Resistance of Tumor Interstitial Pressure to the Penetration of Intraperitoneally Delivered Antibodies into Metastatic Ovarian Tumors

Metastatic ovarian and colorectal carcinoma remain devastating diseases because tumor nodules (5-10 mm diameter) on the peritoneum cannot be detected by oncologic surgeons (1, 2). Interperitoneal therapy for ovarian carcinoma with small molecular weight drugs (MW < 1,000) has produced a longer survival time (3). To improve detection and treatment, antibodies labeled with metabolic poisons or radioisotopes have been used for tumors that have surface antigens (4–7). Preclinical pharmacokinetic studies on the regional delivery of macromolecules have been limited, and unfortunately, success in human therapy has been poor. Large molecular weight (MW > 50,000 Da) agents such as antibodies typically transport via convection or “solvent drag,” which depends on pressure differences between the therapeutic solution and the targeted tumor (8–10).

In this article, we address the hypothesis that the driving force of hydrostatic pressure in the cavity must be sufficiently superior to that of the tumor to drive the antibody into the target tissue. To test the hypothesis, we use our previously developed tumor-bearing rat model of ovarian carcinoma, in which we showed a modest advantage in short-term penetration with higher i.p. pressures (11). However, when we compared the concentration profiles of the immunoglobulin in both tumors (SKOV3 and OVCAR3) with those of normal muscle, the penetration was much less than in previous observations in abdominal wall tissue of another rodent species (12–14). We made measurements of the interstitial pressure profile in tumors and show that within the first millimeter of tumor tissue below the peritoneal surface, the pressure increases to levels above that in the cavity. We carried out experiments with relatively high i.p. pressures and after several hours showed significant differences between concentration profiles of the immunoglobulin in tumors compared with those in adjacent normal abdominal wall. After showing saturable binding of trastuzumab to the cell surface antigens of both of our xenografts, we carried out analogous experiments with an immunoglobulin G (IgG) with no specific receptors on the tumor cells. The profiles were unchanged and therefore we attributed the lack of penetration to the resistance of the hydrostatic pressure within the tumor. We then tried various maneuvers to lower the pressure and finally succeeded in decreasing the pressure to near zero, but there was still very little additional penetration of the antibody into the tumor. We discuss the possible mechanisms involved with this phenomenon.

RNU nude rats, 150 to 200 g, were purchased from the National Cancer Institute (Frederick, MD). They were housed in the Athymic Animal Facility University of Rochester or at the barrier at the Laboratory Animal Facility at the University of Mississippi Medical Center. The research protocols were approved by each local Animal Care and Use Committees.

Details of the culture of the tumor cells, media, and tumor implantation procedures are contained in our previous publication (11). The two types of cells used in all studies were SKOV3 or OVCAR3 cells, purchased from American Type Culture Collection (Rockville, MD). Cells were harvested and, within 3 hours, 30 × 10⁶ to 40 × 10⁶ were implanted in each side of the anterior abdominal wall to create nodules of tumor cells below the peritoneum but protruding into the cavity (cell viability >85% at implantation; ref. 11). The location of the implantation was designed to facilitate xenograft growth into the peritoneal cavity and measurement of tumor interstitial pressures from the exterior of the abdominal wall. All animals were further immunosuppressed with daily s.c. injections of cyclosporine (20 mg/kg).

Preparation for an acute experiment. Animals were anesthetized initially with i.m. injection of pentobarbital sodium 60 mg/kg, followed by i.v. injections, after placement of the i.v. catheter. A tracheostomy was done, and catheters were placed intra-arterially to monitor blood pressures and sample blood. Mean arterial blood pressure was always greater than 80 mm Hg. Rectal temperature was always maintained at 36.5 ± 1.5°C with the use of a heating lamp and warming blanket.

Whole-cavity experiment. To carry out a whole-cavity experiment, an i.p. catheter was placed through the abdominal wall for instillation and hourly sampling of the i.p. solution. The catheter was connected via a three-way valve and tubing to a reservoir, which was raised above the right heart to set i.p. pressure between 0 and 8 mm Hg (0-11 cm H₂O). This pressure was verified every 15 minutes by a glass manometer.

