Contributed by Samuel J. Danishefsky, June 6, 2008 (received for review February 4, 2007)
The epothilones represent a promising class of natural product-based antitumor drug candidates. Although these compounds operate through a microtubule stabilization mechanism similar to that of taxol, the epothilones offer a major potential therapeutic advantage in that they retain their activity against multidrug-resistant cell lines. We have been systematically synthesizing and evaluating synthetic epothilone congeners that are not accessible through modification of the natural product itself. We report herein the results of biological investigations directed at two epothilone congeners: iso-fludelone and iso-dehydelone. Iso-fludelone, in particular, exhibits a number of properties that render it an excellent candidate for preclinical development, including biological stability, excellent solubility in water, and remarkable potency relative to other epothilones. In nude mouse xenograft settings, iso-fludelone was able to achieve therapeutic cures against a number of human cancer cell lines, including mammarian-MX-1, ovarian-SK-OV-3, and the fast-growing, refractory, subcutaneous neuroblastoma-SK-NAS. Strong therapeutic effect was observed against drug-resistant lung-A549/taxol and mammary-MCF-7/Adr xenografts. In addition, iso-fludelone was shown to exhibit a significant therapeutic effect against an intracranially implanted SK-NAS tumor.
In the search for cancer therapeutic agents, the natural product estate has proven itself to be a leading source of anticancer compounds. In fact, >60% of all clinically approved anticancer drugs are derived from structures found in nature. Although some of these agents are used in their unaltered natural forms, still more are natural product derivatives that have been modified in ways that lead to improved pharmacological profiles and therapeutic efficacy. We have long been engaged in the synthesis and evaluation of biologically promising natural products and their synthetic analogs, which we access through a process that we term “diverted total synthesis” (1).
Recently, efforts by natural product isolation chemists have led to the identification of a number of particularly promising lead anticancer candidates, notably the epothilones. Epothilones A and B, isolated from the Sorangium cellulosum myxobacterium (2), were reported to display potent in vitro cytotoxicity (3). Like taxol, the epothilones promote the stabilization of microtubule polymerization [supporting information (SI) Fig. S1], interrupting the process of cell division and promoting apoptosis. However, unlike taxol and most clinically available anticancer agents, the epothilones do not appear to suffer from a loss of effectiveness against multidrug-resistant (MDR) tumor cells.
Upon learning of the unique properties of the epothilones, we launched a program directed toward the total synthesis of epothilone A (EpoA) and epothilone B (EpoB). In 1996, we reported the total synthesis of EpoA (4), and shortly thereafter, EpoB (5). In preliminary in vivo studies with synthetic material, EpoB was found to be highly toxic in mice, even at subtherapeutic dosages. We suspected that this nonspecific toxicity might be, at least to some extent, attributable to the epoxide linkage at C12–C13 of the natural product, and we elected to “edit out” this structural feature (Fig. 1). Thus, 12,13-desoxy-epothilone B (dEpoB) was prepared and shown to be very well tolerated in a number of in vivo settings (6–8). Furthermore, although dEpoB is notably less potent than the parent compound (EpoB), it does retain its activity against MDR cell lines. On the basis of its performance in preclinical studies, dEpoB was advanced to clinical trials and was recently evaluated in late Phase II settings against breast cancer (9).
Chemical structures and rational approach to molecular design and synthesis.
In our second generation analogs, we sought to restore some of the potency that had been forfeited when progressing from the natural product to dEpoB. We postulated that stability and potency could be enhanced through the installation of structural features that would confer rigidity to the molecule. We found that by installing a second double bond at the C9–C10 position, we could synthesize a promising analogue that demonstrated a marked increase in both intrinsic potency and biological stability. This compound, 9,10-dehydro-dEpoB, is termed dehydelone (KOS-1584) and shows significantly enhanced potency in in vivo mouse settings. In addition, dehydelone demonstrates increased serum stability when compared with dEpoB. Dehydelone is currently in Phase II clinical trials against breast cancer.
