Commentary. Looking back

The opportunity to participate in this Festschrift provides an irresistible temptation to look back to the early 1950s when the Institute of Environmental Medicine was established and Norton Nelson's pivotal role in its devel? opment took shape. My participation was limited to four happy and productive years in the Institute from 1951 to 1955. It began with an invitation from Anthony Lanza and Norton Nelson to return to NYU from the Sloan Kettering Institute to which I had gone to join David Pressman on a study of antigens in normal and tumor cells. As I recall the circumstances, the invitation grew out of a grant from the Standard Oil Co. of New Jersey to the Institute for research in immunology involving skin. The grant pro? vided $10,000 per year and it was intended to support the research and to provide a stipend for a young faculty person, with the understanding that the stipend could be augmented because of its limited nature. When offered the grant, I accepted without hesitation. The amount it provided seemed a princely sum at the time; and the need I faced to support a growing family seemed also to be pro? vided for by the opportunity to practice medicine parttime in mid-town Manhattan, in an office that was a 10-min walk from the laboratory. The laboratory was then located in handsome quarters in the then new medical sciences building on 1st Avenue. I have never for a moment regreted accepting that of? fer. As an assistant professor, I was completely independent to develop a research program in immunology; and the stipulation that the skin be involved was no restriction at all, for it fit very well with my intention to depend upon allergic skin reactions to simple chemicals as an as? say system for examining the structure-function relationships in immune responses to haptens. My move to the In? stitute proved all the more congenial where I was joined by a group of effective enthusiastic colleagues. Sid Belman was nominally a technician, but quickly became a re? search colleague, then a graduate student, and later (af? ter I had gone to St. Louis) a member of the faculty. Leo Orris, and then Mary Carsten and Milton Tkbachnick as

Summary.-A serious limitation of chemotherapy for acute myeloid leukaemia (AML), Hodgkins disease and some classes of breast cancer is that, even when clinically evident disease responds well, the same chemotherapy when given during remission does not affect the rate of relapse after chemotherapeutic or surgical ablation of the primary disease. This cannot, in general, be caused by genetic adaptation of the residual cancer cells which renders them resistant to specific drugs, because after relapse further remissions can be obtained with the same drugs that were ineffective by chronic administration in prolonging remission. The resistance of the residual cells may arise from mechanisms such as inaccessibility for anatomical or other reasons, or because of a change in metabolic state which causes these cells temporarily to cease division, when they cannot be harmed by cycle-dependent drugs and repair damage sustained from cycle-independent drugs. Limited differentiation has been shown capable of reversal and this may be a mechanism which leads to quiescence and associated "resistance", particularly in the case of AML. Where such resistance occurs treatment during remission-or as an adjuvant to surgery and radiotherapy-may have to rely on mechanisms which are independent of cellular proliferation such as processes associated with graft-versus-host-disease or the induction of terminal differentiation. A model for studying the nature of resistance of residual cancer and for testing treatments that might be active against cancer cells in this state may be dormant metastases. The latter are malignant cells which appear to be in peaceful co-existence with their host and which in experimental systems have been induced to grow into lethal metastases by perturbation of the host by surgical trauma, by hormonal manipulation or by immunosuppression.
IT IS A WIDELY ACCEPTED PREMISE in cancer chemotherapy that the response to treatment of clinically evident disease provides a measure of responsiveness of residual disease. This concept, when coupled with the model developed by Skipper and Schabel (cf. Schabel, 1975) that the whole tumour burden, including micrometastases, is reduced progressively by chemotherapy according to first-order kinetics, describes well the responses of some transplanted tumours, and has provided a satisfactory explanation for the treatment protocols which have proved so successful in the cure of a substantial proportion of children with acute lymphoblastic leukaemia (ALL) who require long periods of maintenance treatment after complete remission has been achieved.
Such a model does not seem to apply to all types of cancer, and there are malignant diseases where clinically evident disease responds to antiproliferative chemotherapy, yet the same chemotherapy during remission does not affect the rate of recurrence after induction by chemo-therapy or after complete remission by ablation of the primary disease by surgery or radiotherapy.

