We found that normal neural progenitor cells and oligodendrocytes of the CNS are exceptionally vulnerable to the toxic effects of the chemotherapeutic agents BCNU, cisplatin, and cytarabine. Vulnerability to these drugs was observed for all classes of lineage-restricted progenitor cells that can be readily grown as purified cell populations. Moreover, vulnerability was not restricted to dividing cells, as nondividing oligodendrocytes were also targets of these drugs, at exposure levels routinely achieved during treatment. In vitro analyses of purified cell populations were highly predictive of effects seen following systemic treatment with any of the chemotherapeutic agents in vivo. Comparative analysis of multiple cancer cell lines from different tissues only identified one cell line in which vulnerability was comparable to that observed for primary neural progenitor cells, with most such cell lines being more resistant to these agents than the normal cells (despite often being chosen because of their previous use in studies on the response to the drugs studied). Thus, it appears that the vulnerability of multiple normal cell populations of the CNS to cisplatin, BCNU, and cytarabine rivals the vulnerability of cancer cells themselves. The fact that toxicities for neuronal precursors, glial precursors, and oligodendrocytes, and toxicity in three different regions of the CNS, are associated with systemic application of chemotherapeutic drugs is of particular concern, as such toxicity would be applicable to treatment of all forms of cancer. Moreover, our studies demonstrate that the adverse effects of systemic application are not limited to the classes of DNA cross-linking agents represented by BCNU and cisplatin, but also are observed with the antimetabolite cytarabine. Thus, the adverse effects observed in the present studies may be relevant in understanding the side effects of multiple classes of chemotherapeutic drugs.
This is the first study of which we are aware that demonstrates that neural progenitor cells and oligodendrocytes are exceptionally vulnerable to the action of chemotherapeutic drugs in vitro
and in vivo
, even when applied extra-cranially. This study also suggests that, at least in the CNS, it is progenitor cells and not stem cells that are the most vulnerable targets. Adverse effects are known to occur clinically with all the agents we studied, both acutely and as delayed neurotoxicities (such as cognitive impairment) that may only become apparent years after treatment. For example, BCNU treatment has been associated with significant changes in mental status and with white matter degeneration [23
]. Cisplatin at high doses has been associated with leukoencephalopathy and destruction of CNS white matter [25
]. Application of cytarabine, the third drug examined in our studies, has also been associated with acute encephalopathy, confusion, memory loss, and white matter changes [25
]. The vulnerability of neural progenitor cells and oligodendrocytes to these drugs, which was also observed in our antigenic analysis of TUNEL-labeled cells in the CNS of animals exposed to BCNU in vivo
and BrdU-labeled cells exposed to either BCNU or cytarabine, may provide an explanation for the neurotoxic consequences of the treatments and also may be relevant to understanding long-term toxicities. It was particularly striking that O-A/OPCs and oligodendrocytes were one to two orders of magnitude more sensitive to cisplatin or cytarabine than has previously been observed in studies on multiple neuronal populations from both the CNS and the peripheral nervous system [65
The toxicities seen in our studies occurred well within the concentration ranges achieved for these agents in CSF during cancer therapy. For example, cisplatin toxicity for multiple neural cell types was observed at concentrations as low as 0.1 mM (Figure ). CSF concentrations for cisplatin in conventional or low-dose intravenous applications are between 0.6 and 2.8 μM [33
] but can reach up to 80 μM in high-dose applications [36
]. Moreover, much higher concentrations in brain tissue and CSF have been reported after intra-arterial applications, liposomal encapsulations, after previous radiation, or in cases of blood-brain barrier disruption [37
]. BCNU toxicity was observed at concentrations as low as 5 μM. This agent is highly lipophilic, and about 80% of plasma levels are detectable in CSF and brain tissue [34
]. CSF concentrations of BCNU are in the range of 8–10 μM after intravenous applications [35
], but can be 100–1,000-fold higher after local applications via biodegradable polymer wafers [39
]. Similarly, the toxicity of cytarabine was already apparent at concentrations as low as 0.1 μM or less, compared with CSF concentrations during conventional treatments in the range 0.1–0.3 μM, and concentrations that are ten times higher in high-dose applications, and 10,000 times higher following intrathecal application [63
In vitro studies further indicated that the toxicity of BCNU, cisplatin, and cytarabine is not limited to the induction of cell death, but is also associated with the suppression of cell division of O-2A/OPCs even when applied transiently at levels that cause little or no cell death, and that represent small fractions of the CSF concentrations achieved with systemic chemotherapy. The suppression of division was particularly striking in that a single transient exposure of dividing O-2A/OPCs to BCNU, cisplatin, or cytarabine was sufficient to cause a marked reduction in subsequent cell division at the clonal level. Such a loss of dividing cells would compromise the ability of dividing progenitor cells to contribute to repair processes, and could also contribute to long-term or delayed toxicity reactions.
