Systemic 5-fluorouracil treatment causes a syndrome of delayed myelin destruction in the central nervous system
© Han et al.. 2008
Received: 19 June 2007
Accepted: 19 February 2008
Published: 22 April 2008
Skip to main content
© Han et al.. 2008
Received: 19 June 2007
Accepted: 19 February 2008
Published: 22 April 2008
Cancer treatment with a variety of chemotherapeutic agents often is associated with delayed adverse neurological consequences. Despite their clinical importance, almost nothing is known about the basis for such effects. It is not even known whether the occurrence of delayed adverse effects requires exposure to multiple chemotherapeutic agents, the presence of both chemotherapeutic agents and the body's own response to cancer, prolonged damage to the blood-brain barrier, inflammation or other such changes. Nor are there any animal models that could enable the study of this important problem.
We found that clinically relevant concentrations of 5-fluorouracil (5-FU; a widely used chemotherapeutic agent) were toxic for both central nervous system (CNS) progenitor cells and non-dividing oligodendrocytes in vitro and in vivo. Short-term systemic administration of 5-FU caused both acute CNS damage and a syndrome of progressively worsening delayed damage to myelinated tracts of the CNS associated with altered transcriptional regulation in oligodendrocytes and extensive myelin pathology. Functional analysis also provided the first demonstration of delayed effects of chemotherapy on the latency of impulse conduction in the auditory system, offering the possibility of non-invasive analysis of myelin damage associated with cancer treatment.
Our studies demonstrate that systemic treatment with a single chemotherapeutic agent, 5-FU, is sufficient to cause a syndrome of delayed CNS damage and provide the first animal model of delayed damage to white-matter tracts of individuals treated with systemic chemotherapy. Unlike that caused by local irradiation, the degeneration caused by 5-FU treatment did not correlate with either chronic inflammation or extensive vascular damage and appears to represent a new class of delayed degenerative damage in the CNS.
Most treatments used to kill cancer cells also kill a diverse range of normal cell types, leading to a broad range of adverse side effects in multiple organ systems. In the hematopoietic system, the tissue in which such adverse effects have been most extensively studied, their detailed analysis has led to the discoveries that bone marrow transplants and cytokine therapies can improve the outcome of many forms of cancer treatment. In contrast, there has been no comparable level of analysis for most other organ systems compromised by cancer treatments.
One of the tissues for which adverse side effects of cancer treatment are clinically important is the central nervous system (CNS). Although it has long been appreciated that targeted irradiation of the CNS may be associated with neurological damage, it has become increasingly clear that systemic chemotherapy for non-CNS cancers also can have a wide range of undesirable effects. This has been perhaps most extensively studied in the context of breast cancer (for examples, see [1–13]). For example, it has been reported that 18% of all breast cancer patients receiving standard-dose chemotherapy show cognitive defects after treatment , with such problems reported in over 30% of patients examined two years after treatment with high-dose chemotherapy ; this is a greater than eightfold increase over the frequency of such changes in control patients. Adverse neurological sequelae include such complications as leukoencephalopathy, seizures and cerebral infarctions, as well as cognitive impairment [14–18]. Adverse neurological effects have been observed with almost all categories of chemotherapeutic agents [19–22], including antimetabolites (such as cytosine arabinoside (Ara-C) , 5-fluorouracil (5-FU) [24, 25], methotrexate [26–28], DNA cross-linking agents (such as BCNU  and cisplatin ) and even anti-hormonal agents [31–37]. Given the large number of individuals treated for cancer, these adverse neurological changes easily may affect as many people as some of the more extensively studied neurological syndromes.
One of the most puzzling aspects of chemotherapy-induced damage to the CNS is the occurrence of toxicity reactions with a delayed onset. Although this has been particularly well documented in children exposed to both chemotherapy and cranial irradiation [15, 38–47], delayed toxicity reactions also occur in individuals treated only with systemic chemotherapy. For example, white matter changes induced by high-dose chemotherapy for breast cancer, and detected in up to 70% of treated individuals, usually arise only several months after treatment is completed [48, 49].
One widely used chemotherapeutic agent associated with both acute and delayed CNS toxicities is 5-FU. Acute CNS toxicities associated with systemically administered 5-FU (most frequently in combination with other chemotherapeutic agents) include a pancerebellar syndrome and subacute encephalopathy with severe cognitive dysfunction, such as confusion, disorientation, headache, lethargy and seizures. With high-dose treatment, as many as 40% of patients show severe neurological impairments that may progress to coma [50–52]. In addition, a delayed cerebral demyelinating syndrome reminiscent of multifocal leukoencephalopathy has been increasingly identified following treatment with drug regimens that include 5-FU, with diagnostic findings obtained by both magnetic resonance imaging (MRI) and analysis of tissue pathology [24, 53–78].