Perhaps as a consequence of its enhanced potency, dehydelone is more toxic than dEpoB, and, as a result, lower dosages are tolerated in vivo. In the cases of some particularly refractory tumors, dehydelone is unable to achieve ideal levels of tumor eradication, presumably because of its somewhat narrow therapeutic index. In designing a third-generation epothilone-based drug candidate, we sought to attenuate toxicity, thus broadening the therapeutic index. Remarkably, through the substitution of a trifluoromethyl group at C12 in place of the methyl group of the parent compound, we were able to achieve a significant improvement in terms of therapeutic index (10–12). This analogue, termed fludelone, is markedly less toxic than dehydelone. Although fludelone is also less potent than dehydelone, its therapeutic index is far superior, thereby eradicating a number of tumors that are particularly difficult to treat. Fludelone, which is in late preclinical investigations, has emerged as a very promising lead therapeutic candidate. The results of preclinical in vivo evaluations of fludelone may be found elsewhere (9, 10). Herein, we focus on a promising candidate in the fludelone family: iso-fludelone.
In seeking to optimize still further in the fludelone series, we hoped to restore some of the potency that had been lost in the progression from dehydelone to fludelone. With this view, and the hope of increasing the duration of action of the agent, we synthesized the 17-iso-oxazole analogue of fludelone, iso-fludelone (11). For comparison, we also prepared the 17-iso-oxazole congener of dehydelone (iso-dehydelone). In fact, our hopes of enhanced potency were realized. As will be shown, iso-fludelone is a highly promising epothilone analogue. As a consequence of its remarkable potency and stability, iso-fludelone is able to achieve complete remission (CR) and therapeutic cures in certain mouse xenograft models with as few as four doses at relatively infrequent dosage intervals of up to 12 days. We believe iso-fludelone to be a most remarkable drug for reasons described herein.
A comparative evaluation of the in vitro activities of various epothilone derivatives against a range of tumor cell lines reveals some important trends (Table S1). Each compound in the iso-oxazole series (iso-dehydelone and iso-fludelone) is more potent than its parent compound (dehydelone and fludelone). Fortunately, much of the in vitro potency that had been lost in progressing from dehydelone to fludelone was restored through the installation of the isoxazole moiety. Also, iso-fludelone generally displays markedly enhanced potency when compared with the established cancer chemotherapeutic agents vinblastine and taxol. Most importantly, the epothilone analogues retain their activity against the MDR cell lines CCRF-CEM/VBL and CCRF-CEM/taxol (Table S1).
The physico-chemical and metabolic properties of the various epothilone analogues are summarized in Table S2. As shown, the microtubule stabilization activity of each derivative is within the range of taxol. The metabolic stability of the epothilones in mouse plasma and human liver microsomal S-9 fraction is greatly enhanced through the installation of the second C9–C10 double bond in the epothilone framework. Remarkably, the synthesized isoxazole-containing compounds iso-fludelone and iso-dehydelone demonstrate greatly enhanced levels of stability over all previously evaluated epothilone derivatives. Furthermore, the enhanced water solubility of the fludelone family offers advantages related to formulation. These newly developed compounds also retain promising levels of lipophilicity that promote in vivo tissue permeation. The methods of analyzing physico-chemical properties and in vitro metabolic studies are described in SI Methods.
In our extensive investigation of the in vivo properties of the epothilones, conducted over the course of many years (6–12), we have formulated some general conclusions regarding the optimal modes of administration and dosing schedules of these compounds, noting that the epothilones demonstrate a very strong dependency on the mode of their administration. We have found that the use of slow i.v. infusion brings greatly reduced toxicity (Table S3), which results in a significantly broadened therapeutic index. When epothilones are administered over 6-hour i.v. infusion, the tolerated dose is much higher than that observed for other traditional modes of administration, such as bolus injection. In addition, if any significant sign of toxicity occurs (e.g., marked body-weight decreases), intermittent cessation of treatment is required to allow toxicity to subside. This approach is particularly effective in avoiding delayed neurotoxicities, such as peripheral neuropathy. These findings, observed primarily in mouse models, may well translate to human clinical settings. Unless otherwise specified, all of our in vivo studies reported herein for epothilones use 6-h i.v. infusion. For each xenograft study, tumor-size and body-weight changes were monitored simultaneously. The body-weight changes are provided in corresponding figures in the SI Methods (Fig. S2).
Efficacy and toxicity may also be strongly affected by dose, interval, and schedule of treatment, thus requiring preliminary studies before formal experimentation. In the in vivo studies described herein, we report the results achieved at approximate maximum tolerated doses, those doses that do not result in death to the animal or delayed toxicity. The sole exception to this rule is the survival lifespan studies of SK-NAS brain-tumor-bearing mice (see below).
