d Neurosurg 76:679-686, 1992
Laser-induced fluorescence: experimental intraoperative delineation of tumor resection margins WAI S. POON, M.B.,CH.B., F.R.C.S., KEVlN T. SCHOMACKER,PH.D., THOMAS F. DEUTSCH, PH.D., AND ROBERT L. MARTUZA, M.D.
Department of Surgery (Neurosurgery) and Molecular Neurogenetics Laboratory, Wellman Laboratories of Photomedicine, and Department of Dermatology, Massachusetts General Hospital, Harvard Medical School, Boston. Massachusetts v, The ability of laser-induced fluorescence spectroscopy to delineate tumor margins intraoperatively was studied using a rat intracerebral glioma model. A fluorescent dye, chloro-aluminum phthalocyanine tetrasulfonate (CIAIPcS,), was injected intravenously 24 hours before tumor resection. The animals underwent tumor resection under the operating microscope, guided by laser-induced fluorescence measurement in one group (Group 1) and visual assessment in the other (Group 2). The Group 1 rats had a significantly reduced volume of residual tumor following resection (0.5 _+0.2 cumm vs. 13.7 _+4.0 cu ram, mean _+ standard error of the mean, p < 0.02). Three of the nine animals in Group 1 were tumor-free at 2 weeks following resection, compared with none of the 10 rats in Group 2 (p < 0.05). Interference from brain autofluorescence was minimized using spectrally resolved detection and the C1A1PcS4dye, which has a 680-nm fluorescence peak significantly higher than the 470-nm autofluorescence peak of normal brain. Contrast ratios of up to 40:1 were found for glioma:normal brain fluorescence signals. Spatially-resolved spectra were acquired in approximately 5 seconds using a fiberoptic probe. This study demonstrates the ability of an intrao0erative laserinduced fluorescence system to detect tumor margins that could not be identified with the operating microscope. KEy WORDS phthalocyanine
9
9
fluorescence detection 9 glioma tumor resection 9 rat
URVIVALand quality of life in patients with supratentorial malignant astrocytoma has been shown to improve with a more extensive surgical resection? ,~8 However, the majority of these tumors do not have a distinct boundary, making complete resection difficult or impossible. Standard neurosurgical techniques ~5and intraoperative uitrasonography ~ have been recommended to achieve the goal of gross total resection. The use of intraoperative computerized tomography (CT) has been reported, ~6but is not widely available to neurosurgeons. In a recent series of 31 patients where total resection was attempted using standard microsurgical technique, early postoperative contrastenhanced CT showed significant residual tumor in more than one-third of the patients.-' Visual detection of neoplastic tissue in intracranial tumors following intravenous infusion of fluorescent compounds, such as hematoporphyrin derivative 2-' and phthalocyanine, has been reported previously. 33 The exposed brain was examined under ultraviolet light to localize and delineate the margin of the red fluorescent
S
J. Neurosurg. / Volume 76/April 1992
9
photoradiation therapy
glioma in five patients. Such visual detection schemes are limited by interference from the autofluorescence of the normal brain in the region (400-600 rim). In addition, maximum sensitivity would require examination in a nearly dark environment by dark adapted eyes. Laser-induced fluorescence of the photosensitizer, hematoporphyrin derivative, has been used by several investigators to distinguish various tumors, such as those in the lung, bladder, or colon, from normal tissues. 3-~'2~ Since the clearance of hematoporphyrin derivative from normal tissue is more rapid than from tumor, it acts as a tumor-localizing dye. When ultraviolet or violet wavelengths are used to excite laserinduced fluorescence of hematoporphyrin derivative, the natural fluorescence from tissue at the fluorescence peak of hematoporphyrin derivative (630 nm) reduces the contrast between the fluorescence from hematoporphyrin derivative in tumor and normal tissue. However, sensitive charge-coupled device cameras used in connection with analog or computer-aided image process-
679
W. S. Pooh, et al. ing techniques can enhance the contrast significantly. Such image enhancement techniques have been combined with digital video ftuorescence microscopy to study the distribution of hematoporphyrin derivative in a rat 9L gliosarcoma) More recently, a number of second-generation photosensitizers have been under investigation for use in photoradiation therapy. The phthalocyanines have several advantages over hematoporphyrin derivative for laser-induced fluorescence localization of tumors. Their quantum yield for fluorescence is significantly greater than hematoporphyrin derivative (0.52 vs. < 0.1) and their fluorescence peak lies at 680 nm, a level at which background due to tissue autofluorescence is low enough to be ignored. In addition, they clear the skin rapidly, avoiding the effects of skin phototoxicity that can lead to severe sunburn in patients who are exposed to sunlight after being injected with hematoporphyrin derivative. The distribution of aluminum sulfonated phthalocyanine in a number of rodent tumor models, including brain tumors, has been studied; 2s ratios of tumor to normal brain concentrations as high as 28:1 were found in malignant gliomas. In the present study, we have explored the laser-induced fluorescence technique to delineate the resection margin of a subcortically implanted rat glioma. Materials and Methods
Tumor Model The C6BAG cell line, a nitrosourea-induced rat glioma ~ containing the Escherichia coil lac Z (~-galactosidase) gene, 24'25was grown to confluence in monolayer culture over a 5-day period, dispersed with trypsin, and suspended for injection in Dulbecco's modified Eagle's medium at a concentration of 101 cells/5 ul. Male CD Fischer rats, each weighing between 150 and 175 gin, were anesthetized with intraperitoneal pentobarbital (50 to 75 mg/kg) and secured in a stereotactic frame.* A 5 x 5-mm right-sided occipitoparietal craniotomy was created using a dental drill. Tumor cells were injected subcortically in the posterior parietal region at the middle of the craniotomy via a 10-ml syringe+ mounted on the stereotactic frame. A No. 26 needle was advanced to puncture the dura to the point where the bevelled end was facing posteriorly just underneath the dura. The injection of tumor cells was carried out during a l-minute period and the needle was kept in place for 2 minutes, followed by gradual withdrawal over a span of 2 minutes. The craniotomy bone flap was discarded and the skin was closed with staples. Twenty percent of the animals died within 24 hours after injection due to either the anesthesia or postoperative intracerebral hematoma. All survivors developed sizable cortical tumors (tumor volume 26.7 _ 8.3 cu
* Small-animal stereotactic frame manufactured by David Kopf Instruments, Tujunga, California. t Syringe manufactured by Hamilton Co., Reno, Nevada. 680
mm, mean _+standard error of the mean) after 2 weeks and were used in the experiments.
Fluorescent Compound Chloro-aluminum phthalocyanine tetrasulfonate (C1A1PcS4):~ crystals were dissolved in phosphate-buffered saline (PBS) (pH 7.0) to a concentration of 10 mg/ml and administered at a rate of 5 mg/kg of body weight. This compound has been determined to contain approximately 95% of the tetrasulfonated compound, with the trisulfonate as the primary impurity]
Laser-Induced Fluorescence System The laser-induced fluorescence system is shown schematically in Fig. 1. The output of a pulsed nitrogen laserw (337-nm wavelength, 3-nsec pulse width, 10- to 20-Hz repetition rate, 200-t.d pulse energy) is coupled via a 1-in. focal length quartz lens into a single quartz optical fiber with a 600-urn core diameter.n Typical energies to tissue drop to about 40 uJ after attenuation with a 0.5-optical density neutral density filter and coupling, reflection, and fiber losses. Fluorescence from the tissue is passed back through the same optical fiber and is coupled to a quartz optical fiber bundle. The arrangement of fibers at the input of the fiber bundle is circular, while the output is linear with a 0.1 x 2.5mm dimension. The output of the bundle defines the entrance to the polychromator (f-number 3.8, focal length 0.275 m).* Fluorescence intensities for wavelengths between 300 and 800 nm are recorded using an intensified 1024-diode array controlled by a multichannel analyzer.t The intensifier is gated with 100-nsec pulses centered around the 3-nsee laser pulse. The multichannel analyzer allows the recording of a complete spectrum with each excitation pulse. Fifty spectra are collected and averaged for each measurement. Since the laser operates at 10 Hz, approximately 5 seconds are needed to acquire a spectrum.
Techniques of Extraction and Assay of CIAIPcS4 The technique used to extract C1A1PcS4 from tissue is similar to a previously reported protocol.9'2s'29Tumor samples were removed from areas of the rat brain used for laser-induced fluorescence measurements and placed in 2 ml of 0.1 N NaOH solution. The weight of the tumor specimens ranged from 3 to 80 rag. Samples of normal brain, each weighing between 300 to 600 rag, were taken from the contralateral cerebral cortex and immersed in 2 ml of 0.1 N NaOH solution. Samples
~tChloro-aluminum phthalocyanine tetrasulfonate fluorescent dye manufactured by Porphyrin Products, Logan, Utah. wVSL-337NDpulsed nitrogen laser manufactured by Laser Science, Inc., Cambridge, Massachusetts. II Superguideoptical fiber manufactured by Fiberguide Industries, Inc., Stirling, New Jersey. * Monospec 27 polychromator manufactured by Anaspec, Acton, Massachusetts. 1"OMA III multichannel analyzer manufactured by Princeton Applied Research, Princeton, New Jersey.