Chamber experiment. In some experiments, an alternative treatment technique was done with a plastic chamber (15) glued to the surface of the peritoneum over the tumor to isolate the tumor and ensure contact of the therapeutic solution with the tumor surface. The height of the fluid in the chamber determined the pressure exerted on the tumor surface. Sampling was done hourly through a port at the top of the chamber after gently mixing the chamber solution with a syringe and needle/catheter. Plasma samples were taken via the arterial catheter at the same time as the chamber samples were taken.

The details of the isotonic solution used are in our previous publication (11). The monoclonal antibody (mAb) to the human epidermal growth factor receptor-2, trastuzumab, was purchased from the University of Mississippi Medical Center Pharmacy and carefully labeled with ¹²⁵I in accordance with the technique in our previous publication (11). Each experimental day, before the use of an ¹²⁵I-labeled protein, the tracer was repeatedly diluted and filtered via centrifugation to ensure a free ¹²⁵I fraction of ≤1% (8). Following each experiment, samples of the plasma and peritoneal fluid were checked for free isotope by precipitation by tricarboxylic acid; results showed free ¹²⁵I of <1.3%. Total radioactivity in the urine was less than 0.1% of the total dose absorbed.

Anti-rabbit ¹²⁵I-labeled IgG (produced in donkey, immunoabsorbed for human and rat antigens) was purchased from Amersham Corp.(Piscataway, NJ) to act as a nonspecific control antibody. Its binding was checked with methods below and found to be equivalent to the nonspecific binding of trastuzumab and other IgG (8). This labeled compound was separated from free ¹²⁵I as previously described (11).

Gamma counting was done with a Packard Cobra II Auto-gamma Counter or a Beckman 8000 gamma Counter.

Measurement of interstitial pressure. To facilitate multiple determinations of the overall pressure in a single experiment, the wick-in-needle (WIN) technique, as documented by Fadnes et al. (16), Boucher et al. (17), and Wiig et al. (18), was used to measure the pressure beyond 1 to 2 mm. Because the WIN method is somewhat cumbersome to obtain exact profiles, particularly close to the tumor surface, the micropipette/servo-null technique of Wiig et al. (18) and Boucher et al. (17), as modified by Flessner and Schwab (19), was used to measure the hydrostatic pressure profile in the normal abdominal wall muscle and within the tumor parenchyma. Details of this technique are in our previous publication (11).

Quantitative autoradiography. Details of this procedure are in our previous publications (12, 13). Briefly, sections of 10 to 20 μm are cut from the frozen specimen in a cryomicrotome and heat dried or freeze dried to prevent further transport of the labeled antibody. Sections are placed with isotopic standards against an X-ray film (Kodak Biomax MR, Rochester, NY) to produce autoradiograms. These are later analyzed with a computerized densitometer (MCID Imaging Research, Inc., St. Catherine's, Ontario, Canada).

Calculations and statistics. Calculations were done with Microsoft Excel (version 97 or 2000). Statistical calculations were done with NCSS (Provo, UT). Data were represented as mean ± SE or SD. Tissue concentrations were typically normalized by dividing by the time average concentration of the peritoneal solution. Quantities were judged to be significantly different if the probability of a type 1 error was <0.05 (P ≤ 0.05).

The purpose of studying binding to each tumor of immunoglobulin and trastuzumab was to ascertain the degree of saturable binding to specific cell surface receptors and to test whether there were differences in penetration that could be attributed to binding to tumor cells. It has been theorized that binding to cell surface receptors can prevent the penetration of mAbs into tumors (20, 21).