In many of our in vivo evaluations, we have continued to monitor mice for several months after the cessation of treatment. Our objective has been to determine the extent to which the tumor cells have been eradicated. Because of the unusually long-term nature of these studies, we differentiate between CR and therapeutic cure as follows: CR is defined as disappearance of tumor, whereas a de facto cure is defined as CR without relapse for at least 15% of the lifespan (4 months in mouse models). We further attempt to estimate a log-cell kill (LCK) in tumor mass (11), based on the length of time of CR and the time required for the tumor to double in size (see SI Methods). Typically, a tumor cell kill that reaches below one-billionth to one-trillionth [i.e., LCK > (9–12)] is deemed to be a cure.
Nude mice bearing an extra-large MX-1 xenograft (up to 7.8% body weight) were treated with iso-fludelone (30 mg/kg, Q12Dx4) by 6-hour i.v. infusion, on day 25 after tumor implantation. Complete tumor remission was observed in four of four mice after administration of the third dose, and a fourth dose was given for consolidation. There was no relapse in any of the mice on day 158 (Fig. 2A). It was encouraging to find that the maximal body-weight decreases were 12%, 10%, and 6% after sequential administration of three doses, indicating not only a lack of cumulative toxicity, but also an alleviation of toxicity over time (Graph 2A of Fig. S2). Photographs of the mice, taken on days 25, 39, and 53 are shown in Fig. 2B. The LCK of iso-fludelone against MX-1 is estimated to be ≈10.94. Thus, based on our definition, a therapeutic cure was achieved against the MX-1 xenograft.
Treatment of mammary and ovarian tumors. (A) Therapeutic effect of iso-fludelone against extra-large mammary MX-1 tumor xenografts. Initial tumor size, 2.26 ± 0.17g, is ≈7.8% of mice body weight. (B) Typical photos of mice on day 25, day 39 and day 53 for the experiment shown in A. (C) Oral therapy of MX-1 xenograft tumor with dehydelone, iso-dehydelone, iso-fludelone, taxotere, and capecitabine. Dose, route, and schedule are as shown. (D) Therapy of ovarian SK-OV-3 tumor xenograft tumor. Dose, route, and schedule for dehydelone, iso-fludelone, and taxol are as shown.
As summarized in Table S4 and Fig. S3, our relevant epothilone analogues—dehydelone, fludelone, iso-dehydelone, iso-fludelone, as well as taxol—have been shown to achieve CR against normal-sized MX-1 xenografts under our standard experimental protocols, in which treatment was started when the MX-1 tumors had grown to 100–200 mm3 in size. Fludelone, iso-dehydelone, and iso-fludelone showed long-term no-relapse cure.
Nude mice bearing MX-1 xenografts were treated orally beginning on day 11 after tumor implantation with dehydelone, iso-dehydelone, iso-fludelone, capecitabine, and taxotere at the doses and schedules specified in Fig. 2C. In this study, taxotere, which is known to have poor oral bioavailability, served as a negative control. Of these compounds, iso-fludelone (45 mg/kg, Q8Dx7) achieved CR in four of four mice on D35 without relapse up to day 168. The LCK of iso-fludelone is ≈13.3, indicating the achievement of an oral therapeutic cure of the MX-1 xenograft. At the dose and schedule examined (45 mg/kg, Q9Dx6), dehydelone achieved CR on day 32. Although one of four mice relapsed on D71, the remaining three achieved a cure, with no relapse on day 122 (LCK ≥ 9.03). The manageable body-weight loss observed in this course of treatment suggests that optimal conditions can be identified to produce an even better outcome for dehydelone. Treatment with iso-dehydelone (25 mg/kg, Q10Dx3) led to 90.4% suppression on day 20 without further suppression with subsequent doses. In contrast to the remarkable oral efficacy of the epothilone compounds, the established oral anticancer agent, capecitabine (800 mg/kg, QDx7), achieved only 77.6% tumor suppression on day 20 and produced body-weight losses of up to 21%. As predicted, taxotere (30 mg/kg, Q3Dx4) was only able to achieve 26.6% tumor suppression on day 20, without detectable body-weight decreases, presumably because of its poor oral bioavailability.