J. Neurosurg. / Volume 76/April, 1992
Delineation of tumor margins via laser-induced fluorescence
FIG. 1. Schematic diagram illustrating the laser-induced fluorescence system; f.l. = focal length quartz lens.
were homogenized using a microhomogenizer pestle, then left to stand in the dark overnight. Samples were rehomogenized with the pestle and centrifuged at 16,000 G for 30 minutes at 10*C. Fluorescence from the supernatant was recorded at wavelengths between 630 to 800 nm using a fluorimeter.r The excitation wavelength was 610 nm. The peak intensity values at 675 nm were recorded after subtracting a background spectrum obtained from the homogenate without CIA1PcS4. The C1AIPcS, tissue concentrations were found by comparing these peak intensities to values obtained from a calibration curve. The calibration curve was linear to 1 #g/ml, as measured using serial dilutions of a stock solution of known concentration, and was insensitive to the presence of brain homogenate in 0.1 N NaOH solution.
Histological Analysis All brain specimens were fixed in a 2% paraformaldehyde PBS solution, left overnight, and then dehydrated with increasing concentrations of sucrose solution (10%, 20%, and 30%) until they sank. Sections 40 um thick were cut directly onto gelatin-impregnated slides using a freezing microtome. The B-galactosidase histochemical staining technique was modified from the method of Turner and Cepko. 3~ Briefly, 5-bromo4-chloro-indolyl-~-D-galactoside (X-gal)w was prepared as a 2% solution in dimethyi sulfoxide. Mounted sections of the brain specimen were immersed overnight at room temperature in a solution of PBS (pH 7.3) containing 2 mM MgCI2, 35 mM K3Fe(CN)6, 35 mM K4Fe(CN)6, 0.01% sodium deoxycholate, 0.02% nitroprusside-40, and 0. I% X-gal, and then rinsed in PBS, counterstained with hematoxylin and eosin, and coverslipped with Permount.II r DM 1B fluorimeter manufactured by Spex Industries, Edison, New Jersey. wX-gal supplied by Fischer Biotech, Fairlawn, New Jersey. II Permount supplied by Fischer Scientific, Fairlawn, New Jersey.
J. Neurosurg. / Volume 76/April. 1992
FIG. 2. Graph showing the laser-induced fluorescence spectra for gray and white matter and glioma in rat brain obtained 24 hours after injection of C1A1PcS4. The autofluoreseence of normal brain tissue lies primarily in the 400- to 600-nm range and does not interfere with the dye fluorescence at 680 nm. The contrast between the glioma and white matter fluorescence is approximately 25:1 for this animal. Inset: The expanded C1AIPcS4 fluorescence signal from white and gray matter indicating the low noise on these signals. Resection was terminated when the glioma signal reached twice that of the background, approximately 4000 counts in this case. This tracing also indicates that resection could easily have been continued until the laser-induced fluorescence signal was at a level 20% above background before the noise on the background signal became significant.
Residual and recurrent tumors were identified by the characteristic bright blue cytoplasm. Tumor volume was documented in cu m m and estimated using the following formula: 4/3 x abe/2, where "a" was the depth, "b" the transverse diameter in a representative section showing maximum tumor, and "e" the longitudinal diameter, estimated by multiplying 40 um by the number of sections showing tumor. Results
Pharmacokinetics of CIAIPcS4 Distribution To determine the pharmacokinetics of C1A1PcS4 distribution in tumors and in normal brain, groups of four to 10 tumor-bearing rats were injected with C1AIPcS4 (5 mg/kg) via the tail vein at 4, 8, 16, and 24 hours and 2, 3, and 7 days prior to laser-induced fluorescence measurement of the tumor and contralateral normal brain. The animals were sacrificed by neck dislocation, and corresponding tissues were obtained for extraction and assay of CIAIPcS4 tissue concentration. Figure 2 shows a typical fluorescence spectrum for glioma, as well as white and gray matter. The fluorescence peak at 682 nm is well above the broad brain autofluorescence lying between 400 and 600 nm, illustrating the need to choose a marker dye that fluoresces in a spectral region free of autofluorescence. In practice, this means using dye with fluorescence wavelengths longer than 650 nm because there is usually some 681
W. S. Poon, et at.
FIG. 3. Graph showing the laser-induced fluorescence signal for normal brain (filled circles) and tumor (filled triangles) and the laser-induced fluorescence ratio between tumor and normal brain (open triangles) as a function of time. Fluorescence intensity is adequate for tumor detection from 8 hours to 7 days after C1AIPcS(administration. Note that the highest tumor:normal brain ratio (47:1) occurs at 24 hours.
FIG. 4. Graph showing the correlation between the laserinduced fluorescence signal and the absolute concentration of CIA1PcS4, as obtained from alkaline extraction.
TABLE 1
Concentration of CIAIPcS, as a function of time for normal brain and glioma tissue* Time (hrs)
CIAIPc,S( (~g/gm tissue) Normal
Glioma
Ratio (tumor.normal brain)
Brain 4 8 16 24 48 72 168
0.09 0.36 0.14 0.I I O.lO O. 12 0.15
0.68 5.03 3.91 5.15 3.72 4.05 2.04
8 14 28 47 37 34 14
* The calculation assumes a calibrationfactor of 11.7 kcounts/ug C1AIPcSdgmtissue.
autofluorescence at 630 nm due to porphyrins. Figure 2 (inset) shows the white and gray matter fluorescence on an expanded scale. The noise on these spectra at 682 nm is less than 10%, which is significant in determining how close to the background level that guided resection can proceed, as discussed below. Figure 3 shows the laser-induced fluorescence signal as a function of time for both normal brain and tumor tissue. At 24 hours after injection of C1AIPcS4, the laser-induced fluorescence signal from the tumor tissue was more than 40 times that of normal brain, indicating a very high contrast between the two tissues (Table 1). A fluorescence intensity level adequate for tumor detection was present from 8 hours to 7 days after C1AIPcS4 administration (Fig. 3). Figure 4 shows the correlation between the laser-induced fluorescence signal and the absolute concentration of C1A1PcS4, as obtained from the extraction measurements (r 2 = 0.84, slope = 11.7 kcounts/ #g ClA1PcSdgm tissue, intercept = 0).
682
FIG. 5. Diagram showing the volume for tumors resected via visual (upward triangles) and laser-induced fluorescence (LIF, reversed triangles) guidance and for unresected tumors (diamonds) (controls) in animals sacrificed immediately after resection (le)2panel) and 2 weeks after resection (right panel). Thefilled circles with error bars show the mean tumor volume _+standard error of the means for the three groups. Note that, for the laser-induced fluorescence-guided resections, two animals in the immediate-sacrifice group show tumor volumes at the estimated detection limit of 1.5 x 10-scu mm as do three animals in the group sacrificed 2 weeks after resection.
Tumor Growth To determine the extent of tumor growth in this rat glioma model, no tumor resection was performed in 18 control animals. Ten animals were sacrificed at 2 weeks and eight at 4 weeks after implantation of glioma cells. The tumors were trimmed to the level of the cortical surface) 2 The rat brains were removed and fixed in a 2% paraformaldehyde solution prior to sectioning and histochemical staining for estimation of tumor volume. The mean volume of the "unresected tumors" was 26.7 +_ 8.3 c u m m for the rats sacrificed at 2 weeks and 119.9 _+ 31.7 c u m m for those killed at 4 weeks (t -- 2.6, p < 0.02) (Fig. 5). Figure 6A shows a typical unreJ. Neurosurg. / Volume 76/April, 1992
Delineation of tumor margins via laser-induced fluorescence
FIG. 6. Low-power photomicrographs of rat glioma specimens. The C1AIPcS4glioma cells, stained bright blue by the galactosidase histochemical technique, appear dark in these photographs (solid arrow.~j. The tumor bed is marked by open arrows. Bar = 0.1 ram. A: A typical unresected tumor (arrow) at 2 weeks after ~mplantation of glioma cells. B-D: The nodules of residual tumor were missed by the laser-induced fluorescence probe at random in four animals (B). Residual tumor was missed in the cortical margin of the tumor bed in two rats (C) because laser-induced fluorescence measurements were carried out systematically only within the tumor bed. In one animal, infiltrative tumor at the depth of the tumor bed was not detected by la~er-induced fluorescence measurement (D).