The observations that BCNU, cisplatin, and cytarabine all cause dividing O-2A/OPCs to undergo a greater extent of oligodendrocyte generation are as predicted from our studies on the role of intracellular redox state in controlling the balance between self-renewal and differentiation [41
], and from observations that all three agents cause cells to become more oxidized [69
]. In our studies on redox regulation of precursor cell function, we found that O-2A/OPCs that are slightly (around 20%) more oxidized have a higher probability of undergoing differentiation, whether this oxidative status is due to cell-intrinsic mechanisms, exposure to pharmacological pro-oxidants or to physiological inducers of oligodendrocyte generation (such as thyroid hormone) [41
]. Even when this shift in differentiation probability is relatively small [76
], cumulative effects over multiple cell generations can lead to differentiation outcomes in which clonal composition is clearly different but in which analysis at delayed time points is required for the reduction in progenitor cell representation to translate into markedly smaller clonal sizes.
In vitro studies on purified cell populations appeared to accurately predict sensitivities observed in vivo. Combined analysis of TUNEL and antigen expression demonstrated death of both neuronal and glial precursors, as well as of oligodendrocytes. Combined analysis of BrdU labeling and antigen expression similarly revealed reductions in BrdU incorporation in neuronal precursors of the hippocampus and in glial precursor cells of the CC. The high level of correlation between in vitro and in vivo outcomes suggests that purified populations of the cell types studied can provide a means of rapidly analyzing other cancer therapies.
Although all chemotherapeutic drugs examined were associated with toxicity in vivo
, there were important differences between them. BCNU was associated with particularly severe and prolonged cell death in vivo
, while cell death induced by cisplatin was less severe and eventually returned to normal values. Cytarabine was associated with increased cell death for at least 14 days after treatment ended, with values tending towards or at base-line levels of TUNEL labeling at 56 days post-treatment. Whether the less severe effects of cisplatin in this regard were due to different drug characteristics in terms of blood-brain barrier permeability is not known (although, in this regard, it should be noted that cisplatin application in vivo
may actually cause opening of this barrier [77
All three agents examined were associated, moreover, with continued reductions in cell division in one or more CNS regions after treatment ended, suggesting a long-lasting depletion of populations required for cell replenishment. Nonetheless, the fact that some BrdU-incorporating cells remained in all brain regions examined raises the question of whether treatments analogous to those used to enhance bone-marrow function after cancer treatment may be applicable some day to enhancing the function of the normal dividing cells of the CNS during or after cancer treatment, possibly even using the same cytokines that are used to enhance cell repopulation from the bone marrow [78
The effects of cytarabine on the different cell populations that incorporated BrdU in vivo
were particularly surprising in the context of previous observations that cytarabine exposure in vivo
(delivered by infusion onto the cortex for 7 days) is associated with a repopulation of the SVZ after treatment ceases [21
]. In contrast, our own studies indicate that this repopulation of dividing cells does not occur in the CC or DG, and may not endure in the SVZ (Figure ). Although previous studies differ from our own in delivery methods and dosages applied, it may also be that the capacity for repopulation of dividing cells differs in different regions of the CNS. Moreover, it may be that the repopulation of the dividing cells of the SVZ is a transient phenomenon, as the latest time point examined in our studies was associated with a fall in the levels of BrdU incorporation to levels seen 1 day after treatment ended.
Taken together with recent studies on the effects of irradiation on the CNS [82
], our results indicate that damage to CNS progenitor cells is an apparent correlate of both the main treatments for cancer. Monje et al
] suggested that the adverse effects of irradiation on the hippocampus might be causally related to the neurological symptoms and cognitive decline associated with this treatment. This suggestion would also apply to the effects of chemotherapy.