To measure the specific and nonspecific binding of trastuzumab as a function of time to the tumors, we used the combined work of two investigators (22, 23) and our previous experience (24). Tumor implants were grown to 0.5 to 1.0 cm in diameter in the abdominal wall of athymic rats, which were sacrificed with an overdose of sodium pentobarbital, and the tumors were excised and frozen in isopentane at −70°C. These were sliced to 10-μm sections in a cryomicrotome (model OTF, Bright-Hacker, Fairfield, NJ or with a Leica CM3050S, Leica Microsystems, Nussloch, Germany) and taken up on a glass slide and dried on a warming plate. All tumor slices were preincubated in 5% bovine serum albumin/5% chicken serum in PBS for 30 minutes to decrease binding of the labeled antibody to the glass slide. To determine the total binding (total binding = sum of saturable and nonsaturable bindings), the slices were incubated in five to six different concentrations of labeled antibody for 10, 30, 60, or 240 minutes. After washing and drying, the slides were placed against an X-ray film (Kodak Biomax MR) along with radioactive standards (13, 25) to develop quantitative autoradiograms. These were analyzed with a computerized densitometer and MCID software (Imaging Research) to produce data consisting of counts per minute bound per milligram of tissue. The counts per minute per milligram of tissue data were converted to picomoles per gram by dividing the counts per minute by the specific activity of the isotope (cpm/pmol). The bound antibody concentration (pmol/g) was then plotted versus the free concentration in the solution (pmol/mL). Because the concentration varied for the longer incubations, the incubation solutions were sampled every hour and the concentrations were averaged to provide an average concentration of tracer [C̄_free (pmol/mL), deviation from mean: ± 2%]. These data were fitted to the equation of the form:

where b_ns(t) (mL/g) equals the slope of the nonsaturable binding curve; B_max(t) (pmol/g) and K(t) (pmol/mL) are coefficients for saturable binding modeled by Michaelis-Menten kinetics.

To determine nonsaturable binding, the incubations of a separate set of slides were repeated with unlabeled antibody in the solution at a concentration of 0.12 mg/mL (100-300 × the concentration of labeled antibody). The unlabeled antibody competes with the labeled antibody for the saturable binding sites so that presumably only nonsaturable sites are available for detection. Quantitative autoradiograms were developed and analyzed as above. Data were fitted to the following equation:

Saturable binding was determined for each incubation time by subtracting the nonsaturable binding curve (Eq. B) from the total binding curve (Eq. A).

In addition, the dissociation coefficient (b_r, h⁻¹) of the antibody from the tumor was determined by saturating tumor slices with labeled trastuzumab, and then measuring the amount left on the slide versus time of incubation in a solution free of the tracer over 24 hours. Data were fitted to the following equation:

Paclitaxel to reduce tumor pressure. Because of previous publications of the use of taxanes to reduce interstitial pressure (26), animals with established tumors (dia > 5 mm) were injected with a 0.125 mL i.v. bolus containing ∼12.5 mg of paclitaxel (50-60 mg/kg). After 48 to 72 hours, the animal was anesthetized and pressure measurements were made, and a transport experiment was carried out as described above. Because of the necessity to dissolve the paclitaxel in Cremophor and ethanol, two animals died immediately after the injection. We subsequently switched to the water-soluble Taxotere.

Taxotere to reduce tumor pressure. Taxotere was administered i.p. 3 to 4 days ahead of the experiment. Dosages were typically 50 to 60 mg/kg i.p. (6 to 9 mg per animal). After 3 to 4 days, the animal was prepared for a whole-cavity experiment in which the Krebs-Ringer solution contained 0.02% dioctyl sodium sulfosuccinate, which has been shown to facilitate coverage of the entire peritoneal surface with the i.p. solution (27). Animals underwent 3 hours of a full-cavity experiment with a solution containing ¹²⁵I-labeled trastuzumab at i.p. pressures of 4 to 6 mm Hg (∼6-8 cm H₂O). At the end of experiment, the animal was euthanized, the fluid was removed, and the tumor was rapidly harvested, frozen, and processed in the cryotome for quantitative autoradiography. Pressures were measured with WIN method and, when possible, with the micropipette/servo-null system (11, 19).

