set of tumor vessels, variations in the PS/Q̇ ratio between vessels will additionally complicate the net exchange observed by the large window method. Because S and Q̇ were increased in the BCC, as shown by morphometry and laser Doppler fluximetry, it is not possible to conclude that permeability was raised. For a rapid, partly flow-limited exchange process, a 2.6-fold increase in initial plasma clearance could be explained largely by the ∼3-fold increase in flow.

In 7 of 15 BCCs, Phase I and Phase II were not separable. A steep, brief accumulation phase was quickly followed by the removal phase. This type of kinetics is to be expected in tumors with (a) very high PS/Q̇ values (because interstitial concentration will then closely mirror the plasma concentration, Fig. 7 A) and/or (b) high Q̇/V_i ratios (because a big blood supply to a small interstitial volume allows rapid equilibration of the interstitium). Particularly high perfusion rates were indeed observed in BCCs with a merged Phases I and II.

Phase III was the simplest to analyze in that all BCCs had a clear Phase III, and there were no rapid changes in plasma concentration. The slow decay of venous concentration during this phase (Fig. 7,B) is due at least partly to renal clearance. Although the assumption in equation 12, that plasma concentration is constant, was not strictly valid in practice, the good linearization of Phase III by equation 14 (Fig. 6 A) indicates that the error introduced by the slow concentration decay is acceptably small.

The Phase III rate constants told a similar story to Phase II, which is on average ∼2-fold greater in BCCs than controls. This reinforces the conclusion that the rate constant determinants PS/Q̇ and/or Q̇/V_i (equations 6c and 9) are increased in BCC.

An interesting feature of Phase III was that its rate constants were an order of magnitude slower than in Phase II, both for BCC and for controls. Transcapillary diffusion is a passive, inherently symmetrical process, so the rate constants should in principle be the same for an unbound, diffusible solute (equations 6c and 9). The slow clearance of fluorescein relative to accumulation may be attributable to binding in the extravascular space (24). Indeed, depots of fluorescein-labeled dextran injected into the dermis during fluorescence lymphography sometimes remain visible for weeks as yellow stain. Alternative explanations would be a major vectorial contribution to plasma-to-tissue fluorescein transport by fluid filtration or active transport. We are not aware of any evidence for the latter. Filtration is often raised in tumors (1, 2), but diffusional transport normally so greatly outpaces convective transport that truly massive increases in filtration rate would be needed to generate an order-of-magnitude difference between the accumulation and removal rate constants. Moreover the plots of d(FI)/dt versus FI conformed well with the laws of diffusion upon which the analysis is based. It seems more likely, therefore, that a reversible, moderate affinity binding of fluorescein by extravascular elements accounts for the difference between Phase II and Phase III rate constants.

As pointed out above, the fluorescein FI curves neither prove nor disprove a change in tumor endothelial permeability to small solutes. Most evidence for increased permeability in tumor microvessels relates to macromolecules, usually dextrans and albumin, not small solutes (e.g., Refs. 1, 2, 3, 4, 9). On the dual pore theory (46), increased macromolecular permeability is caused by an increase in the large pore population. For a solute the size of fluorescein the diffusion capacity of the more abundant small pore system normally outweighs that of the large pore system by approximately three orders of magnitude. Consequently, even big increases in macromolecular permeability such as 8-fold (1) need not imply similar increases in small solute permeability.

The aggregate small pore area is determined by the extent of breaks in the junctional strands between endothelial cells (47). Tumor endothelium does not form a normal monolayer and has loose intercellular junctions with focal gaps up to 2 μm (9). It is possible, therefore, that strand breaks are more extensive in tumor microvessels, as in venules compared with arterial capillaries (48, 49). Such changes seem likely in tumors because tumor necrosis factor and IFN-γ both increase paracellular permeability and cause fragmentation of zonula occludens-1 protein organization in vitro (50). It is also relevant to note, in view of the increased blood flow in BCC, that endothelial permeability to fluorescein in mesenteric microvessels increases acutely as a function of flow (27). The effect of flow on fluorescein permeability is abolished by nitro-l-arginine and is attributed to increased numbers of pores of radius ∼0.8 nm.

In conclusion, the study showed that human BCC is associated with angiogenesis, increased perfusion, and faster exchange kinetics for a small, rapidly diffusing solute and by implication for unbound drugs. Whole-field FI is not, by itself, a sufficiently discriminatory tool for the assessment of vascular permeability to rapidly diffusing solutes (compare Refs. 29, 31). Large window FA can yield macromolecule apparent permeability if combined with small window measurements of intravascular fluorescence (3), but this approach has not yet been applied in man. Quantitative assessment of the heterogeneity of small solute exchange, which could influence local drug access, will require small window measurements.