sected tumor at 2 weeks after implantation of the glioma cells. Visual vs. Laser-Induced Fluorescence-Guided Resection In order to test the completeness of visual versus laser-induced fluorescence-guided resection, tumors were implanted in 19 rats. The C1AIPcS4 dye (5 mg/kg) was administered intravenously 24 hours prior to the experiment. Tumor resection was carded out by one of us (W.S.P.) under the operating microscope, with gentle suction using a l-mm glass pipette and bipolar cautery. The tumor appeared gray, whereas normal brain was white. Complete resection via visual guidance was confirmed by a second observer (K.T.S.). To ensure that there was no bias in the randomization procedure, we measured the mean laser-induced fluorescence signals of the four quadrants and the center of the tumor bed after visual resection but prior to random assignment J. Neurosurg. / Volume 76/April, 1992
into test groups. There was no statistical difference between any treatment group prior to laser-induced fluorescence-guided resection (Table 2). The uniformity of the mean laser-induced fluorescence signal for the four groups indicates quantitatively that the visually guided resections were performed in a consistent manner. It does not address the question of whether another neurosurgeon would have obtained fewer or more laserinduced fluorescence counts (indications of residual tumor at the margin), but our goal was only to demonstrate that laser-induced fluorescence could reduce the number of tumor cells remaining at the margin. One-half of the rats were randomly selected for further resection of residual tumor under laser-induced fluorescence guidance. Resection continued until the laser-induced fluorescence signal was reduced to approximately twice the background level of 1000 to 2000 counts; this quantitative criterion served to give the neurosurgeon an objective means of deciding when to 6@3
W. S. Poon, et aL TABLE 2 Laser-inducedfluorescence signal of the fimr quadrants and the center of the tumor bed*
Tr~tment
Laser-InducedFluorescence Signal (kcounts)
Sacrificed Sacrificed Immediately After 2 Wks visual resection 9.47 _+3.45t 9.29 _+3.67r (10 rats) (I0 rats) laser-induced fluorescence 9.21 ___1.63t 8.24 _+3.32~t guided resection (9 rats) (9 rats) * Data obtainedfollowingvisualresectionbut prior to randomizing into varioustreatment grou~. Assumingthat a laser-induced fluorescence signal abovebackgroundin tumor beds represents the presence of residual tumor, comparablelaser-induced fluorescencesignals for all four groups suggestcomparableextent of tumor resection in each of the four groupsprior to randomization. Valuesexpressedas means -+ standard error of the means. Significanceof difference: p < 0.5; tt = 0.1, r = 0.3.
terminate resection and avoided surgeon bias. This was a very conservative resection criterion; the spectra in Fig. 2 show that the resection could have been continued until the laser-induced fluorescence signal was at a level about 20% above background before noise on the signal became a problem. All 19 animals were sacrificed within 30 minutes after the surgical procedure, and the brains were removed for histological study and estimation of residual tumor volume. The nine rats that underwent laserinduced fluorescence-guided resection had a mean residual tumor volume of 0.5 + 0.2 cu mm (range 0 to 2.0 cu mm), which was significantly less than that of the 10 rats that underwent visual resection (13.7 _+_4.09 cu mm, range 0.1 to 36.5 cu mm; t = 2.5, p < 0.05) (Fig. 5). The detection limit of residual tumor was estimated as 1.5 • 10-f eu mm by calculating the volume of the region using the area of a cell (diameter 20 um) and the thickness of the microscope sections (40 urn). The pattern of failure of laser-induced fluorescence-guided resection in leaving residual tumor in the tumor bed is demonstrated by Fig. 6.
Tumor Regrowth To test whether laser-induced fluorescence-guided resection altered the rate of tumor regrowth in the rat giioma model, the procedure described above was repeated in 24 animals, making allowance for a 20% procedure-related mortality rate. Following tumor resection, these animals were permitted to survive for 2 weeks in order to compare the recurrent tumor volume between the group with laser-induced fluorescenceguided resection (Group 1) and the group with visual resection (Group 2). Three animals in Group 1 and two in Group 2 died immediately postoperatively, leaving nine and 10 animals, respectively, for analysis. The mean recurrent tumor volumes were 36.1 + 18.6 cu mm (range 0 to 175.5 cu mm) in Group 1 rats and 66.5 _+ 24.0 cu m m (range 9.0 to 175.8 cu mm) in Group 2 684
rats (t = 1.3, 0.1 < p < 0.5). The growth rate of the tumors, assuming exponential growth, was proportional to the natural logarithm of the ratio of tumor volume measured 2 weeks after resection of tumor volume calculated immediately after resection. The growth rate of the tumors in Group 1 rats was about 2.7 times that of the tumors in Group 2 animals. This may reflect one of several factors. The growth rate may be dependent in part upon the amount of space available; immediately after resection, the average volume of the residual tumors in Group 1 animals was about 27 times smaller than the average volume of those in Group 2 rats and therefore the larger tumor was closer to the size limits imposed by the volume available for growth. Additionally, as gliomas grow, the central areas of larger tumors may become hypoxic or necrotic and show reduced growth. Two of the nine animals in Group 1 died at 10 days and three of the 10 in Group 2 died at 10 to 12 days (X2 = 0.15, p > 0.5). Three Group 1 rats were tumorfree at 2 weeks, whereas every animal in Group 2 had gross tumor recurrence (X2 = 4.0, p < 0.05). In all animals, the mean laser-induced fluorescence signals of the four quadrants and the center of the tumor bed after visual resection and before randomization into groups had almost identical values, suggesting a comparable extent of residual tumor (Table 2).
Discussion This is the first report of the use of laser-induced fluorescence to increase the accuracy with which rat brain-tumor margins are defined during resection, and, to our knowledge, the first use of laser-induced fluorescence in vivo to delineate tumor margins during resection in any tumor model.
Reduction of Tumor Volume Substantial reduction of brain-tumor volume prolongs life, improves the quality of survival, and may make adjuvant therapy more effective, z~3`~s'~s'19'3~'3:We have demonstrated that laser-induced fluorescence detection of phthalocyanine in a rat glioma model is effective in maximizing the extent of tumor resection; however, in those animals that were tumor-free at 2 weeks, we have not excluded the possibility that tumor recurrence from small numbers of infiltrating tumor cells might occur at a later time. Indeed, in humans, infiltrating glioma cells are the norm. Thus, this technique is not aimed at tumor cure but rather at extensive defined reduction of the bulk of the tumor. Nonetheless, the ability to resect the bulk of a glioma may allow for localized forms of therapy, 2j'2g including implants of chemotherapeutic agents ~~ or other locally applied biological agents. '4'~ Potential Advantages of the Technique There are several differences between this work and prior studies of brain-tumor fluorescence. The autofluorescence spectrum of normal brain, excited by 337 J. Neurosurg / Volume 76/April, 1992
Delineation of tumor margins via laser-induced fluorescence nm radiation, is concentrated between 350 and 650 nm. This background fluorescence severely hampers tumor localization using either visual or broad-band photographic detection. Since our detection system uses a gating polychromator to reject the broad fluorescence emission of normal brain, only the weak background fluorescence of dye in normal brain at our 680-rim measurement wavelength must be corrected for. The spectral selectivity of our detection system manifests itself in the high 40:1 ratio of tumor:normal brain fluorescence. Additionally, some previously used fluorescent dyes, such as fluorescein and hematoporphyrin derivative, have fluorescence emission peaks overlapping the autofluorescence of normal brain and are not optimum for delineating tumor margins. While we used the photosensitizer CIA1PcS4, similar results would be expected from other sensitizers emitting at wavelengths greater than 650 nm with similar quantum efficiencies. Selectivity is primarily the result of the blood-brain barrier, which prevents the sensitizer from entering the brain except in tumor regions, where there is a proliferation of immature leaky neovessels. Clinical Applications This system should be easily applied to surgery. The present unit is mounted on a computer trolley and is therefore portable. The fiberoptic probe and the short (5-sccond) data acquisition time allow multiple measurements to be made in a resection cavity. Since the information needed to guide resection involves only the fluorescence intensity at a single wavelength, the present polychromator-multichannel analyzer could be replaced by a simpler, less expensive, and more compact system. Only the 600-~m quartz fiber needs to be sterilized. Potential applications include identification of tumor resection margins, localization of subeortical tumor during open operation, and confirmation of target in stereotactic tumor biopsies. Additionally, the technique may enhance the effectiveness of photodynamic therapy by reducing residual tumor volume, thus aiding light penetration into a thick tumor, 19and by targeting localized areas suggestive of infiltrative tumor rather than treating the entire tumor bed. Finally, since the laser-induced fluorescence signals are well correlated with tissue concentrations of CIA1PcS4 below 10 gg/gm tissue, this technique may be used to correlate dinical results with drug concentration and light dose. This real-time information may allow the development of a drug/light-dose protocol that is tumoricidal but causes minimal brain damage; it may also permit a decision to forego treatment when the drug concentration in the tumor is not optimal. Limitations and Areas for Further Development While these techniques may be readily applied in the clinical setting, caution must be exercised in extrapolating these animal studies to neurosurgical practice. Selective uptake of porphyrin compounds is not tumorJ. Neurosurg. / Volume 76/April, 1992