Despite the existence of multiple clinical studies describing delayed CNS damage associated with systemic exposure to chemotherapy, almost nothing is known about the basis for these effects. For example, because of the multi-drug regimens most frequently used in cancer treatment, it is not even known whether delayed toxicities require exposure to multiple drugs. Nor is it known whether such delayed changes can be caused solely by exposure to chemotherapy or if they represent a combination of the response to chemotherapy and, for example, physiological changes caused by the body's reaction to the presence of a tumor. In addition, the roles of ongoing inflammation or damage to the vasculature in inducing such delayed CNS damage are wholly unknown. Moreover, the absence of animal models for the study of delayed damage makes progress in the biological analysis of this important problem difficult.
Here, we demonstrate that delayed CNS damage in mice is caused by short-term systemic treatment with 5-FU. Our experiments demonstrate that CNS progenitor cells and oligodendrocytes are vulnerable to clinically relevant concentrations of 5-FU in vitro and in vivo. More importantly, 5-FU exposure in vivo was followed by degenerative changes that were markedly worse than those observed shortly after completion of chemotherapy and that grew still worse with time. Systemic application of 5-FU in vivo (three injections interperitoneally (i.p.) over 5 days) was sufficient to induce delayed degeneration of CNS white-matter tracts. We observed this degeneration using functional, cytological and ultrastructural analysis and by altered expression of the transcriptional regulator Olig2, which is essential for generation of functional oligodendrocytes. The degeneration was not associated with either the prolonged inflammation or the extensive vascular damage to the CNS caused by local irradiation. This study provides the first animal model of delayed damage to white-matter tracts of individuals treated with systemic chemotherapy and suggests that this important clinical problem might represent a new class of damage, different from that induced by local CNS irradiation.
We first examined the effects of exposure to clinically relevant concentrations of 5-FU in vitro, as in our previous studies on the chemotherapeutic agents cisplatin, BCNU (carmustine) and cytarabine . To estimate clinically relevant concentrations, we used the following information: routinely used continuous intravenous infusions of 5-FU can result in steady-state plasma and cerebrospinal fluid (CSF) concentrations in the range 0.3–71.0 μM, and continuous pump infusions result in 3- to 25-fold higher levels of exposure . High-dose (bolus) injections of 5-FU can even expose brain tissue to peak concentrations in the millimolar range [80, 81], with tri-exponential elimination half-time values of 2, 12 and 124 minutes , and CSF elimination half-times can be greatly extended after localized application to brain tissue using slowly biodegradable polymer microspheres [83, 84].
We found that progenitor cells and oligodendrocytes were vulnerable to clinically relevant levels of 5-FU. Exposure to 1 μM 5-FU for 24 hours (which is at the low end of the range of concentrations observed in the CSF of individuals treated with 5-FU by intravenous infusion ) caused a 55–70% reduction in viability of dividing O-2A/OPCs and also of non-dividing oligodendrocytes (Figure 1b). Exposure for 24 hours to 5 μM 5-FU killed about 80% of O-2A/OPCs and oligodendrocytes and more than 50% of GRP cells and HUVECs. Even at concentrations as low as 0.5 μM, 5-FU reduced the survival of O-2A/OPCs and oligodendrocytes by approximately 45%. Exposure to 5 μM 5-FU for 5 days killed almost all the oligodendrocytes (Figure 1c), and exposure to 1 mM 5-FU for just 1 hour reduced the number of viable oligodendrocytes by more than 55% (Figure 1d). In marked contrast, these doses of 5-FU had no effect on any of a variety of cancer cell lines, in agreement with previous studies on the breast cancer lines examined [90, 91]. Thus, cell division was not sufficient to confer vulnerability to 5-FU, and a lack of division by oligodendrocytes was not sufficient to make them resistant.
Purified astrocytes and rapidly dividing NSCs were less vulnerable to 5-FU than progenitor cells and oligodendrocytes (Figure 1b–d), although even these populations showed some evidence of vulnerability when exposure time was extended to 120 hours (as is often associated with continuous intravenous infusion; Figure 1c). The relative resistance of NSCs to 5-FU (as compared with O-2A/OPCs, GRP cells and oligodendrocytes) demonstrates that, even in primary cell populations, cell division is not by itself sufficient to confer vulnerability to 5-FU.