Prostaglandin E₁ to reduce tumor pressure. There has been recent evidence that prostaglandin E₁ (PGE₁) causes changes in the tumor microcirculation if administered in the s.c. tissue surrounding transplanted carcinomas, and lowers tumor pressures by 30% after a 15-μg injection (28). In our protocol, s.c. injections of 100 μL of PGE₁ solution (each 100 μL containing 15 μg purchased from Sigma Corporation, St. Louis, MO) were made at four points around the base of the tumor every 30 minutes. No injections were made directly into the parenchyma of the tumor. Because each animal had multiple tumors, one tumor was treated with PGE₁, and another tumor was treated with 100 μL of vehicle (isotonic Krebs solution) and acted as a control. WIN measurements of the pressure before and during the procedure were made. These animals were treated with a whole cavity solution containing ¹²⁵I-labeled trastuzumab at pressures of 4 to 6 mm Hg (∼6-8 cm H₂O). In a similar fashion to the above experiments, at 3 hours the final plasma samples and peritoneal samples were taken and the animal was euthanized and the tumor tissue was rapidly frozen and processed for quantitative autoradiography.

Penetration in euthanized animals (zero microvascular pressure). Although PGE₁ treatments decreased tumor pressures by ∼50% to 75%, the remaining tumor pressure often exceeded the pressure which could be attained in the cavity. It has been hypothesized that high interstitial pressures in the tumors are a result of an abnormally high microvascular pressures (26, 29, 30). Animals were therefore euthanized, and blood pressure was observed to decrease to zero, and tumor pressures were determined with WIN measurements. Whole-cavity experiments were carried out with solutions containing ¹²⁵I-labeled trastuzumab, and the tumor concentrations were measured with quantitative autoradiography.

Decapsulation of tumors to lower tumor interstitial pressures in vivo. Animals were anesthetized and the tumor peritoneal surface was exposed. The tumors were decapsulated or unroofed through careful dissection to expose the tumor parenchyma directly to the solution containing ¹²⁵I-labeled trastuzumab. Tumor pressures were measured with the WIN technique and then a transport chamber was applied over the tumor to carry out an experiment as described above. At the end of the experiment, the animals were euthanized and the tumors were harvested, frozen, and prepared for quantitative autoradiography.

To understand the forces within the ovarian xenografts which might resist penetration of macromolecules, tumor interstitial pressure profiles were determined with a micropipette/servo-null system. Figure 1A illustrates the pressure profiles measured in 12 SKOV3 tumors. The origin of the x-axis of Fig. 1A is the s.c. edge of the tumor, and the points deep in the tumor lie toward the peritoneum. The range of pressure was 0 to 32 mm Hg, whereas the median pressure was 15 mm Hg and the mean ± SE (SD) was 9.7 ± 2.4 (8.3). The tumor pressures were quite variable within individual tumors and between tumors. The typical size of these tumors was on the order of 3 to 10 mm in diameter, some with oval or cylindrical shape and others with more spherical shape. In the same fashion, the OVCAR tumors (n = 15) displayed significantly high pressures (Fig. 1B), with a median pressure of 12 mm Hg and mean ± SE (SD) was 12.5 ± 1.8 (7.0).

Note that the pressures are lower towards the edge of the tumor but increase steeply as the micropipette enters the tumor. Because convection of a large molecule into the tumor requires a superior pressure in the cavity relative to the tumor, the i.p. pressure would have to be raised above 12 to 15 mm Hg for macromolecular convection into the center of the majority of tumors. In the rat model, it is not possible to raise the i.p. pressure above 8 mm Hg without compromising the circulation through the portal vein and causing venous stasis in the gut (31). In addition, the large volume required to attain this pressure exerts force on the diaphragm, which compromises respiration.

To predict the potential role in transport resistance due to binding of an antibody to specific sites on the surface of tumor cells, the binding of immunoglobulins to the two ovarian xenografts was determined.