The clinical efficacy of taxol in treating ovarian cancer has been well documented. Our third-generation epothilone analogues are also able to effectively achieve therapeutic cures against the SK-OV-3 xenograft. As shown in Fig. 2D, iso-fludelone (15 mg/kg, Q6Dx5), dehydelone (20 mg/kg, Q6Dx7), and iso-dehydelone (10 mg/kg, Q14Dx4) were able to achieve therapeutic cures in fewer doses, without relapse up to day 202 and with LCK ≥ 11.7.
We next evaluated iso-fludelone against the refractory lung carcinoma A549 cell line and the taxol-resistant subline, A549/taxol (574-fold resistant in vitro) (see SI Methods for tumor cell line generation). Each mouse was simultaneously inoculated s.c. with both the A549 and A549/taxol xenograft tumors on the left and right sides of the flank, to minimize interanimal variations for the two tumors and to take advantage of the fact that the two xenografts grow at similar rates. Treatment commenced on day 11 when each tumor had grown to ≈100 mm3 in size.
As shown in Fig. 3A, both iso-fludelone (15 mg/kg, Q7Dx4) and taxol (25 mg/kg, Q3Dx9) were able to achieve >99.7% tumor suppression against A549, but without tumor shrinkage. Iso-dehydelone (10 mg/kg, Q12Dx3) and dehydelone (20 mg/kg, Q7Dx4) achieved only 89.5% and 77% tumor suppression, respectively.
Treatment of refractory tumors. (A) Therapy of nude mice bearing both refractory human lung carcinoma A549 (right side flank) and taxol-resistant, A549/taxol (left side flank) with dehydelone, iso-dehydelone, iso-fludelone, and taxol. Dose, route, and schedule are as shown. (B) Therapy of adriamycin-resistant human mammary adenocarcinoma MCF-7/Adr with fludelone, iso-fludelone, and taxol. Dose, route, and schedule are as shown. (C) Therapy against s.c. implanted human neuroblastoma SK-NAS xenograft with dEpoB, dehydelone, fludelone, iso-dehydelone, iso-fludelone, and taxol. Dose, route, and schedule are as shown. (D) Therapy against refractory, intracranially implanted human neuroblastoma SK-NAS xenograft in nude mice with iso-fludelone, BCNU, taxol, and temodar. Dose, route, and schedule are as shown. The time courses of percent survival in treated and untreated control are shown in a Kaplan–Meier plot.
The therapeutic effects of the epothilones and taxol against the A549/taxol100 xenograft are presented in Fig. 3A. As expected, the efficacy of taxol was greatly diminished against the taxol-resistant xenograft, accomplishing suppression of tumor growth by 42%. Under the conditions examined, iso-fludelone demonstrated the best efficacy, suppressing tumor growth by 80%. Dehydelone and iso-dehydelone achieved growth suppression of 72% and 57%, respectively.
Nude mice bearing the adriamycin-resistant MCF-7/Adr xenograft were treated with fludelone (25 mg/kg, Q2Dx4), iso-fludelone (15 mg/kg, Q7Dx2), or taxol (25 mg/kg, Q2Dx4) by 6-h i.v. infusion (Fig. 3B). Fludelone and iso-fludelone were able to effectively suppress tumor growth by 98.2% and 100%, respectively (Day 18), whereas taxol was only able to achieve ≈3.7% tumor suppression.
The s.c. SK-NAS xenograft doubles in tumor size in<1.7 days, rendering this refractory xenograft the most aggressive used in our studies. When treatment commenced on day 11, the tumor size was ≈200 mm3 (Fig. 3C). By day 17, the control-group tumor size had grown to 4,000 mm3 and the control animals were killed. On day 17, taxol (30 mg/kg, Q2Dx4) and dEpoB (30 mg/kg, Q2Dx5) exhibited tumor suppression of 74.4% and 81.4%, respectively.
Treatment with dehydelone (20 mg/kg, Q6Dx4) induced CR in one of three mice on day 37, with relapse on day 43 (LCK = 1.06). Fludelone (20 mg/kg, Q2Dx9) achieved CR in one of three mice on day 33, which relapsed on day 63 (LCK = 5.31), and iso-dehydelone (10 mg/kg, Q12Dx2) also yielded CR in one of three mice (Day 19), which relapsed on day 43 (LCK = 4.25). By contrast, only iso-fludelone (45 mg/kg, Q12Dx4) was able to induce CR in four of four mice, with no relapse on day 120 (LCK ≥ 14.60). Thus, among the epothilone analogues evaluated herein, iso-fludelone uniquely achieved a therapeutic cure against the refractory, fast-growing SK-NAS xenograft. Further detailed comparisons of epothilones against SK-NAS xenografts are given in Table S5.