We next investigated whether exposure to sublethal concentrations of 5-FU would disrupt normal progenitor cell function by suppressing cell division, as we have seen with BCNU, cisplatin and cytarabine . Analysis of clonal growth in these experiments was used as it provides more detailed information on both cell division and progenitor cell differentiation than does analysis in mass culture. Progenitors, grown at cell densities that allow the study of single clonally derived families of cells (as in, for example, [92–94]), were exposed for 24 hours to 0.05 μM 5-FU (a concentration equivalent to less than 10% of that found in the CSF in standard-dose applications ), followed by 5 days of clonal growth.
Confocal microscopic analysis of immunolabeling and terminal deoxynucleotidyltransferase-mediated dUTP nick-end labeling (TUNEL) staining confirmed that the vulnerability of cells in vivo was similar to that observed in vitro (Figure 3d). In untreated animals, TUNEL+ cells (which are apoptotic cells) were very rare, but such cells were frequently found in the SVZ, DG and CC of animals receiving chemotherapy. In the SVZ and DG, the majority of TUNEL+ cells observed after 5-FU treatment were double-cortin+ (DCX+) neuronal progenitors , followed by GFAP+ cells (a subset of which may be stem cells in the SVZ ). The SVZ also contained a smaller number of TUNEL+ Olig2+cells, which could be ancestors of oligodendrocytes [97, 98]. In the DG, there was also a very small amount of NeuN+ mature neurons that were TUNEL+. In the CC, approximately 70% of the TUNEL+ cells were Olig2+, and thus would be either oligodendrocyte progenitors or oligodendrocytes. Most of the remaining TUNEL+ cells in the CC were GFAP+, which in this tissue would mean they are astrocytes. The specificity of TUNEL labeling is demonstrated by representative images of TUNEL+ cells that were DCX+, Olig2+ or GFAP+ (Additional data file 1).
In the SVZ, 5-FU exposure was associated with a 40.9 ± 2.6% decrease in numbers of BrdU+ cells on day 1, with a transient re-population of BrdU+ cells at days 7 and 14, followed by a subsequent decrease in animals examined at day 56 and 6 months after completion of treatment. It was striking that the most significant inhibition of DNA synthesis in the SVZ was seen at 6 months post-treatment, when there was a 67.7 ± 3.0% decrease in the number of BrdU+ cells compared with control animals (Figure 4a). In the DG, suppression of DNA synthesis started on day 14 after treatment, and the greatest inhibition (60.7 ± 7.8%) was also seen at 6 months (Figure 4c). In the CC, in contrast, cell proliferation was significantly suppressed at all time points examined (Figure 4b).
To determine whether exposure to 5-FU preferentially reduced DNA synthesis in any particular cell population(s) in vivo, we combined BrdU labeling with cell-type-specific antibodies and analyzed individual BrdU+ cells by confocal microscopy (see Materials and methods). We analyzed the CNS of animals sacrificed 1 day and 56 days after the completion of 5-FU treatment in order to examine the acute and long-term effects of treatment.
In contrast with effects on putative O-2A/OPCs, there was a somewhat enhanced loss of DCX+ cells (which would have been neuronal progenitors ) from among the BrdU+ population in both the SVZ and the DG (Figure 5a–d). In the SVZ, at 1 day after treatment, there was a disproportionate and significant reduction in the percentage of DCX+ BrdU+ cells, which represented 50.2 ± 1.9% of the cells incorporating BrdU in control animals and only 30.7 ± 3.9% in animals treated with three injections of 5-FU (p < 0.01). At day 56 the proportion of BrdU+ cells that were DCX+ was not different between controls and treated animals, although the total number of BrdU+ cells in the SVZ of treated animals continued to be significantly lower than that of the control group (only 67.7 ± 4.9% compared with control animals at the same time point; p < 0.01). In contrast, in the DG, a reduction in the number of DCX+ cells was also seen, both at day 1 (with DCX+ cells comprising only 34.3 ± 4.4% of the BrdU+ population in 5-FU-treated mice compared with 63.2 ± 3.4% in the control mice; p < 0.01), and at day 56 (23.7 ± 3.9% in 5-FU-treated mice versus 52.2 ± 2.8% in the control mice; p < 0.01). In the CC, exposure to 5-FU was also associated with a small increase in the proportion of GFAP+ cells among the BrdU-incorporating populations at both day 1 and day 56, although such cells continued to represent a minority of the BrdU+ cells in this tissue. In addition, BrdU+ cells that were not labeled with any of the cell-type-specific antibodies used in these studies were more prominent in treated animals than in controls at day 1 (but not at day 56) in the SVZ and were found in the DG at both time points (data not shown). The DG was the only tissue in which these unlabeled cells made up >10% of the entire BrdU+ population. Such cells represented about 40% and 50% of all BrdU-labeled cells in 5-FU-treated animals at days 1 and 56, respectively, compared with about 2% and 20%, respectively, of all BrdU-labeled cells in control animals.