specific. Hematoporphyrin fluorescence was also demonslrated in traumatized brain and normal pituitary stalk s~ and choroid plexus tissue. 26 However, in one study using tritiated hematoporphyrin derivative in a dog glioma model 26 the highest concentrations of the fluorescent drug were found in the cytoplasm of the tumor cells, while the concentration in peritumoral edematous brain or brain adjacent to tumor was at the same low level as in normal brain. This concentration was slightly raised in a rat 9L gliosarcoma studied using digital video fluorescence microscopyJ These results may reflect the different natures of the tumor involved. In general, the uptake of photosensitizer in brain adjacent to tumor can be different from that in normal brain; changes in the vasculature of brain adjacent to tumor can affect its uptake. Uptake to phthalocyanines in brain adjacent to tumor has not been studied. However, it should again be pointed out that CIAIPcS4 has the following advantages over hematoporphyrin derivative: 1'23 the skin phototoxicity is negligible and the longer wavelength of the fluorescence band at 680 nm reduces interference from tissue autofluorescence. Our criterion for discontinuing resection (reducing the laser-induced fluorescence signal to twice the background) was very conservative and may not be optimal; resection until the laser-induced fluorescence signal is barely detectable above the background may be better. Current development of photosensitizing drugs for photodynamic therapy emphasizes chromophores absorbing at 650 to 800 nm, a wavelength range that allows deeper light penetration. Such photosensitizers would be even less sensitive to interference from autofluorescence than CIAIPcS4. The pattern of failure of laser-induced fluorescence detection of residual tumor at the resection margin (Fig. 6) suggests that more point measurements at and within the vicinity of the tumor bed and/or a real-time fluorescent imaging technique s may improve its performance. In preliminary experiments, we have been able to obtain images of the tumor bed using a standard commercial video camera in conjunction with an electronic system that allows the integration of the video signal for periods of up to 30 seconds. Conclusions
This study has demonstrated that intraoperative laser-induced fluorescence measurement is an effective means of delineating tumor margin during resection in glioma-beafing rats injected with CIA1PcS4, resulting in significant reduction of tumor volume compared with standard microsurgical resection. These studies need to be confirmed and extended in other animal models, and additional studies should be performed before this technique can he applied to humans. References
1. AbernatheyCD, Anderson RE, Kooistra KL, et al: Activity of phthalocyanine photosensitizers against human glioblastoma in vitro. Neurosurgery21:468-473, 1987 685
W. S. Poon, et al. 2. Ammirati M, Vick N, Liao Y, et al: Effect of the extent of surgical resection on survival and quality of life in patients with supratentorial glioblastoma and anaplastic astrocytomas. Neurosurgery 21:201-206, 1987 3. Ankerst J, Montan S, Svanberg K, et al: Laser-induced fluorescence studies of hematoporphyrin derivative (HPD) in normal and malignant tissue of rat. Appl Spectr 38:890-896, 1984 4. Balchum O J, Profio AE, Razum N: Ratioing fluorometer probe for localizing carcinoma in situ in bronchi of the lung. Photaehem Photohiol 46:887-891, 1987 5. Baumgartner R, Fisslinger H, Jocham D, et al: A fluorescence imaging device for endoscopic detection of early stage cancer - - instrumental and experimental studies. Photoehem Photobial 46:759-763, 1987 6. Benda P, Someda K, Messer J. et al: Morphological and immunochemical studies of rat glial tumors and clonal strains propagated in culture. J Neurosurg 34:310-323, 1971 7. Berg K, Bummer JC, Moan J: Evaluation of sulfonated aluminum phthalocyanines for use in photochemotherapy. Cellular uptake studies. Cancer Lett 44:7-15, 1989 8. Boggan JE, Walter R, Edwards MSB, el al: Distribution of hematoporphyrin derivative in the rat 9L gliosareoma brain tumor analyzed by digital video fluorescence microscopy. J Neurosurg 61:1113-1119, 1984 9. Bown SG, Tralau CJ, Coleridge Smith PD, et al: Photodynamic therapy with porphyrin and phthalocyanine sensitisation: quantitative studies in normal rat liver. Br J Cancer 54:43-52, 1986 10. Brem H, Mahaley MS Jr, Vick NA, el al: Interstitial chemotherapy with drug polymer implants for the treatment of recurrent gliomas. J Neurosurg 74:441-446, 1991 1I. Chandler WF, Knake JE, McGillieuddy JE, et al: Inlraoperative use of real-time ultrasonography in neurosurgery. J Neurosurg 57:157-163, 1982 12. Cheng MK, McKean J, Boisvert D, et al: Effects of pho~.oradiation therapy of normal rat brain. Neurosurgery 15:804-810, 1984 13. Cooper JS, Borok TL, Ransohoff J, et al: Malignant gliurea. Results of combined modality therapy. JAMA 248: 62-65, 1982 14. Ezzedine ZD, Martuza RL, Short MP, et al: Selective killing of giioma cells in culture and in vivo following retrovirus transfer of the herpes simplex virus thymidine kinase gene. New Biolog 3:608-614, 1991 15. Harsh GR IV, Wilson CB: Neuroepithelial tumors of the adult brain, in Youmans JR (ed): Neurological Surgery, ed 3. Philadelphia: WB Saunders, 1990, pp 3040-3136 16. Kelly PJ, Kall BA, Goerss SJ: Special stereotactic techniques: stereotactic laser resection of deep-seated tumors, in Heilbrun MP (ed): Concepts in Neurosurgery, Vol 2: Stereotactie Neurosurgery. Baltimore: Williams & Wilkins, 1988, pp 233-240 17. Martuza RL, Malick A, Markert JM, el al: Experimental therapy of human glioma by means of a genetically engineered virus mutant. Science 252:854-856, 1991 18. MRC (Medical Research Council) Brain Tumor Working Party: Prognostic factors for high grade malignant glioma: development of a prognostic index. J Neurooncol 9: 47-55, 1990 19. Muller PJ, Wilson BC: Photodynamic therapy of malig-
686
20. 21. 22.
23. 24.
25. 26.
27. 28.
29.
30. 31. 32. 33.
nant primary brain tumours: clinical effects, post-operative ICP, and light penetration of the brain. Photochem Photobiol 46:929-935, 1987 Profio AE: Laser excited fluorescence of hematoporphyfin derivative for diagnosis of cancer. IEEE J Quantum Electron QE-20 12:1502-1507, 1984 Profio AE, Doiron DR, King EG: Laser fluorescence bronchoscopy for localization of occult lung tumors. Meal Phys 6:523-525, 1979 Rassmussen-Taxdal DS, Ward GE, Figge FHJ: Fluorescence of human lymphatic and cancer tissues following high doses of intravenous hematoporphyrin. Cancer 8: 78-81, 1955 Sandeman DR, Bradford R, Buxton P, et al: Selective necrosis of malignant gliomas in mice using photodynamic therapy. Br J Cancer 55:647-649, 1987 Shimohama S, Rosenberg MB, Fagan AM, et at: Grafting genetically modified ceils into the rat brain: characteristics of E. coil beta-galactosidaseas a reporter gene. Brain Res Mol Brain Res 5:271-278, 1989 Short MP, Choi BB, Lee JK, et al: Gene delivery to glioma cells in rat brain by grafting of a retrovirus packaging cell line. d Neurosci Res 27:427-433, 1990 Steichen JD, Dashner K, Martuza RL: Distribution of hematoporphyrin derivative in canine glioma following intraneoplastic and intrapefitoneal injection. J Neurosurg 65:364-369, 1986 Svanberg K, Kjellrn E, Ankerst J, et al: Fluorescence studies of bematoporphyrin derivative in normal and malignant rat tissue. Cancer Res 46:3803-3808, 1986 Tralau CJ, Ban" H, Sandeman DR, et at: Aluminum sulfonated phthalocyanine distribution in rodent tumors of the colon, brain and pancreas. Photochem Photohiol 46:777-781, 1987 Tralau CJ, MacRobert AJ, Coleridge-Smith PD, et al: Photodynamic therapy with phthalocyanine sensitisation: quantitative studies in a transplantable rat fibrosarcoma. Br J Cancer 55:389-395, 1987 Turner DL, Cepko CL: A common progenitor for neurons and glia persists in rat retina late in develoment. Nature 328:131-136, 1987 Walker MD: Malignant glioma, in Wilson CB, Huff JT {eds): Current Surgical Management of Neurologic Disease. New York: Churchill Livingstone, 1980, pp 72-78 Walker RW, Posner JB: Central nervous system neoplasm. Curr Neurol 5:285-322, 1984 Wise BL, Taxdal DR: Studies of the blood-brain barrier utilizing hematoporphyrin. Brain Res 4:387-389, 1967
Manuscript received January 30, 1991. Accepted in final form August 16, 1991. This research was supported in part by the SDIO-MFEL program under Office of Naval Research Contract N0001486-K-0117 and by a grant from Neurofibromatosis, Inc., Massachusetts. Dr. Poort was the recipient of a grant from the Faculty of Medicine, Chinese University of Hung Kong. Address reprint requests to: Robert L. Martaza, M.D., Department of Neurosurgery, Georgetown University Medical Center, 3800 Reservoir Road NW, Washington, D.C. 200(37.