SKOV3 tumors. The time-dependent binding characteristics of ¹²⁵I-labeled trastuzumab to SKOV3 tumors are illustrated in Fig. 2A. Saturation of receptors rapidly occurs and approaches a steady state after 30 minutes. The nonsaturable binding continues to increase with time of incubation. The total binding curve at 60 minutes is as follows (bound mAb in pmol/g; free mAb in pmol/mL):

The saturable binding curve is found by subtracting the nonsaturable binding curve at 60 minutes from the total binding curve:

The half-saturation concentration (K) is relatively low at 0.90 pmol/mL. The maximum number of binding sites (B_max) is very high at ∼179 pmol/g, indicating a large number of receptors. The binding is nearly irreversible with a dissociation coefficient (b_r) equal to 0.0021 h⁻¹. This corresponds to a dissociation T_1/2 = 330 hours.

The total binding of the anti-rabbit ¹²⁵I-labeled IgG to SKOV3 and OVCAR3 was determined with the same techniques and found to have binding equivalent to the nonsaturable binding of trastuzumab (see straight lines in Fig. 2A and B).

OVCAR3 tumors. The time-dependent binding characteristics of ¹²⁵I-labeled trastuzumab to OVCAR3 tumors are illustrated in Fig. 2B. Saturation of receptors occurs at a slower pace but approaches a steady state after 30 minutes. The nonsaturable binding continues to increase with time of incubation. The total binding curve at 60 minutes is as follows (bound mAb in pmol/g; free mAb in pmol/mL):

The saturable binding curve is found by subtracting the nonsaturable binding curve at 60 minutes from the total binding curve:

The half-saturation concentration (K) is relatively low at ∼6 pmol/mL. When compared with SKOV3 tumors, the maximum number of binding sites (B_max) is relatively low at ∼8.8 pmol/g and indicates a much lower number of OVCAR3 specific receptor sites when compared with SKOV3. The process of dissociation is slow with a rate of dissociation equal to 0.019 h⁻¹. This corresponds to a dissociation T_1/2 = 36.5 hours.

Effect of binding on penetration of ¹²⁵I-labeled trastuzumab into SKOV3. Because the SKOV3 tumors possess such significant saturable binding in the concentration typically used in our experiments (10-20 pmol/mL of antibody), further experiments were done with an IgG antibody with no specific binding (binding characteristics were nearly the same as the nonsaturable binding for trastuzumab; illustrated in Fig. 2A). Shown in Fig. 3A are the trastuzumab profiles from our previously published article (11), the concentration profile measured in normal adjacent abdominal wall muscle and that of the IgG in the tumor. The experiment in the animals with a nonspecifically binding IgG showed the same curve as with trastuzumab and markedly less than that of normal tissue. From these results, we conclude that binding is not the major cause of the lack of penetration.

To further illustrate the apparent resistance to penetration in tumors, Fig. 3B displays a SKOV3 tumor surrounded by a layer of muscle. Transport occurs from left to right with the i.p. pressure averaging ∼6 mm Hg; there is significant penetration of trastuzumab in the muscle but the concentration quickly decreases in the tumor.

Taxanes. Treatment of SKOV3 tumor–bearing rats with paclitaxel reduced tumor pressures to an average (mean ± SE) of 6.4 ± 2.7 mm Hg (n = 5 animals). Unfortunately, this was not enough of a tumor pressure reduction to gain a significant superiority of i.p. pressure, and the mean profile for ¹²⁵I-labeled trastuzumab (mean ± SE, n = 92 profiles) in Fig. 4 is not different from the data from the corresponding high pressure curve in SKOV3 tumors (Fig. 3A), which is plotted for comparison. Further experiments with Taxotere (n = 9 animals, 135 profiles) showed similar ¹²⁵I-labeled trastuzumab concentration profiles to the SKOV3 control animals and to those treated with paclitaxel (see Fig. 4). Despite modest reduction in the mean tumor pressures, the SKOV3 tumors were apparently resistant to taxanes and their effect on interstitial pressure (26).