To determine whether the exposure of experimental animals to 5-FU was associated with functional impairment, we investigated hearing function in treated animals at various time points after treatment. Damage to the auditory system is a well known correlate of treatments with cisplatin [102, 103]. This damage is associated with death of cochlear outer hair cells, increases in the auditory brainstem response (ABR) thresholds and decreases in transient evoked otoacoustic emissions (TEOAE) and distortion product otoacoustic emissions (DPOAE), all of which are indicators of compromised cochlear function.
We examined the DPOAE as an indicator of cochlear function and ABRs to provide information on changes in conduction velocity from the ear to the brain, an indicator of myelination status. Different peaks (called P1, P2, and so on) in the ABR response are thought to correspond to different steps in the transmission of information, and prior analysis of ABR inter-peak latencies shows that loss of myelin (as in, for example, CNS myelin-deficient mouse models [104, 105]) causes increases in specific ABR inter-peak latencies (P2-P1 and P3-P1). Such measurements have been used by several investigators to study myelination-associated problems in impulse conduction in children with iron deficiency [106–109].
Our analysis of auditory function in 5-FU-treated animals revealed what seems to be a previously unrecognized consequence of chemotherapy exposure: increased latencies of impulse transmission. Consistent with the absence from the literature of reported deficits in cochlear function associated with 5-FU administration, DPOAEs in treated animals were not significantly different from those in untreated animals. In contrast, treated animals showed a progressive alteration in ABRs when inter-peak latencies were examined at days 1, 7, 14 and 56 after completion of treatment and compared with baseline measurements of each individual 1 day before 5-FU application.
The results of our ABR analysis raised the possibility that 5-FU-treated animals show a syndrome of delayed white matter damage. Although our analysis of cell division and cell death following systemic treatment with 5-FU revealed a long-lasting suppression of cell division in the CC, we observed only an increased level of apoptosis in this tissue at one day after the cessation of treatment. We therefore conducted a more detailed analysis of the CC, the major myelinated tract in the rodent CNS.
The occurrence of delayed damage to the CNS following irradiation has been a subject of interest for many years, and both vascular damage and delayed inflammatory reactions have been implicated as being important in the adverse effects of this treatment on the CNS [113–116]. To begin to determine whether similar mechanisms might be relevant to analysis of the delayed effects of 5-FU administration, we examined microglial activation and endothelial cell apoptosis in 5-FU-treated animals.
Damage to the vasculature following irradiation has also been suggested as a possible contributor to delayed CNS damage but, as for inflammation, it seems unlikely that such damage contributed to the delayed effects of 5-FU administration. Analysis of TUNEL labeling in 5-FU-treated animals revealed a subset of animals (four out of ten total treated animals from two independent experiments) that showed markedly increased diffuse TUNEL+ nuclei; the distribution, morphology and size of these nuclei resembled those of the microvasculature endothelial cells of the CNS (Figure 11c–e). Double-labeling to visualize expression of the vascular endothelial cell marker PECAM/CD31  confirmed that these apoptotic cells were vascular endothelial cells (Figure 11c–e). However, these indications of vascular damage were seen only in a subset of animals examined 1 day after the cessation of treatment and were not observed in animals examined at any later time points.
Our studies demonstrate that systemic treatment with 5-FU is associated with both acute and delayed toxicity reactions, outcomes that are of particular concern because of the use of this agent in the treatment of many cancers. As in our recent studies on cisplatin, cytarabine and carmustine , in vitro analysis of vulnerability to 5-FU revealed that lineage-restricted progenitor cells of the CNS and non-dividing oligodendrocytes were vulnerable to the effects of 5-FU at or below clinically relevant exposure levels. Thus, toxicity of 5-FU was not limited to dividing cells. Toxicity of 5-FU was not limited to induction of cell death and was also associated with suppression of O-2A/OPC division, even when applied transiently at exposure levels that represent small fractions of the CNS concentrations achieved during cancer treatment. Although previous in vitro studies on neurons and oligodendrocytes also observed vulnerability of these cells [119, 120], the effective concentrations used in our present study are considerably lower than those used in previous studies. Our in vitro analyses also predicted the acute in vivo effects of 5-FU with considerable accuracy, just as was the case with our previous studies on cisplatin, BCNU and cytarabine . 5-FU exposure transiently increased apoptosis and suppressed proliferation for extended periods of time in the SVZ, DG and CC. Cell-type-specific analyses confirmed that the main populations affected in vivo were also progenitor cells and oligodendrocytes. Suppression of progenitor cell proliferation was also seen in vitro in analyses of division and differentiation in clonal families of cells.