J. Neurosurg. / Votume 76/April, 1992
Laser-induced fluorescence: experimental intraoperative delineation of tumor resection margins WAI S. POON, M.B.,CH.B., F.R.C.S., KEVlN T. SCHOMACKER,PH.D., THOMAS F. DEUTSCH, PH.D., AND ROBERT L. MARTUZA, M.D.
Department of Surgery (Neurosurgery) and Molecular Neurogenetics Laboratory, Wellman Laboratories of Photomedicine, and Department of Dermatology, Massachusetts General Hospital, Harvard Medical School, Boston. Massachusetts v, The ability of laser-induced fluorescence spectroscopy to delineate tumor margins intraoperatively was studied using a rat intracerebral glioma model. A fluorescent dye, chloro-aluminum phthalocyanine tetrasulfonate (CIAIPcS,), was injected intravenously 24 hours before tumor resection. The animals underwent tumor resection under the operating microscope, guided by laser-induced fluorescence measurement in one group (Group 1) and visual assessment in the other (Group 2). The Group 1 rats had a significantly reduced volume of residual tumor following resection (0.5 _+0.2 cumm vs. 13.7 _+4.0 cu ram, mean _+ standard error of the mean, p < 0.02). Three of the nine animals in Group 1 were tumor-free at 2 weeks following resection, compared with none of the 10 rats in Group 2 (p < 0.05). Interference from brain autofluorescence was minimized using spectrally resolved detection and the C1A1PcS4dye, which has a 680-nm fluorescence peak significantly higher than the 470-nm autofluorescence peak of normal brain. Contrast ratios of up to 40:1 were found for glioma:normal brain fluorescence signals. Spatially-resolved spectra were acquired in approximately 5 seconds using a fiberoptic probe. This study demonstrates the ability of an intrao0erative laserinduced fluorescence system to detect tumor margins that could not be identified with the operating microscope. KEy WORDS phthalocyanine
9
9
fluorescence detection 9 glioma tumor resection 9 rat
URVIVALand quality of life in patients with supratentorial malignant astrocytoma has been shown to improve with a more extensive surgical resection? ,~8 However, the majority of these tumors do not have a distinct boundary, making complete resection difficult or impossible. Standard neurosurgical techniques ~5and intraoperative uitrasonography ~ have been recommended to achieve the goal of gross total resection. The use of intraoperative computerized tomography (CT) has been reported, ~6but is not widely available to neurosurgeons. In a recent series of 31 patients where total resection was attempted using standard microsurgical technique, early postoperative contrastenhanced CT showed significant residual tumor in more than one-third of the patients.-' Visual detection of neoplastic tissue in intracranial tumors following intravenous infusion of fluorescent compounds, such as hematoporphyrin derivative 2-' and phthalocyanine, has been reported previously. 33 The exposed brain was examined under ultraviolet light to localize and delineate the margin of the red fluorescent
S
J. Neurosurg. / Volume 76/April 1992
9
photoradiation therapy
glioma in five patients. Such visual detection schemes are limited by interference from the autofluorescence of the normal brain in the region (400-600 rim). In addition, maximum sensitivity would require examination in a nearly dark environment by dark adapted eyes. Laser-induced fluorescence of the photosensitizer, hematoporphyrin derivative, has been used by several investigators to distinguish various tumors, such as those in the lung, bladder, or colon, from normal tissues. 3-~'2~ Since the clearance of hematoporphyrin derivative from normal tissue is more rapid than from tumor, it acts as a tumor-localizing dye. When ultraviolet or violet wavelengths are used to excite laserinduced fluorescence of hematoporphyrin derivative, the natural fluorescence from tissue at the fluorescence peak of hematoporphyrin derivative (630 nm) reduces the contrast between the fluorescence from hematoporphyrin derivative in tumor and normal tissue. However, sensitive charge-coupled device cameras used in connection with analog or computer-aided image process-
679
W. S. Pooh, et al. ing techniques can enhance the contrast significantly. Such image enhancement techniques have been combined with digital video ftuorescence microscopy to study the distribution of hematoporphyrin derivative in a rat 9L gliosarcoma) More recently, a number of second-generation photosensitizers have been under investigation for use in photoradiation therapy. The phthalocyanines have several advantages over hematoporphyrin derivative for laser-induced fluorescence localization of tumors. Their quantum yield for fluorescence is significantly greater than hematoporphyrin derivative (0.52 vs. < 0.1) and their fluorescence peak lies at 680 nm, a level at which background due to tissue autofluorescence is low enough to be ignored. In addition, they clear the skin rapidly, avoiding the effects of skin phototoxicity that can lead to severe sunburn in patients who are exposed to sunlight after being injected with hematoporphyrin derivative. The distribution of aluminum sulfonated phthalocyanine in a number of rodent tumor models, including brain tumors, has been studied; 2s ratios of tumor to normal brain concentrations as high as 28:1 were found in malignant gliomas. In the present study, we have explored the laser-induced fluorescence technique to delineate the resection margin of a subcortically implanted rat glioma. Materials and Methods
Tumor Model The C6BAG cell line, a nitrosourea-induced rat glioma ~ containing the Escherichia coil lac Z (~-galactosidase) gene, 24'25was grown to confluence in monolayer culture over a 5-day period, dispersed with trypsin, and suspended for injection in Dulbecco's modified Eagle's medium at a concentration of 101 cells/5 ul. Male CD Fischer rats, each weighing between 150 and 175 gin, were anesthetized with intraperitoneal pentobarbital (50 to 75 mg/kg) and secured in a stereotactic frame.* A 5 x 5-mm right-sided occipitoparietal craniotomy was created using a dental drill. Tumor cells were injected subcortically in the posterior parietal region at the middle of the craniotomy via a 10-ml syringe+ mounted on the stereotactic frame. A No. 26 needle was advanced to puncture the dura to the point where the bevelled end was facing posteriorly just underneath the dura. The injection of tumor cells was carried out during a l-minute period and the needle was kept in place for 2 minutes, followed by gradual withdrawal over a span of 2 minutes. The craniotomy bone flap was discarded and the skin was closed with staples. Twenty percent of the animals died within 24 hours after injection due to either the anesthesia or postoperative intracerebral hematoma. All survivors developed sizable cortical tumors (tumor volume 26.7 _ 8.3 cu
* Small-animal stereotactic frame manufactured by David Kopf Instruments, Tujunga, California. t Syringe manufactured by Hamilton Co., Reno, Nevada. 680
mm, mean _+standard error of the mean) after 2 weeks and were used in the experiments.
Fluorescent Compound Chloro-aluminum phthalocyanine tetrasulfonate (C1A1PcS4):~ crystals were dissolved in phosphate-buffered saline (PBS) (pH 7.0) to a concentration of 10 mg/ml and administered at a rate of 5 mg/kg of body weight. This compound has been determined to contain approximately 95% of the tetrasulfonated compound, with the trisulfonate as the primary impurity]
Laser-Induced Fluorescence System The laser-induced fluorescence system is shown schematically in Fig. 1. The output of a pulsed nitrogen laserw (337-nm wavelength, 3-nsec pulse width, 10- to 20-Hz repetition rate, 200-t.d pulse energy) is coupled via a 1-in. focal length quartz lens into a single quartz optical fiber with a 600-urn core diameter.n Typical energies to tissue drop to about 40 uJ after attenuation with a 0.5-optical density neutral density filter and coupling, reflection, and fiber losses. Fluorescence from the tissue is passed back through the same optical fiber and is coupled to a quartz optical fiber bundle. The arrangement of fibers at the input of the fiber bundle is circular, while the output is linear with a 0.1 x 2.5mm dimension. The output of the bundle defines the entrance to the polychromator (f-number 3.8, focal length 0.275 m).* Fluorescence intensities for wavelengths between 300 and 800 nm are recorded using an intensified 1024-diode array controlled by a multichannel analyzer.t The intensifier is gated with 100-nsec pulses centered around the 3-nsee laser pulse. The multichannel analyzer allows the recording of a complete spectrum with each excitation pulse. Fifty spectra are collected and averaged for each measurement. Since the laser operates at 10 Hz, approximately 5 seconds are needed to acquire a spectrum.
Techniques of Extraction and Assay of CIAIPcS4 The technique used to extract C1A1PcS4 from tissue is similar to a previously reported protocol.9'2s'29Tumor samples were removed from areas of the rat brain used for laser-induced fluorescence measurements and placed in 2 ml of 0.1 N NaOH solution. The weight of the tumor specimens ranged from 3 to 80 rag. Samples of normal brain, each weighing between 300 to 600 rag, were taken from the contralateral cerebral cortex and immersed in 2 ml of 0.1 N NaOH solution. Samples
~tChloro-aluminum phthalocyanine tetrasulfonate fluorescent dye manufactured by Porphyrin Products, Logan, Utah. wVSL-337NDpulsed nitrogen laser manufactured by Laser Science, Inc., Cambridge, Massachusetts. II Superguideoptical fiber manufactured by Fiberguide Industries, Inc., Stirling, New Jersey. * Monospec 27 polychromator manufactured by Anaspec, Acton, Massachusetts. 1"OMA III multichannel analyzer manufactured by Princeton Applied Research, Princeton, New Jersey.