The relation in vitro between fluorescein mass m and fluorescence intensity I was shown to be a straight line through the origin of a slope z.

Linearity is also observed in the hamster cheek pouch in vivo(42) and is assumed to apply for BCC tissue over the depth of focus used here. Nonlinearity would not alter the study’s fundamental demonstration of increased exchange kinetics but would complicate the analysis that follows.

The diffusional transport of a small solute across a single perfused microvessel (neglecting the minor convective component) is given by the well known Renkin-Crone expression (51).

where Q̇ is plasma flow, C_A and C_i are arterial plasma and interstitial fluid concentrations, P is endothelial permeability, and S is endothelial surface area. This expression is derived for a uniform interstitial concentration around the vessel. From equations 1b and 2, the rate of change of fluorescence dI/dt attributable to solute escape from a single vessel is given by the following expression.

Total fluorescein mass in a field of unit area is the sum of the fluorescein in the interstitial fluid (volume V_i) and in the intravascular plasma (volume V_p, mean concentration C_p). Therefore, from equation 1a, the total field intensity I (referred to as FI in the main text) is given by the following.

The above expressions relate strictly to a single uniform vessel. If tumor microvessels have heterogeneous PS/Q̇ values, net dI/dt will be disproportionately weighted by vessels of high PS/Q̇ because flux is a nonlinear (exponential) function of PS/Q̇. In the following analysis of Phases II and III, PS/Q̇ represents a nonarithmetical average for multiple, potentially heterogeneous microvessels.

Equations 1b, 2, and 4b combine to give the rate of increase in total FI during Phase II.

This can be written as

where

and

A plot of dI/dt (tangent to the Phase II curve in Fig. 3) versus I will be linear if α and β are constants. The slope of the plot is the accumulation fractional rate constant β (units s⁻¹), which depends on the vascular clearance parameter Q̇(1 − exp[−PS/Q̇]) relative to interstitial fluid volume. Because V_p/V_i is small and C_p < C_A, intercept α approximates to zQ̇(1 − exp[−PS/Q̇])C_Ain units of GL s⁻¹(equation 6b). If C_A is essentially constant during Phase II (Fig. 7,A), α can be evaluated by linear extrapolation of a plot of dI/dt versus I_(t) to t = 0 = I (Fig. 5 A). Plasma clearance at this point (t = 0) equals the solute flux (dI/dt)_t=0z⁻¹ (= αz⁻¹) divided by C_A. If C_A is the same for identical injections in a given subject, the ratio of BCC α to control α estimates the plasma clearance ratio.

Because the clearance of a partly flow-limited small solute depends on blood flow and surface area as well as permeability, it is not possible to determine permeability from the accumulated FI alone (contrast with Ref. 29).

The integration of equation 6 provides an alternative linearization procedure and a useful definition of the parameters determining FI after a fixed time. If C_A does not vary significantly with time over Phase II, integration of equations 6a, 6b, and 6c, and evaluation of the integration constant gives

Thus FI is given by

where rate constant β is again Q̇(1 − exp[PS/Q̇]) / V_i. FI thus depends not only on permeability but also on the vascular parameters Q̇ and S and depends inversely on the interstitial parameter V_i.

Let the mass of fluorescein in the interstitium be m₀ at the onset of Phase III (onset time defined as t = 0) and m_i at any subsequent time t. Substitution of C_i = m_i/V_i into equation 2 gives removal rate as a function of interstitial fluorescein mass.

Plasma concentration changes only a little during Phase III (Fig. 7 B), so as an approximation, C_A was treated as a constant. Integration and evaluation of the integration factor shows that interstitial fluorescein mass during the clearance phase is a negative exponential function of time.

where the removal rate constant β is again Q̇(1 − exp[−PS/Q̇]) / V_i, and C_AV_i is interstitial fluid fluorescein mass at equilibrium, m_∞.Removal rate constant β is identical to the accumulation rate constant β in Phase II (equation 6c) for a solute that is not bound by interstitial matrix components. Conversion of fluorescein mass to FI using equations 1a and 4a leads to an exponential expression for whole-field fluorescence inclusive of plasma.

Taking natural logarithms we have

If the fluorescein is mainly extravascular up to 30 min, as inspection indicates, the intravascular contribution to I is negligible, and the left hand term approximates to ln I, giving the approximate linear relation

The linearity of ln I versus time plots in practice provided support for the simplifying approximations. The removal rate constant β is given by the slope of a plot of ln I versus time and depends on three vascular parameters, Q̇, P, and S, and on one extravascular factor, V_i.

We thank Dr. R. Allan Marsden (Department of Dermatology, St. George’s Hospital) for help with recruitment of patients and the patients themselves.