This study is the first to demonstrate that delayed degenerative damage can be caused by systemic application of a single chemotherapeutic agent (5-FU) and does not require the concurrent presence of cancer to manifest, as well as the first to provide an animal model of delayed damage to white matter associated with the systemic administration of chemotherapy. These results are of particular interest in the context of many clinical reports that have identified neurotoxicity as a complication of treatment regimens in which 5-FU is a component. Although most reports of 5-FU-associated neurotoxicity indicate a relatively acute onset, a delayed demyelinating cerebral complication reminiscent of multifocal leukoencephalopathy has also been increasingly reported in patients treated with chemotherapy regimens that include 5-FU [24, 53–78]. Although 5-FU is used most extensively in the treatment of colorectal cancers, it is also an important component of adjuvant therapies for the treatment of a variety of other cancers, including breast [121–128], gastric [129–136], pancreatic [137–142] and lung [129, 143, 144], and is thus given to large numbers of patients. Neurological symptoms may occur in some patients several months after adjuvant therapy with 5-FU and include declines in mental status, ataxia and the appearance of prominent multifocal enhancing white matter lesions detectable by MRI. In addition, both acute and delayed neurological side effects have been observed for many other chemotherapeutic agents [9, 14, 23, 26, 27, 29–31, 33, 145–154], and it will be of interest to determine whether the pattern of degenerative changes observed with 5-FU exposure is representative of delayed changes associated with other chemotherapeutic agents.
We also have provided several novel findings regarding the problem of delayed white matter damage caused by 5-FU exposure. Our findings of aberrant regulation of Olig2 expression, with the presence of many Olig2-negative oligodendrocytes at 56 days after treatment, provide the first indication that chemotherapy alters the normal expression of important transcriptional regulators in oligodendrocytes. Our ultrastructural studies demonstrate extensive myelin pathology at this time point, along with indications of neuronal pathology. It is not yet known whether damage to myelin precedes damage to neurons (as is thought to occur in multiple sclerosis (see, for example [155–162]), or whether neuronal damage occurs concurrent with or preceding myelin pathology. The vulnerability of oligodendrocytes to 5-FU in vitro and the increased apoptosis in these cells following 5-FU exposure in vivo, however, suggests strongly that oligodendrocytes are a direct target of this anti-metabolite. Although this is a somewhat surprising result (in that 5-FU has been thought to target dividing cells specifically, while oligodendrocytes do not divide in the conditions used in our experiments), previous studies have shown that experimental derivatives of 5-FU, and its metabolites, also cause myelin damage in vitro and in vivo [163, 164]. Whether 5-FU derivatives such as capecitabine (an orally active form of 5-FU) cause similar damage is not yet known, but the presence of the activating enzyme for this drug (thymidine phosphorylase) in white-matter tracts  makes this a matter of concern.
Although the continuing presence of at least some BrdU+ cells in the CC at 56 days offered the possibility that the damage to myelin occurring at this time point might be reversible, analysis at 6 months demonstrated a striking loss of cells and of MBP. Thus, it appears that even a short-term exposure to 5-FU can cause long-term and apparently irreversible damage to white-matter tracts.
Analysis of alterations in myelination caused by chemotherapy would benefit enormously from the ability to conduct functional analysis in a non-invasive manner, and our analysis of alterations in inter-peak latencies in ABRs provides a tool of particular potential interest in this regard, as well as revealing a novel form of chemotherapy-induced neurological damage. Despite extensive investigations of ototoxicity induced by exposure to cisplatin (for reviews, see [166–171]), such studies appear to have been focused exclusively on the effects of chemotherapy on hair cells and cochlear function and have not used ABR analysis of inter-peak latencies to analyze changes that may be related to white matter damage. Thus, our ABR analyses seem to provide the first demonstration of adverse effects of chemotherapy on a functional outcome related to CNS myelination. ABR inter-peak latency analysis has been used, however, to study myelination-related maturation and function of the auditory pathway in normal infants in conditions in which myelination is compromised (for example, iron deficiency, fetal cocaine syndrome) and in experimental animals [105, 106, 108, 172–175]. Thus, this approach provides a non-invasive functional analysis of a myelination-related outcome measure that can be used in both experimental animals and human populations.