J. Neurosurg. / Volume 76/April, 1992
Delineation of tumor margins via laser-induced fluorescence
FIG. 1. Schematic diagram illustrating the laser-induced fluorescence system; f.l. = focal length quartz lens.
were homogenized using a microhomogenizer pestle, then left to stand in the dark overnight. Samples were rehomogenized with the pestle and centrifuged at 16,000 G for 30 minutes at 10*C. Fluorescence from the supernatant was recorded at wavelengths between 630 to 800 nm using a fluorimeter.r The excitation wavelength was 610 nm. The peak intensity values at 675 nm were recorded after subtracting a background spectrum obtained from the homogenate without CIA1PcS4. The C1AIPcS, tissue concentrations were found by comparing these peak intensities to values obtained from a calibration curve. The calibration curve was linear to 1 #g/ml, as measured using serial dilutions of a stock solution of known concentration, and was insensitive to the presence of brain homogenate in 0.1 N NaOH solution.
Histological Analysis All brain specimens were fixed in a 2% paraformaldehyde PBS solution, left overnight, and then dehydrated with increasing concentrations of sucrose solution (10%, 20%, and 30%) until they sank. Sections 40 um thick were cut directly onto gelatin-impregnated slides using a freezing microtome. The B-galactosidase histochemical staining technique was modified from the method of Turner and Cepko. 3~ Briefly, 5-bromo4-chloro-indolyl-~-D-galactoside (X-gal)w was prepared as a 2% solution in dimethyi sulfoxide. Mounted sections of the brain specimen were immersed overnight at room temperature in a solution of PBS (pH 7.3) containing 2 mM MgCI2, 35 mM K3Fe(CN)6, 35 mM K4Fe(CN)6, 0.01% sodium deoxycholate, 0.02% nitroprusside-40, and 0. I% X-gal, and then rinsed in PBS, counterstained with hematoxylin and eosin, and coverslipped with Permount.II r DM 1B fluorimeter manufactured by Spex Industries, Edison, New Jersey. wX-gal supplied by Fischer Biotech, Fairlawn, New Jersey. II Permount supplied by Fischer Scientific, Fairlawn, New Jersey.
J. Neurosurg. / Volume 76/April. 1992
FIG. 2. Graph showing the laser-induced fluorescence spectra for gray and white matter and glioma in rat brain obtained 24 hours after injection of C1A1PcS4. The autofluoreseence of normal brain tissue lies primarily in the 400- to 600-nm range and does not interfere with the dye fluorescence at 680 nm. The contrast between the glioma and white matter fluorescence is approximately 25:1 for this animal. Inset: The expanded C1AIPcS4 fluorescence signal from white and gray matter indicating the low noise on these signals. Resection was terminated when the glioma signal reached twice that of the background, approximately 4000 counts in this case. This tracing also indicates that resection could easily have been continued until the laser-induced fluorescence signal was at a level 20% above background before the noise on the background signal became significant.
Residual and recurrent tumors were identified by the characteristic bright blue cytoplasm. Tumor volume was documented in cu m m and estimated using the following formula: 4/3 x abe/2, where "a" was the depth, "b" the transverse diameter in a representative section showing maximum tumor, and "e" the longitudinal diameter, estimated by multiplying 40 um by the number of sections showing tumor. Results
Pharmacokinetics of CIAIPcS4 Distribution To determine the pharmacokinetics of C1A1PcS4 distribution in tumors and in normal brain, groups of four to 10 tumor-bearing rats were injected with C1AIPcS4 (5 mg/kg) via the tail vein at 4, 8, 16, and 24 hours and 2, 3, and 7 days prior to laser-induced fluorescence measurement of the tumor and contralateral normal brain. The animals were sacrificed by neck dislocation, and corresponding tissues were obtained for extraction and assay of CIAIPcS4 tissue concentration. Figure 2 shows a typical fluorescence spectrum for glioma, as well as white and gray matter. The fluorescence peak at 682 nm is well above the broad brain autofluorescence lying between 400 and 600 nm, illustrating the need to choose a marker dye that fluoresces in a spectral region free of autofluorescence. In practice, this means using dye with fluorescence wavelengths longer than 650 nm because there is usually some 681
W. S. Poon, et at.
FIG. 3. Graph showing the laser-induced fluorescence signal for normal brain (filled circles) and tumor (filled triangles) and the laser-induced fluorescence ratio between tumor and normal brain (open triangles) as a function of time. Fluorescence intensity is adequate for tumor detection from 8 hours to 7 days after C1AIPcS(administration. Note that the highest tumor:normal brain ratio (47:1) occurs at 24 hours.
FIG. 4. Graph showing the correlation between the laserinduced fluorescence signal and the absolute concentration of CIA1PcS4, as obtained from alkaline extraction.
TABLE 1
Concentration of CIAIPcS, as a function of time for normal brain and glioma tissue* Time (hrs)
CIAIPc,S( (~g/gm tissue) Normal
Glioma
Ratio (tumor.normal brain)
Brain 4 8 16 24 48 72 168
0.09 0.36 0.14 0.I I O.lO O. 12 0.15
0.68 5.03 3.91 5.15 3.72 4.05 2.04
8 14 28 47 37 34 14
* The calculation assumes a calibrationfactor of 11.7 kcounts/ug C1AIPcSdgmtissue.
autofluorescence at 630 nm due to porphyrins. Figure 2 (inset) shows the white and gray matter fluorescence on an expanded scale. The noise on these spectra at 682 nm is less than 10%, which is significant in determining how close to the background level that guided resection can proceed, as discussed below. Figure 3 shows the laser-induced fluorescence signal as a function of time for both normal brain and tumor tissue. At 24 hours after injection of C1AIPcS4, the laser-induced fluorescence signal from the tumor tissue was more than 40 times that of normal brain, indicating a very high contrast between the two tissues (Table 1). A fluorescence intensity level adequate for tumor detection was present from 8 hours to 7 days after C1AIPcS4 administration (Fig. 3). Figure 4 shows the correlation between the laser-induced fluorescence signal and the absolute concentration of C1A1PcS4, as obtained from the extraction measurements (r 2 = 0.84, slope = 11.7 kcounts/ #g ClA1PcSdgm tissue, intercept = 0).
682
FIG. 5. Diagram showing the volume for tumors resected via visual (upward triangles) and laser-induced fluorescence (LIF, reversed triangles) guidance and for unresected tumors (diamonds) (controls) in animals sacrificed immediately after resection (le)2panel) and 2 weeks after resection (right panel). Thefilled circles with error bars show the mean tumor volume _+standard error of the means for the three groups. Note that, for the laser-induced fluorescence-guided resections, two animals in the immediate-sacrifice group show tumor volumes at the estimated detection limit of 1.5 x 10-scu mm as do three animals in the group sacrificed 2 weeks after resection.
Tumor Growth To determine the extent of tumor growth in this rat glioma model, no tumor resection was performed in 18 control animals. Ten animals were sacrificed at 2 weeks and eight at 4 weeks after implantation of glioma cells. The tumors were trimmed to the level of the cortical surface) 2 The rat brains were removed and fixed in a 2% paraformaldehyde solution prior to sectioning and histochemical staining for estimation of tumor volume. The mean volume of the "unresected tumors" was 26.7 +_ 8.3 c u m m for the rats sacrificed at 2 weeks and 119.9 _+ 31.7 c u m m for those killed at 4 weeks (t -- 2.6, p < 0.02) (Fig. 5). Figure 6A shows a typical unreJ. Neurosurg. / Volume 76/April, 1992
Delineation of tumor margins via laser-induced fluorescence
FIG. 6. Low-power photomicrographs of rat glioma specimens. The C1AIPcS4glioma cells, stained bright blue by the galactosidase histochemical technique, appear dark in these photographs (solid arrow.~j. The tumor bed is marked by open arrows. Bar = 0.1 ram. A: A typical unresected tumor (arrow) at 2 weeks after ~mplantation of glioma cells. B-D: The nodules of residual tumor were missed by the laser-induced fluorescence probe at random in four animals (B). Residual tumor was missed in the cortical margin of the tumor bed in two rats (C) because laser-induced fluorescence measurements were carried out systematically only within the tumor bed. In one animal, infiltrative tumor at the depth of the tumor bed was not detected by la~er-induced fluorescence measurement (D).