PGE ₁ treatment.Figure 5 illustrates the results of the PGE₁ injection treatment in both SKOV3 and OVCAR3 tumors and compares the penetration of ¹²⁵I-labeled trastuzumab to that in control (untreated) tumors after 3 hours of dialysis with a solution at an i.p. pressure of 4 to 6 mm Hg. In the case of OVCAR-3, the pressure reduction was 15.5 ± 4.3 (± SE, n = 6) to 4.5 ± 1.8 mm Hg, whereas in SKOV3, the mean tumor pressures were lowered from 20.2 to 3.9. Control tumors varied from 10.4 ± 4.0 before injection of vehicle to 14.6 ± 5.3 after injection of equivalent volumes of vehicle. Figure 5 illustrates the concentration profiles in OVCAR-3 from mean of n = 28 profiles from six paired control or treated tumors. For SKOV3 tumors, the overall penetration was not much different from the OVCAR tumors or from the SKOV3 controls (see Fig. 5; 8 treated animals, n = 71 profiles; 4 control animals, n = 36 profiles). Despite a significant effect of PGE₁ on tumor pressure, penetration of ¹²⁵I-labeled trastuzumab was not enhanced. We attributed this to the fact that the pressure reduction was not sufficient to create a positive pressure difference of more than 1 to 2 mm Hg between the peritoneal cavity and the tumor parenchyma. Therefore, we investigated additional means to lower tumor pressure.

Cessation of blood circulation. In a further effort to decrease the pressure in the tumor, the animals were prepared for a whole-cavity experiment, and their circulation was stopped by barbiturate overdose. The concentration profiles (not shown) in normal abdominal wall and in tumor after 3 hours did not differ from those in Figs. 3 to 5. In general, the abdominal wall had a much higher concentration and the antibody penetrated deeper into the tissue than in the tumors. In measuring the pressure in the tumors with the WIN, it was found that the initial tumor pressures varied from 0 to 48 mm Hg (mean ± SE, 23.4 ± 3.4) but decreased over 5 to 10 minutes to zero, if the needle was left in the tumor parenchyma. If the tumor pressure was measured at another site in the same tumor, it would register at a high value and then slowly decrease after the initial puncture, as if the local pressure (interstitial fluid) was released around the needle after insertion into the tumor. After some experimentation, it was found that if the outer layer of connective tissue or “capsule” of the tumor was dissected away, the pressure tended to drop toward zero.

Decapsulation of tumors to reduce pressure. The invasive procedure of carefully dissecting the apparent tumor capsule reduced the WIN-measured pressure in each tumor to zero. In these experiments, the i.p. pressure was maintained at 4 to 5 mm Hg (6-8 cm H₂O), which set up a hydrostatic pressure gradient to cause convection of protein into the tissue. However, Fig. 6 shows that there remained major differences in the penetration of ¹²⁵I-labeled trastuzumab into both tumors when compared with adjacent normal abdominal wall muscle. The higher concentrations at the peritoneal edge of the abdominal wall versus the tumor may be due to adsorption of ¹²⁵I-labeled trastuzumab to the normal peritoneum (32), which was not present on many of the tumors.

Pressure profiles in OVCAR3 and SKOV3 xenografts.Figure 1A and B shows pressure profiles were generally higher toward the central regions of the tumors and lower near the periphery. This pattern was observed by Boucher et al. (17), who measured similarly high tumor pressures in human melanoma (33) and in cervical cancer (34). These pressures differ from the normal pressure of the abdominal wall muscle, which is slightly negative (35, 36). The slope or gradient of each pressure profile favors movement of macromolecules and fluid from the interior of the tumor toward the periphery of the tumor and presents a significant challenge to the penetration of antibodies from the periphery of the tumor (see below).