The progressive alterations in ABR inter-peak latencies observed in our studies also highlight the fact that at least some of the delayed damage associated with 5-FU administration is greater than the damage observed acutely. The ability to study progressive deterioration in the same animals over prolonged periods will make this approach of particular value in further investigations of these changes. Moreover, because of the ease of conducting such studies in humans, such analysis may provide a simple, non-invasive approach to the analysis of adverse effects on white matter complementary to the imaging-based detection of leukoencephalopathy.
The underlying causes of delayed damage induced by chemotherapy will be the subject of continued investigation, but the observations that vascular damage and inflammatory reactions were rare and were observed only at short intervals after completion of treatment makes it seem unlikely that these are causally important. This is in striking contrast to the effects of irradiation, where inflammation is thought to be essential in delayed suppression of hippocampal neurogenesis [115, 116]. It is possible that the appearance of delayed damage following 5-FU treatment reflects the combined effects of delayed oligodendrocyte death and a loss of the progenitor cell populations required for replacement. Recent findings that aging is associated with a loss of expression of important transcriptional regulators, including Olig2, in oligodendrocytes  and may be associated with degenerative white matter changes [177–185] also raises the possibility, however, that the effect of 5-FU results from an acceleration of the normal aging processes.
Our findings also raise the question of whether multiple pathological changes contribute to the effects of chemotherapy on cognition. The ability of irradiation to the CNS to suppress the generation of new neurons in the hippocampus has been suggested to be relevant to the understanding of cognitive impairment associated with this particular form of cancer treatment . Although reduced numbers of dividing hippocampal neuronal progenitors are also seen in association with exposure to 5-FU, BCNU or cytarabine , the additional damage to white-matter tracts caused by chemotherapy would be expected to impair normal neuronal impulse conduction (in accordance with the changes in ABR latency seen here) and thus might also contribute to alterations in cognition. It is particularly interesting in this regard that recent studies on breast cancer patients treated with adjuvant chemotherapy have revealed that, relative to controls, patients had slower speeded processing and altered fractional anisotropy (a measure of white matter integrity) in the corpus callosum. It has been suggested that these white matter changes are related to the cognitive deficits that may be associated with treatment with systemic chemotherapy .
As adverse effects on several normal tissues have been observed for almost all classes of chemotherapeutic agents [19–22, 187] (including alkylating agents [29, 30], anti-metabolites [23–26, 57], methotrexate  and even anti-hormonal agents [31–37]) and such treatments will clearly remain the standard of care for cancer patients for many years to come, the need to understand such damage better is great. Indeed, some of the most important advances in the treatment of cancer have emerged from the study of such damage, the necessary first step in its prevention. Moreover, evaluation of potential new therapeutics that does not include adequate analysis of these potential toxicities may lead to the approval of treatments that are no better than existing treatments in avoiding serious damage to normal tissue. The clinical study of such side effects does not provide the experimental foundations required for the analysis of such problems. Indeed, treatment for neurological complications of 5-FU treatment has largely been ineffective so far, with some patients responding to immediate discontinuance of chemotherapy and steroid treatment [57, 60], but with others continuing to deteriorate and, in some severe cases, progressing to death . In contrast, recent studies on the toxicities in vitro and in vivo of several chemotherapeutic agents [79, 189], and our discovery of an animal model for delayed damage to the CNS caused by chemotherapy, provide experimental foundations that should prove of great value in the discovery and evaluation of therapies that either allow selective killing of cancer cells or offer selective protection to the normal cells of the body.
Most materials and methods are as described in  and are presented here in brief.
In vitro studies were performed on purified cultures of primary CNS cells isolated from the developing rat CNS. Purified populations of neuroepithelial stem cells, neuron restricted precursor cells, glial restricted precursor cells, O-2A/OPCs, oligodendrocytes and astrocytes were all prepared and grown as described previously . HUVECs (Cambrex) were cultured in endothelial growth medium (EGM-2) and used within two passages after thawing. Cancer cell lines used were established breast cancer cell lines (MCF-7 and MDA-MB-231), ovarian cancer (ES-2) cells, L1210 lymphocytic leukemia and EL-4 lymphoma cells, a meningioma cell line and two cell lines isolated from patients with glioblastoma multiforme (UT-4 and T98 cell lines); these were grown as previously described .
In vitro toxicity studies involved microscopic analysis of staining with the 3,(4,5-dimethylthiazol-2-yl) 2,5-diphenyl-tetrazoliumbromide (MTT) assay in combination with 4',6-diamidino-2-phenylindole (DAPI) and staining with cell-type-specific antibodies, as previously described . Each experiment was carried out in quadruplicate and was repeated at least twice in independent experiments. Data points represent means from single experiments and error bars shown in figures represent ± standard error of the mean (s.e.m).