sected tumor at 2 weeks after implantation of the glioma cells. Visual vs. Laser-Induced Fluorescence-Guided Resection In order to test the completeness of visual versus laser-induced fluorescence-guided resection, tumors were implanted in 19 rats. The C1AIPcS4 dye (5 mg/kg) was administered intravenously 24 hours prior to the experiment. Tumor resection was carded out by one of us (W.S.P.) under the operating microscope, with gentle suction using a l-mm glass pipette and bipolar cautery. The tumor appeared gray, whereas normal brain was white. Complete resection via visual guidance was confirmed by a second observer (K.T.S.). To ensure that there was no bias in the randomization procedure, we measured the mean laser-induced fluorescence signals of the four quadrants and the center of the tumor bed after visual resection but prior to random assignment J. Neurosurg. / Volume 76/April, 1992
into test groups. There was no statistical difference between any treatment group prior to laser-induced fluorescence-guided resection (Table 2). The uniformity of the mean laser-induced fluorescence signal for the four groups indicates quantitatively that the visually guided resections were performed in a consistent manner. It does not address the question of whether another neurosurgeon would have obtained fewer or more laserinduced fluorescence counts (indications of residual tumor at the margin), but our goal was only to demonstrate that laser-induced fluorescence could reduce the number of tumor cells remaining at the margin. One-half of the rats were randomly selected for further resection of residual tumor under laser-induced fluorescence guidance. Resection continued until the laser-induced fluorescence signal was reduced to approximately twice the background level of 1000 to 2000 counts; this quantitative criterion served to give the neurosurgeon an objective means of deciding when to 6@3
W. S. Poon, et aL TABLE 2 Laser-inducedfluorescence signal of the fimr quadrants and the center of the tumor bed*
Tr~tment
Laser-InducedFluorescence Signal (kcounts)
Sacrificed Sacrificed Immediately After 2 Wks visual resection 9.47 _+3.45t 9.29 _+3.67r (10 rats) (I0 rats) laser-induced fluorescence 9.21 ___1.63t 8.24 _+3.32~t guided resection (9 rats) (9 rats) * Data obtainedfollowingvisualresectionbut prior to randomizing into varioustreatment grou~. Assumingthat a laser-induced fluorescence signal abovebackgroundin tumor beds represents the presence of residual tumor, comparablelaser-induced fluorescencesignals for all four groups suggestcomparableextent of tumor resection in each of the four groupsprior to randomization. Valuesexpressedas means -+ standard error of the means. Significanceof difference: p < 0.5; tt = 0.1, r = 0.3.
terminate resection and avoided surgeon bias. This was a very conservative resection criterion; the spectra in Fig. 2 show that the resection could have been continued until the laser-induced fluorescence signal was at a level about 20% above background before noise on the signal became a problem. All 19 animals were sacrificed within 30 minutes after the surgical procedure, and the brains were removed for histological study and estimation of residual tumor volume. The nine rats that underwent laserinduced fluorescence-guided resection had a mean residual tumor volume of 0.5 + 0.2 cu mm (range 0 to 2.0 cu mm), which was significantly less than that of the 10 rats that underwent visual resection (13.7 _+_4.09 cu mm, range 0.1 to 36.5 cu mm; t = 2.5, p < 0.05) (Fig. 5). The detection limit of residual tumor was estimated as 1.5 • 10-f eu mm by calculating the volume of the region using the area of a cell (diameter 20 um) and the thickness of the microscope sections (40 urn). The pattern of failure of laser-induced fluorescence-guided resection in leaving residual tumor in the tumor bed is demonstrated by Fig. 6.
Tumor Regrowth To test whether laser-induced fluorescence-guided resection altered the rate of tumor regrowth in the rat giioma model, the procedure described above was repeated in 24 animals, making allowance for a 20% procedure-related mortality rate. Following tumor resection, these animals were permitted to survive for 2 weeks in order to compare the recurrent tumor volume between the group with laser-induced fluorescenceguided resection (Group 1) and the group with visual resection (Group 2). Three animals in Group 1 and two in Group 2 died immediately postoperatively, leaving nine and 10 animals, respectively, for analysis. The mean recurrent tumor volumes were 36.1 + 18.6 cu mm (range 0 to 175.5 cu mm) in Group 1 rats and 66.5 _+ 24.0 cu m m (range 9.0 to 175.8 cu mm) in Group 2 684
rats (t = 1.3, 0.1 < p < 0.5). The growth rate of the tumors, assuming exponential growth, was proportional to the natural logarithm of the ratio of tumor volume measured 2 weeks after resection of tumor volume calculated immediately after resection. The growth rate of the tumors in Group 1 rats was about 2.7 times that of the tumors in Group 2 animals. This may reflect one of several factors. The growth rate may be dependent in part upon the amount of space available; immediately after resection, the average volume of the residual tumors in Group 1 animals was about 27 times smaller than the average volume of those in Group 2 rats and therefore the larger tumor was closer to the size limits imposed by the volume available for growth. Additionally, as gliomas grow, the central areas of larger tumors may become hypoxic or necrotic and show reduced growth. Two of the nine animals in Group 1 died at 10 days and three of the 10 in Group 2 died at 10 to 12 days (X2 = 0.15, p > 0.5). Three Group 1 rats were tumorfree at 2 weeks, whereas every animal in Group 2 had gross tumor recurrence (X2 = 4.0, p < 0.05). In all animals, the mean laser-induced fluorescence signals of the four quadrants and the center of the tumor bed after visual resection and before randomization into groups had almost identical values, suggesting a comparable extent of residual tumor (Table 2).
Discussion This is the first report of the use of laser-induced fluorescence to increase the accuracy with which rat brain-tumor margins are defined during resection, and, to our knowledge, the first use of laser-induced fluorescence in vivo to delineate tumor margins during resection in any tumor model.
Reduction of Tumor Volume Substantial reduction of brain-tumor volume prolongs life, improves the quality of survival, and may make adjuvant therapy more effective, z~3`~s'~s'19'3~'3:We have demonstrated that laser-induced fluorescence detection of phthalocyanine in a rat glioma model is effective in maximizing the extent of tumor resection; however, in those animals that were tumor-free at 2 weeks, we have not excluded the possibility that tumor recurrence from small numbers of infiltrating tumor cells might occur at a later time. Indeed, in humans, infiltrating glioma cells are the norm. Thus, this technique is not aimed at tumor cure but rather at extensive defined reduction of the bulk of the tumor. Nonetheless, the ability to resect the bulk of a glioma may allow for localized forms of therapy, 2j'2g including implants of chemotherapeutic agents ~~ or other locally applied biological agents. '4'~ Potential Advantages of the Technique There are several differences between this work and prior studies of brain-tumor fluorescence. The autofluorescence spectrum of normal brain, excited by 337 J. Neurosurg / Volume 76/April, 1992
Delineation of tumor margins via laser-induced fluorescence nm radiation, is concentrated between 350 and 650 nm. This background fluorescence severely hampers tumor localization using either visual or broad-band photographic detection. Since our detection system uses a gating polychromator to reject the broad fluorescence emission of normal brain, only the weak background fluorescence of dye in normal brain at our 680-rim measurement wavelength must be corrected for. The spectral selectivity of our detection system manifests itself in the high 40:1 ratio of tumor:normal brain fluorescence. Additionally, some previously used fluorescent dyes, such as fluorescein and hematoporphyrin derivative, have fluorescence emission peaks overlapping the autofluorescence of normal brain and are not optimum for delineating tumor margins. While we used the photosensitizer CIA1PcS4, similar results would be expected from other sensitizers emitting at wavelengths greater than 650 nm with similar quantum efficiencies. Selectivity is primarily the result of the blood-brain barrier, which prevents the sensitizer from entering the brain except in tumor regions, where there is a proliferation of immature leaky neovessels. Clinical Applications This system should be easily applied to surgery. The present unit is mounted on a computer trolley and is therefore portable. The fiberoptic probe and the short (5-sccond) data acquisition time allow multiple measurements to be made in a resection cavity. Since the information needed to guide resection involves only the fluorescence intensity at a single wavelength, the present polychromator-multichannel analyzer could be replaced by a simpler, less expensive, and more compact system. Only the 600-~m quartz fiber needs to be sterilized. Potential applications include identification of tumor resection margins, localization of subeortical tumor during open operation, and confirmation of target in stereotactic tumor biopsies. Additionally, the technique may enhance the effectiveness of photodynamic therapy by reducing residual tumor volume, thus aiding light penetration into a thick tumor, 19and by targeting localized areas suggestive of infiltrative tumor rather than treating the entire tumor bed. Finally, since the laser-induced fluorescence signals are well correlated with tissue concentrations of CIA1PcS4 below 10 gg/gm tissue, this technique may be used to correlate dinical results with drug concentration and light dose. This real-time information may allow the development of a drug/light-dose protocol that is tumoricidal but causes minimal brain damage; it may also permit a decision to forego treatment when the drug concentration in the tumor is not optimal. Limitations and Areas for Further Development While these techniques may be readily applied in the clinical setting, caution must be exercised in extrapolating these animal studies to neurosurgical practice. Selective uptake of porphyrin compounds is not tumorJ. Neurosurg. / Volume 76/April, 1992