There are many ways in which the effects of chemotherapy may be even more of a concern than the effects of irradiation, beginning with the fact that whereas radiation damage is caused by therapy targeted to the CNS, toxicity after chemotherapy also occurs after systemic administration of these compounds. Moreover, our studies also reveal that the range of CNS cell types vulnerable to the effects of chemotherapy is greater than has been studied for irradiation, and demonstrate toxicity of chemotherapeutic agents for glial progenitor cells and for oligodendrocytes, as well as for the hippocampal precursor cells that have been examined in studies on the effects of irradiation [82
]. Yet another difference between the effects of these two modes of treatment is that irradiation-associated impairment of neurogenesis appears to be a secondary effect of inflammation, and can thus be reduced with anti-inflammatory agents [83
]. In contrast, our preliminary analyses of chemotherapy-treated animals have not revealed any increased microglial activation, a hallmark of CNS inflammation (J. D. and M. N., unpublished work). Thus, there is presently no reason to think the adverse effects of chemotherapy might be ameliorated by control of inflammation. The two sets of studies also differed in severity of outcome, in that our study reveals a partial fall in the representation of DCX+
neuronal precursor cells whereas the studies on irradiation revealed a virtually complete lack of neurogenesis [82
]. While it will be of interest to extend examination of both treatment paradigms, it is nonetheless the case that both studies raise the concern that neurogenesis in the brain is vulnerable to both forms of cancer treatment.
Our studies have multiple implications for future strategies of cancer treatment. As doses of BCNU, cisplatin, and cytarabine that killed even chemosensitive cancer cell lines were equally or more toxic for neural progenitor cells and oligodendrocytes, it seems that any concentration of these chemotherapeutic agents sufficient to harm cancer cells may also damage many cell populations of the CNS. That cisplatin may have less severe long-term effects than BCNU might be construed as encouragement that less toxic treatments can be developed with existing chemotherapeutic agents. It is also possible, however, that our results actually understate the extent of damage that occurs in association with chemotherapy. Such treatment is typically applied for several courses over an extended period of time. Furthermore, current treatment protocols simultaneously apply multiple different chemotherapeutic agents. This issue is of particular concern in the light of reports that agents such as cisplatin or BCNU can cause opening of the blood-brain barrier [77
], which could allow entry of adjunctive non-lipophilic agents into the CNS. In addition, there are multiple therapeutic regimes associated with higher concentrations of drugs than those we have studied (for example, intra-arterial administration, liposome-encapsulated drugs, or locally applied biodegradable wafers in the treatment of brain tumors). Moreover, the advances that have been made in rescuing patients from the toxicity of chemotherapeutic agents for bone marrow have been associated with a tendency to apply yet higher doses of these agents, thus potentially increasing the risk of neurotoxicity.
As chemotherapy will remain a cornerstone of cancer therapy for the foreseeable future, the potential ramifications of this work for present and future cancer treatments seem clear. Plainly, it is of great importance to identify the neural populations at risk during any cancer treatment in order to develop means of reducing neurotoxicity and preserving the quality of life in long-term survivors. This is an issue of great concern, particularly in the light of recent studies favoring the use of more aggressive and high-dose regimens or of newer drugs that target receptor tyrosine kinase signaling pathways that are critical regulators of neural progenitor and stem-cell function. In this context, it will be of particular importance to include more profound analysis of CNS toxicity in the assessment of new candidate chemotherapeutic drugs, an evaluation that currently is not consistently performed. It will also be critical to understand why some patients have adverse side effects (whether neurological or non-neurological), whereas others are spared such damage, and to determine the value of low-dose (metronomic) therapies [85
] in avoiding damage to the CNS without compromising treatment outcome. In this regard, it is of concern that our in vitro
results raise the possibility that even exposure to very low levels of these agents may compromise progenitor cell division. It is clearly vital to identify therapeutic approaches that do not share these problems, either by enabling targeted killing of cancer cells or through selective protection of normal cells during cancer treatment. The strong correlations between our in vitro
and in vivo
analyses indicate that the same approaches we used to identify the reported toxicities can also provide rapid in vitro
screens for analyzing new therapies and discovering means of achieving selective protection or targeted killing. In light of the ease of use of these in vitro
and in vivo
assays, applying them early in the drug-discovery process may enable a more rapid identification of treatments able to eliminate cancer cells without compromising the patient's quality of life.