Clonal analysis of O-2A/OPC division and differentiation was carried out as described previously [92–94]. One day after plating, cells were exposed for 24 h to low-dose 5-FU (0.01 μM), a concentration that did not cause significant killing of O-2A/OPCs in mass culture. The number of undifferentiated progenitors and differentiated oligodendrocytes was determined in each individual clone from a total of 100 clones in each condition by morphological examination and by immunostaining to confirm cell-type identification. Experiments were performed in triplicate in at least two independent experiments.
For in vivo experiments, 6–8-week-old CBA mice were treated with chemotherapy under approved protocols. 5-FU (Sigma) was administered by i.p. injections. Animals received 5-FU as three consecutive injections every other day (40 mg kg-1 body weight). Control animals received equal amounts of 0.9% NaCl i.p. Animals were sacrificed on days 1, 7, 14 and 56 and 6 months after completion of treatment with 5-FU (where day 0 is the time of the last injection of the agent). For all in vivo experiments, animals were perfused transcardially with 4% paraformaldehyde in phosphate buffer (pH 7.4), under deep anesthesia using Avertin (tribromoethanol; Sigma; 250 mg kg-1, 1.2% solution).
We chose the in vivo dosage on the basis of conversion from human treatment dosage to an equivalent mouse dosage and previous animal studies of 5-FU effects in mice. As is standard practice, we used a conversion factor of 3 [190–194] to calculate the equivalent mouse dose range (20–1,167 mg kg-1) from the clinical human treatment dose range (60–3,500 mg m-2). On the basis of animal studies in mice (for example, [195–197]), in which doses of 5-FU used ranged from 40–200 mg kg-1, we first used 60 mg kg-1 every other day for three doses as the initial trial treatment. As this treatment caused death in half of treated CBA mice one week after completion of treatment, we lowered the dosage to 40 mg kg-1, which was tolerated well by the animals (it caused less than a 10% increase in death over a six-month period compared with sham-treated controls). The differences in tolerance to 5-FU treatment between our study and others may result from the different mouse strains used. In clinical practice, patients are often given the highest tolerated dosage of chemotherapy to achieve maximum fractional kill of the malignant cells. Considering this situation, we determined the appropriate in vivo dosage to be 40 mg kg-1.
All in vivo analysis was carried out as described in . Using free-floating sections (40 μm), detection of nuclear profiles with DNA fragmentation, a hallmark of apoptosis, was performed using a TUNEL assay combined with DAPI counterstaining to visualize nuclear profiles. To combine TUNEL staining with immunofluorescence staining for different cell lineage markers, TUNEL staining was performed first, followed by labeling with one of the following primary antibodies for 24–48 h: mouse anti-NeuN (1:500, Chemicon), goat anti-DCX (doublecortin; 1:500, Santa Cruz), rat anti-S-100β (1:2,500, Swant), mouse anti-MBP (1:1,000, Chemicon), mouse anti-GFAP (1:2,500, DAKO), rabbit anti-Olig2 (1:1,000, a gift of David Rowitch), mouse anti-CC1 (1:300, Calbiochem; which was used under conditions  in which specificity for oligodendrocytes is preserved and no double-labeling with GFAP+ astrocytes was observed), rat anti-CD31/PECAM (1:500, Chemicon) and rat anti-F4/80 (Abcam). All secondary antibodies, generated in donkey (anti-rat, anti-rabbit, anti-goat and anti-mouse), were coupled to TritC, FitC or Cy5 (Jackson ImmunoResearch) for in vivo staining and were used according to the species of primary antibody. Fluorescent signals were detected using a confocal laser scanning microscope Leica TCS SP2 and a 40× oil immersion lens, with pinhole settings corresponding to an optical thickness of less than 2 μm used to avoid false positive signals from adjacent cells.
To label the proportion of dividing cells engaged in DNA synthesis in vivo, mice received a single injection of BrdU (50 mg kg-1 body weight, dissolved in 0.9% NaCl) given i.p. 4 h before perfusion. Anti-BrdU antibody was used to identify BrdU+ cells by standard techniques (as in ). A minimum of 50 BrdU+ cells was counted for each labeling condition in each animal (n = 3 animals in each group examined), with the sole exception of the DG of the animals examined 56 days after cytarabine treatment, for which an identical number of sections were examined as in controls, but the frequency of labeled cells was not sufficient to reveal 50 cells in these sections. Quantification of BrdU+ cells was accomplished with unbiased counting methods. BrdU immunoreactive nuclei were counted in one focal plane to avoid over-sampling. Brain structures were sampled either by selecting predetermined areas on each section (lateral SVZ) or by analyzing the entire structure on each section (CC and DG). Differences were considered significant when p < 0.01.