To evaluate the therapeutic effect of iso-fludelone in a difficult-to-treat brain-tumor model, the fast-growing refractory neuroblastoma SK-NAS was implanted intracranially (106 cells per μl) into the right frontal lobe of nude mice brains. As shown in the Kaplan–Meier plot for the time course of survival (Fig. 3D), the median survival time (MST) of the control group of mice was 16.8 ± 0.4 days. Treatment with the brain tumor chemotherapeutic agents BCNU (15 mg/kg, QDx4, i.v.) or Temodar (200 mg/kg, Q2Dx3, oral) yielded increases in MST to 17.8 ± 0.4 days and 19.2 ± 1.3 days, respectively. As expected, taxol (20 mg/kg, Q2Dx7, i.v.) had little therapeutic effect, with an MST of 18.8 ± 1.2 days. By contrast, treatment with iso-fludelone (30 mg/kg, Q10Dx2, 6-h i.v. infusion) yielded a significantly increased MST of 26.8 ± 1.3 days**. These results, against a highly refractory tumor, reveal a marked improvement for iso-fludelone over the established cancer agents evaluated in this study. The findings suggest that iso-fludelone is readily capable of exerting a strong therapeutic effect against brain tumors.
Fig. S4A shows the distribution of various epothilones in the MX-1 tumor, after 3-h i.v. infusion of tumor-bearing nude mice (13). At the low dose of 2.5 mg/kg, iso-fludelone has the highest concentration at 6 h, followed by fludelone and dehydelone. dEpoB, at 10 mg/kg, reached similar levels as dehydelone at 6 h. At 21 h after infusion, the tumor epothilone concentrations were as follows: iso-fludelone > fludelone > dehydelone > dEpoB. It is remarkable that, even at this low dose, the tumor drug concentrations for iso-fludelone and fludelone are several hundred times higher than their in vitro IC50 values: up to 21 h after the end of infusion.
The tumor/brain ratios, after 5-min, 1-h, and 3-h infusion, are shown in Fig. S4 B and C for fludelone and iso-fludelone, respectively. The ratio was as high as 14 for fludelone at 1 h and 22 for iso-fludelone at 3 h, after infusion. These high ratios are rarely seen in antitumor agents. The tumor/brain drug ratios are compared in Fig. S4D, after 3-h i.v.-infusion at 2.5 mg/kg each, with the exception of dEpoB, which was evaluated at 10 mg/kg. At 6 h after infusion, the tumor/brain ratio rank was: iso-fludelone ≫ fludelone ≫ dehydelone > dEpoB. At 21 h after infusion, the rank was: fludelone > iso-fludelone ≫ dehydelone > dEpoB.
The results indicate that the epothilones, particularly fludelone and iso-fludelone, are concentrated selectively in the tumor rather than in the brain (or many other organs) (data not shown). In addition, these compounds persist in the tumor for a very long time (e.g., 21 h) and at very high concentrations (e.g., several-hundred-fold concentrations compared with their IC50 in vitro).
The preclinical findings from the nude mouse/xenograft models presented herein provide valuable directives into the ways in which specific perturbations of the epothilone framework may impact issues of biological efficacy and toxicity. These synthetic epothilone analogues represent a lineage of systematically designed compounds, each of which incorporates features that have been installed with the hope of enhancing different aspects of biological performance.
In terms of in vitro cytotoxicity, the epothilone analogues may generally be ranked as follows: iso-dehydelone > iso-fludelone ≈ dehydelone > fludelone. The analogues in the isoxazole series (iso-dehydelone and iso-fludelone) demonstrate enhanced potency when compared with their congeners, dehydelone and fludelone. The metabolic stability of the epothilones may be generalized as follows: iso-dehydelone ≈ iso-fludelone > fludelone > dehydelone.