Pressure-driven convection of macromolecules in tumors. Whereas diffusion is the most important mode of passive transport of small molecules (MW < 5,000 Da) in biological tissues, convection or solvent drag typically is the major mechanism of movement of large proteins (MW > 40,000 Da; refs. 8, 9, 14). There is the possibility that the radioconjugates were metabolized to smaller molecules at the tumor surface and that the concentration profiles observed in Figs. 3 to 6 resulted from a free ¹²⁵I label which diffused into the tissue and decreased in concentration due to absorption into the tumor microcirculation. Two facts make this less probable: (a) significant amounts of free ¹²⁵I were not found in the urine, plasma, or peritoneal fluid; and (b) the concentrations within tumor would be significantly higher as they are for mannitol (11).

For a given moment in time, the unidirectional transport equation takes the following general form (for more details, see ref. 14):

where D, the solute effective diffusivity (∼10⁻⁸ cm²/s for IgG ref. 14); dC / dx, slope of the solute concentration profile (see Fig. 3A, LP concentration profile for trastuzumab after 3 hours of diffusion); K, effective hydraulic conductivity which depends on tissue properties and molecular size; C(x), solute concentration versus position in the tissue; dP / dx, slope of the hydrostatic pressure profile in the tissue, the driving force for convection; and F_binding, net local binding of antibody. Because the product of D and dC / dx tends to be a very small number, there must be a decreasing pressure profile from the cavity into the tumor to drive a large solute via convection from the peritoneal side of the tumor. The lack of penetration in tumors could be due to a diminished slope of the pressure profile. If the pressure profile were linear across the tumor (thickness ∼5-7 mm) as it is in the abdominal wall (thickness ∼1-2 mm; see Fig. 7), then the slope could be markedly diminished in tumors. When fluid is placed in the cavity, the pressure profile across the rat abdominal wall is linear, as shown in the solid symbols of Fig. 7, along with a hypothetical pressure profile for an i.p. pressure of 8 mm Hg (dashed lines; ref. 35). The maximum pressure which can be sustained for more than a few minutes in a supine, anesthetized rat who is not intubated and ventilated is 8 mm Hg (37). Figure 7 also displays the tumor mean values from Fig. 1; the curves are assumed to be symmetrical so that the pressures measured from the skin side are plotted as if they are measured from the peritoneum. Figure 7 shows that the potential driving pressure (dP / dx = d(P_theoretical − P_tumor) / dx) will be markedly degraded over the first 200 to 300 μm of tumor tissue, where the driving pressure from the cavity exceeds that in the tumor. Therefore, the enhanced penetration, which we observed in the initial 300 μm below the peritoneal surface (Fig. 3, compare HP versus LP curves) after raising the i.p. pressure from zero to ∼5 to 7 mm Hg, is likely due to increased convection from the cavity into the superficial portion of the tumor where the pressure from the cavity is above the tumor pressure. We hypothesized that the lack of more significant penetration was due to the high tumor interstitial pressures in the deeper regions of the tumor and that a greater degree of penetration could be obtained if this pressure could be lowered to zero. However, we needed to first rule out a “binding site barrier” (20).

Binding of trastuzumab to SKOV3 and OVCAR3.In vivo binding of trastuzumab to either SKOV3 or OVCAR3 tumor cells could result in retardation of its penetration into tumor parenchyma (20, 38). Figure 2B shows that there is relatively little saturable binding of the antibody to specific receptors on OVCAR3 tumors. However, the rapid and nearly irreversible binding of trastuzumab to a large number of receptors on SKOV3 (Fig. 2A) required investigations with an IgG with no specific receptors on SKOV3 tumor cells. Because the IgG did not penetrate to any greater degree than trastuzumab, a binding site barrier due to saturable binding to specific receptor sites was not responsible for retardation of antibody through the tumor.