specific. Hematoporphyrin fluorescence was also demonslrated in traumatized brain and normal pituitary stalk s~ and choroid plexus tissue. 26 However, in one study using tritiated hematoporphyrin derivative in a dog glioma model 26 the highest concentrations of the fluorescent drug were found in the cytoplasm of the tumor cells, while the concentration in peritumoral edematous brain or brain adjacent to tumor was at the same low level as in normal brain. This concentration was slightly raised in a rat 9L gliosarcoma studied using digital video fluorescence microscopyJ These results may reflect the different natures of the tumor involved. In general, the uptake of photosensitizer in brain adjacent to tumor can be different from that in normal brain; changes in the vasculature of brain adjacent to tumor can affect its uptake. Uptake to phthalocyanines in brain adjacent to tumor has not been studied. However, it should again be pointed out that CIAIPcS4 has the following advantages over hematoporphyrin derivative: 1'23 the skin phototoxicity is negligible and the longer wavelength of the fluorescence band at 680 nm reduces interference from tissue autofluorescence. Our criterion for discontinuing resection (reducing the laser-induced fluorescence signal to twice the background) was very conservative and may not be optimal; resection until the laser-induced fluorescence signal is barely detectable above the background may be better. Current development of photosensitizing drugs for photodynamic therapy emphasizes chromophores absorbing at 650 to 800 nm, a wavelength range that allows deeper light penetration. Such photosensitizers would be even less sensitive to interference from autofluorescence than CIAIPcS4. The pattern of failure of laser-induced fluorescence detection of residual tumor at the resection margin (Fig. 6) suggests that more point measurements at and within the vicinity of the tumor bed and/or a real-time fluorescent imaging technique s may improve its performance. In preliminary experiments, we have been able to obtain images of the tumor bed using a standard commercial video camera in conjunction with an electronic system that allows the integration of the video signal for periods of up to 30 seconds. Conclusions
This study has demonstrated that intraoperative laser-induced fluorescence measurement is an effective means of delineating tumor margin during resection in glioma-beafing rats injected with CIA1PcS4, resulting in significant reduction of tumor volume compared with standard microsurgical resection. These studies need to be confirmed and extended in other animal models, and additional studies should be performed before this technique can he applied to humans. References
1. AbernatheyCD, Anderson RE, Kooistra KL, et al: Activity of phthalocyanine photosensitizers against human glioblastoma in vitro. Neurosurgery21:468-473, 1987 685
W. S. Poon, et al. 2. Ammirati M, Vick N, Liao Y, et al: Effect of the extent of surgical resection on survival and quality of life in patients with supratentorial glioblastoma and anaplastic astrocytomas. Neurosurgery 21:201-206, 1987 3. Ankerst J, Montan S, Svanberg K, et al: Laser-induced fluorescence studies of hematoporphyrin derivative (HPD) in normal and malignant tissue of rat. Appl Spectr 38:890-896, 1984 4. Balchum O J, Profio AE, Razum N: Ratioing fluorometer probe for localizing carcinoma in situ in bronchi of the lung. Photaehem Photohiol 46:887-891, 1987 5. Baumgartner R, Fisslinger H, Jocham D, et al: A fluorescence imaging device for endoscopic detection of early stage cancer - - instrumental and experimental studies. Photoehem Photobial 46:759-763, 1987 6. Benda P, Someda K, Messer J. et al: Morphological and immunochemical studies of rat glial tumors and clonal strains propagated in culture. J Neurosurg 34:310-323, 1971 7. Berg K, Bummer JC, Moan J: Evaluation of sulfonated aluminum phthalocyanines for use in photochemotherapy. Cellular uptake studies. Cancer Lett 44:7-15, 1989 8. Boggan JE, Walter R, Edwards MSB, el al: Distribution of hematoporphyrin derivative in the rat 9L gliosareoma brain tumor analyzed by digital video fluorescence microscopy. J Neurosurg 61:1113-1119, 1984 9. Bown SG, Tralau CJ, Coleridge Smith PD, et al: Photodynamic therapy with porphyrin and phthalocyanine sensitisation: quantitative studies in normal rat liver. Br J Cancer 54:43-52, 1986 10. Brem H, Mahaley MS Jr, Vick NA, el al: Interstitial chemotherapy with drug polymer implants for the treatment of recurrent gliomas. J Neurosurg 74:441-446, 1991 1I. Chandler WF, Knake JE, McGillieuddy JE, et al: Inlraoperative use of real-time ultrasonography in neurosurgery. J Neurosurg 57:157-163, 1982 12. Cheng MK, McKean J, Boisvert D, et al: Effects of pho~.oradiation therapy of normal rat brain. Neurosurgery 15:804-810, 1984 13. Cooper JS, Borok TL, Ransohoff J, et al: Malignant gliurea. Results of combined modality therapy. JAMA 248: 62-65, 1982 14. Ezzedine ZD, Martuza RL, Short MP, et al: Selective killing of giioma cells in culture and in vivo following retrovirus transfer of the herpes simplex virus thymidine kinase gene. New Biolog 3:608-614, 1991 15. Harsh GR IV, Wilson CB: Neuroepithelial tumors of the adult brain, in Youmans JR (ed): Neurological Surgery, ed 3. Philadelphia: WB Saunders, 1990, pp 3040-3136 16. Kelly PJ, Kall BA, Goerss SJ: Special stereotactic techniques: stereotactic laser resection of deep-seated tumors, in Heilbrun MP (ed): Concepts in Neurosurgery, Vol 2: Stereotactie Neurosurgery. Baltimore: Williams & Wilkins, 1988, pp 233-240 17. Martuza RL, Malick A, Markert JM, el al: Experimental therapy of human glioma by means of a genetically engineered virus mutant. Science 252:854-856, 1991 18. MRC (Medical Research Council) Brain Tumor Working Party: Prognostic factors for high grade malignant glioma: development of a prognostic index. J Neurooncol 9: 47-55, 1990 19. Muller PJ, Wilson BC: Photodynamic therapy of malig-
686
20. 21. 22.
23. 24.
25. 26.
27. 28.
29.
30. 31. 32. 33.
nant primary brain tumours: clinical effects, post-operative ICP, and light penetration of the brain. Photochem Photobiol 46:929-935, 1987 Profio AE: Laser excited fluorescence of hematoporphyfin derivative for diagnosis of cancer. IEEE J Quantum Electron QE-20 12:1502-1507, 1984 Profio AE, Doiron DR, King EG: Laser fluorescence bronchoscopy for localization of occult lung tumors. Meal Phys 6:523-525, 1979 Rassmussen-Taxdal DS, Ward GE, Figge FHJ: Fluorescence of human lymphatic and cancer tissues following high doses of intravenous hematoporphyrin. Cancer 8: 78-81, 1955 Sandeman DR, Bradford R, Buxton P, et al: Selective necrosis of malignant gliomas in mice using photodynamic therapy. Br J Cancer 55:647-649, 1987 Shimohama S, Rosenberg MB, Fagan AM, et at: Grafting genetically modified ceils into the rat brain: characteristics of E. coil beta-galactosidaseas a reporter gene. Brain Res Mol Brain Res 5:271-278, 1989 Short MP, Choi BB, Lee JK, et al: Gene delivery to glioma cells in rat brain by grafting of a retrovirus packaging cell line. d Neurosci Res 27:427-433, 1990 Steichen JD, Dashner K, Martuza RL: Distribution of hematoporphyrin derivative in canine glioma following intraneoplastic and intrapefitoneal injection. J Neurosurg 65:364-369, 1986 Svanberg K, Kjellrn E, Ankerst J, et al: Fluorescence studies of bematoporphyrin derivative in normal and malignant rat tissue. Cancer Res 46:3803-3808, 1986 Tralau CJ, Ban" H, Sandeman DR, et at: Aluminum sulfonated phthalocyanine distribution in rodent tumors of the colon, brain and pancreas. Photochem Photohiol 46:777-781, 1987 Tralau CJ, MacRobert AJ, Coleridge-Smith PD, et al: Photodynamic therapy with phthalocyanine sensitisation: quantitative studies in a transplantable rat fibrosarcoma. Br J Cancer 55:389-395, 1987 Turner DL, Cepko CL: A common progenitor for neurons and glia persists in rat retina late in develoment. Nature 328:131-136, 1987 Walker MD: Malignant glioma, in Wilson CB, Huff JT {eds): Current Surgical Management of Neurologic Disease. New York: Churchill Livingstone, 1980, pp 72-78 Walker RW, Posner JB: Central nervous system neoplasm. Curr Neurol 5:285-322, 1984 Wise BL, Taxdal DR: Studies of the blood-brain barrier utilizing hematoporphyrin. Brain Res 4:387-389, 1967
Manuscript received January 30, 1991. Accepted in final form August 16, 1991. This research was supported in part by the SDIO-MFEL program under Office of Naval Research Contract N0001486-K-0117 and by a grant from Neurofibromatosis, Inc., Massachusetts. Dr. Poort was the recipient of a grant from the Faculty of Medicine, Chinese University of Hung Kong. Address reprint requests to: Robert L. Martaza, M.D., Department of Neurosurgery, Georgetown University Medical Center, 3800 Reservoir Road NW, Washington, D.C. 200(37.
J. Neurosurg. / Votume 76/April, 1992