Because of their enhanced potency and remarkable metabolic stability, iso-dehydelone and iso-fludelone require fewer dosages of 6-h i.v. infusion with longer intervals to achieve strong therapeutic effects in in vivo xenograft settings. It is clear that iso-fludelone is demonstrably superior to iso-dehydelone in terms of in vivo efficacy. Iso-fludelone clearly displays a more favorable therapeutic index, enabling efficacy without provoking unacceptable levels of toxicity.
In comparing fludelone and iso-fludelone, the identification of a superior compound is not as straightforward. What is clear from these studies and others reported previously (10–12) is that both fludelone and iso-fludelone are remarkable compounds that are able to eradicate highly refractory tumors in a number of challenging xenograft settings. Both achieve therapeutic cures against human mammary MX-1 and ovarian SK-OV-3 xenografts. Both show marked growth suppression or CR of refractory A549, A549/taxol, MCF-7/Adr, and SK-NAS xenografts. Remarkably, iso-fludelone has been found to significantly increase the mean survival time of mice implanted with fast-growing SK-NAS brain tumors, even in comparison with established brain-cancer chemotherapeutic agents. As shown in Fig. S4, the tissue distribution for the tumor/brain ratios of iso-fludelone is much higher than other tested epothilones. The pharmacokinetic profile and tissue distribution for iso-fludelone showed high bioavailability, high tissue penetration, and long retention. Finally, both fludelone and iso-fludelone have been shown to achieve therapeutic cures against human mammary MX-1 xenografts through oral administration.
In conclusion, we have presented the results of extensive in vitro and in vivo evaluations focusing on iso-fludelone, which emerges as a highly promising agent, certainly worthy of further vigorous development. It is worth noting that the progression from EpoB to iso-fludelone, which has brought with it major advances in preclinical therapeutic indices, was enabled by an interfacing of the powers of chemical synthesis and the resources of preclinical pharmacology. It is likely that natural products optimized through chemical synthesis offer many additional opportunities for the discovery of promising agents in oncology.
Athymic nude mice (nu/nu) obtained from National Cancer Institute were used in all human tumor xenograft therapeutic studies. Tumors were implanted s.c. except for the SK-NAS neuroblastoma, which was implanted intracranially. The s.c. tumor size was measured periodically at specified times, and the treatment started when tumor size reached 100–200 mm3 or larger. Tumor size and body weight were recorded every 2, 3, or 4 days. Usually, when tumor size reached 10% of body-weight, animals were killed for humane reasons. Prolonged experiments were carried out when treated animals reached complete tumor remission, so that any relapse or cure could be determined. Full material and method descriptions may be found in SI Methods.
We thank Que-Hui Tan and Luan-Ing Chen for expert technical assistance and Grace Kang for editing the manuscript. This work was supported by National Institutes of Health Grant CA28824 (to S.J.D.) and Sloan-Kettering Institute General Fund and Core Grant CA08748 (to T.C.C.).
Author contributions: T.-C.C., R.J., and S.J.D. designed research; T.-C.C., X.Z., Z.-Y.Z., Y.L., L.F., S.E., D.R.M., N.W., and Y.I.Y. performed research; D.R.M. and S.J.D. contributed new reagents/analytic tools; T.-C.C., X.Z., Z.-Y.Z., R.J., N.W., and S.J.D. analyzed data; and T.-C.C., Z.-Y.Z., R.M.W., and S.J.D. wrote the paper.
A preliminary report of this work has been published (13–15).
Conflict of interest statement: Samuel Danishefsky has served as an advisor at Kosan Biosciences, the company that is developing isofludelone. Kosan Biosciences is the licensee of the epothilone/fludelone/isofludelone portfolio of anticancer agents from the Memorial Sloan-Kettering Cancer Center. Because these relationships pertain across the entire range of compounds, there is no particular advantage in the development of any specific member of the portfolio. From that perspective, there is no real conflict of interest because MSKCC and its licensee, Kosan, dominate this field. Incidentally, the MSKCC/Kosan relationship is well known and appreciated in the chemistry/pharmacology/oncology communities and thus is fully known by all of the referees
This article contains supporting information online at www.pnas.org/cgi/content/full/0804773105/DCSupplemental.
↵** In this study, neurological function impairment was not evaluated. We did observe toxicity through initial body weight decreases, followed by recovery, and finally by sharp decreases in body weight that eventually led to death apparently because of aggravation of implanted brain tumor (Graph 3D of Fig. S2).