Maneuvers to lower tumor interstitial pressure. Several methods were attempted to lower the tumor pressure to establish a positive difference between the solution pressure in the cavity and the tumor. Systemic administration of taxanes and local injection of PGE₁ were both successful in significantly lowering the tumor pressure, as had been observed by others (28, 39). However, Figs. 4 and 5 show no additional penetration of the antibody. Cessation of circulation through euthanasia did not lower the tumor pressure, but we observed that removal of the apparent tumor capsule resulted in near zero pressure. In vivo decapsulation of both OVCAR3 and SKOV3 tumors resulted in very low pressures but unfortunately showed no enhancement of antibody penetration into tumor. It should be noted that cell line–derived experimental tumors may not match all the characteristics of de novo peritoneal tumors. For example, the apparent capsule, which on removal resulted in the decrease of pressure to zero, may or may not reflect all types human metastatic disease. The model does, however, represent a small metastatic tumor which protrudes into the peritoneal cavity and has been thoroughly discussed in our previous publication (11). Our hypothesis is therefore only partially supported by these data, which point to the tumor interstitial structure or “microenvironment” as a major source of transport resistance.

Microenvironment of ovarian tumors as a source of transport resistance. Our results have generated a new hypothesis that ovarian tumor cells integrated with the interstitium, called by some the tumor microenvironment, must present a structural barrier to the entrance of mAbs. Recent work with in vitro multicell layers of tumor cells has shown significant resistance to the penetration of relatively small chemotherapeutic agents (40). The resistance to penetration of antibodies could be due to several mechanisms: (a) the volume of distribution of the tumor for IgG may be miniscule compared with that in the muscle and a relatively high concentration of IgG within the space available to the molecule cannot possibly raise the concentration based on total tissue weight very high; (b) the matrix of the tumor interstitium presents a physical or electrical (repulsive) barrier so as to prevent entry of a negatively charged macromolecule; or (c) interaction between the interstitial matrix and the tumor cells results in a barrier to macromolecules.

The interstitium of normal tissue, such as the muscle of the anterior abdominal wall, seems to be much more porous to IgG antibodies (see Fig. 3A and B). This is despite the fact that the interstitium makes up a much larger portion of the tumor tissue than in normal tissue; interstitial volume relative to the tissue mass was 0.53 ± 0.11 μL/mg in SKOV3, 0.61 ± .03 μL/mg in OVCAR3, and 0.21 ± 0.06 μL/mg in normal abdominal wall muscle (11, 36). The volume of distribution of IgG in muscle has been determined with in vivo methods to be 0.03 to 0.08 μL/mg (24, 41) in nonexpanded tissue. In experiments done in vitro, the volume fraction of neutrally charged dextrans with MW of 70 kDa in fibrosarcoma was determined to be ∼0.10 μL/mg whereas the fraction for a 2,000 kDa dextran was ∼0.04; the interstitial space of this tumor has been shown to vary from 0.33 to 0.60 (42, 43), not unlike the range for the tumors in our studies. Whereas the IgG volume of distribution has not been measured in SKOV3 or OVCAR3, we assume that it is in the same range as normal tissue or fibrosarcoma. If this assumption is valid, then the argument cannot be made that the IgG volume of distribution is too small to allow the concentration within the tumor to increase above small values. However, there is one factor that this argument does not take into account: that the interstitium of the abdominal wall in Fig. 3 or Fig. 6 is likely expanded to twice its normal value (36, 44). The hydraulic permeability increases significantly in the volume-expanded tissue (32). The volumes of distribution for proteins could be altered by this expansion of the available volume (45) or lead to less exclusion of the molecules (46).

The results of this research point to the interstitial matrix-tumor cell structure as a major resistance to macromolecular transport. This could result from exclusion of a negatively charged IgG by the negative charges of the interstitium. Hyaluronan has fixed negative charges and excludes proteins from a tissue volume which is many times its molecular volume (47). Recent research (48) has shown that albumin molecules with a net positive charge are excluded from 16% of the interstitium in skeletal muscle whereas the native albumin with fixed negative charge is excluded from 26%. Fixed negative charges were responsible for 40 % of the total albumin exclusion in skeletal muscle (46). Others have found evidence that tumor diffusivity and hydraulic conductivity are correlated with collagen fiber content, and that treatment of the tumor with collagenase increased the diffusional and convective permeability of tumors (49, 50). Effective i.p. delivery of macromolecules to metastatic ovarian carcinoma will require further research to investigate the nature of this interstitial resistance.