BrdU+ cells were counted in every sixth section (40 μm) from a coronal series between interaural anterior-posterior (AP) +5.2 mm and AP +3.9 mm (the anterior commissure crossing). BrdU+ cells were counted along the lateral ventricular wall up to 200 μm distance from the lateral ventricle wall.
BrdU+ cells were counted in every sixth section (40 μm) from a coronal series between interaural AP +5.2 mm and AP +3.0 mm in the entire extension of the rostral and medial part of the CC and analyzed as for the SVZ.
BrdU+ cells were counted in every sixth section (40 μm) from a coronal series between interaural AP +2.5 mm and AP +1.1 mm. BrdU+ cells were counted in the area of the dentate gyrus, including the hilus, subgranular zone and the granule cell layer and analyzed as for the SVZ. Quantitative data in all figures are presented as mean percentage normalized to control animals. Error bars represent ± s.e.m.
To analyze BrdU incorporation in specific cell types, anti-BrdU immunostaining was combined with immunolabeling to identify DCX+ neuronal precursor cells , Olig2+ oligodendrocyte precursor cells (defined as cells that were BrdU+ and Olig2+, in order to discriminate these cells from Olig2+ non-dividing oligodendrocytes [97, 98, 101]) and GFAP+ cells; the latter would have been astrocytes in the CC or DG or, in the SVZ, may also have been stem cells . Labeling and confocal analysis was carried out as for the combination of immunolabeling with TUNEL staining.
Baseline ABRs were measured in each animal one day before initiation of treatment. After treatment ended, follow-up ABR tests were conducted on each animal at various points during a time course of 56 days. For the measurement, mice were anesthetized with xylazine (20 mg kg-1 i.p.) and ketamine (100 mg kg-1 i.p.) . Needle electrodes were inserted at the vertex and pinna of the ear, with a ground near the tail. ABR potentials were evoked with click stimuli at 80 dB SPL. The response was amplified (10,000×) and 1,024 responses were averaged with an analog-digital board in a LabVIEW-driven data-acquisition system. For comparison of change of latencies, the wave peaks P1, P2 and P3 were identified by visual inspection at recorded wave forms, and the inter-peak latencies of wave P2-P1 and P3-P1 computed. The change of inter-peak latencies was calculated as Lt – L0 (where Lt is the inter-peak latency values at day 1, day 7, day 14, or day 56 post-treatment; and L0 is the baseline inter-peak latency values one day before treatment initiation).
Digital images were captured using a Nikon Eclipse E400 upright microscope with a spot camera (Diagnostic Instruments) and the spot advanced software for Macintosh (Diagnostic Instruments), or using the confocal laser scanning microscope (Leica TCS SP2). Paired or unpaired Student t-tests were used for statistical analyses where applicable.
The animals were anesthetized and injected with heparin and perfused with a 0.1 M phosphate buffered 4.0% paraformaldehyde/2.5% glutaraldehyde fixative. The brain was removed and allowed to fix overnight at 4°C and the CC was then sectioned sagitally and coronally at approximately 1.0 mm. The sections were rinsed in 0.1 M sodium phosphate buffer, post-fixed in buffered 1.0% osmium tetroxide, dehydrated in a graded series of ethanol to 100%, transitioned to propylene oxide and infiltrated with EPON/Araldite epoxy resin overnight. They were embedded into mold capsules (BEEM) and polymerized for 2 days at 70°C. Sections of 1 μm were cut on to glass slides and stained with toluidine blue to determine areas to be thin sectioned at 70 nm onto grids. The grids were stained sequentially with uranyl acetate and lead citrate. A Hitachi 7100 transmission electron microscope was used to examine and digitally capture images using a MegaView III digital camera (Soft Imaging).
Additional data file 1 shows a representative confocal micrograph of TUNEL+ cells co-labeled with cell-type-specific markers. (A-A") show a TUNEL+ DCX+ cell in the sub-ventricular zone; (B-B") show a TUNEL+ Olig2+ cell in the CC; (C-C") show a TUNEL+ GFAP+ cell in the CC. The scale bars represent 10 μm.
It is a pleasure to acknowledge the many helpful discussions with our colleagues regarding this research, in particular Chris Pröschel and Hartmut Land. This research was funded with generous support from the National Institutes of Health (HD39702 and NS44701), from the Susan B. Komen Foundation for the Cure and from the Wilmot Cancer Center.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.