_Nature of residual disease in AJilL
The most compelling example is adult acute myeloid leukaemia (AML) in which, unlike ALL maintenance, chemotherapy contributes little or nothing to the length of remission. This was first demonstrated (Powles et al., 1.979) as a by-product of a controlled study on the role of immunotherapy, in which maintenance during remission by inmmunotherapy only was compared with a regimen of chemotherapy plus immunotherapy. Neither the length of remission nor survival was extended bar chemotherapy during remission (see Fig.  1). In a U.S. trial (Haskell, 1981)  brought into remission by intensive treatment were randomized to receive either no maintenance treatment or repeated rigorous courses of chemotherapy. The rate of relapse was quite unaffected by intensive maintenance chemotherapy.
Similar data have also come from sequential studies at St Bartholomew's Hospital which show that maintenance chemotherapy did not alter the course of the disease after remission induced by intensive regimens (Lister et al., 1981). The explanation that the residual AML cells have become genetically resistant to the drugs used in maintenance therapy, which in general were those used for induction, does not appear to apply here, since in all 3 studies it was found that patients who had relapsed responded again to the drugs that had been uised initially. Moreover, the frequency of second clinical remissions brought abouit by the same drugs was not v-ery different from the frequency of first remissions (Powles et al., 1979). Operationally, of course, the residual cells are resistant, but the nature of this resistance may lie more in their metabolic or anatomical state thian in the acquisition of specific resistance to the chemotherapeutic drtugs.
The status of residual tumour cells in Hodgkin's disease and breast cancer In Hodgkin's disease also, the length of remission and the frequency of relapse are not significantly influenced by prolonged maintenance treatment. This was shown by Young et al. (1973) in a study comparing mnaintenance chemotherapy with no treatment and in an M.R.C. trial (1 979) in which the rate of relapse was the same in a group receiving intensive maintenance therapy as in a group given a less toxic regimen. In Hodgkin's disease, as in AML, the conventional concept of acquired drug resistance cannot be invoked, because a high proportion of the relapsed patients respond again to essentially the same chemotherapy, which when given during remission failed to extend it.
A case can be made that a similar situation applies to certain categories of breast cancer. The treatmeent regimen, CMF (cyclophosphamide, methotrexate and 5-fluorouracil) administered as an adjuvant, treatment to patients who have been brouight into clinical remission by surgery, has not significantly prolonged remission in postmenopausal women. Results of the Milan study, quoted by Carter (1980), show that in the control arm 51-7% were relapse-free 4 years after surgery, compared with 56-5o in the group who received CMF. Yet, CAIF was chosen as an adjuvant treatment because it had prex,iously been shown highlyr effective in patients with advanced disease, in whom 68o% showed a marked clinical response (Canellos et al., 1974). These findings suggest that clinically evident disease can respond better to antiproliferative chemotherapy than residual disease. This behaviour was shown strikingly in a patient with breast cancer of Dr Trevor Powles (Royal Marsden Hospital) who had a massive local recurrence after surgery, with no metastasis. The recurrent tumour responded completely to chemotherapy, but the patient died of a heart attack 6 months later. At postmortem examination there was no evidence of cancer at the site of the recurrence but there were lung secondaries. At the time of chemotherapy, any lung lesions must have been much smaller than the local recurrence, yet the lung deposits were not, eliminated, whereas the clinically evident recurrence was cured. It is rarely possible to get such decisive postmortem data to illustrate that clinically evident disease can be eradicated by a course of chemotherapy, but that disease which was clinically undetectable at the time of chemotherapy, The nature of the "resistant" state Resistance of microscopic residual cancer to chemotherapy when clinically evident disease responds could arise if the residual cancer cells were mitotically quiescent. In this state they would by definition be refractory to cycle-dependent chemotherapeutic agents. However, thev could also show resistance to cycleindependent agents (e.g. many of t;he alkylating agents and ionizing radiations) because lethal damage sustained by the cells can be restituted if the cells are maintained in a non-dividing but metabolically active state. There are manv instances of this in radiation biology. Thus the fraction of mouse leukaemia cells killed in vitro by X-rays is markedly reduced if the cells are maintained after irradiation for some hours at 34°C, at which temperature they are metabolically active but do not divide (Beer et al., 1963). A good in vivo example of mitotically quiescent but metabolically active cells are the parenchymal cells of the adult liver. If after 50 Gy to an exteriorized liver of rats a partial hepatectomy is performed, there is almost total inhibition of l)ifferentiation is measured as the progressive inciease in the number of cells with Fe receptors (i) anct histoclhemically positive for ehloracetate esterase (A) and non-specific esterase (V). Some populations dlifferentiate towards polymorphs an(1 this is associatedl with acquisition of cilloracetate e(sterase, while otlhers differentiate towardIs macropliages an(l be(ome positive for non-specific esterase (Palil et ol., 1 979o; Forl)es et (o., 1981). mitosis if the interval between irradiation and hepatectomy is less than a week, but if the interval is a month the number of mitoses following partial hepatectomy is as high in the irradiated as in the control liver (Weinbren et al., 1960), though with a higher incidence of chromosome breaks (Albert, 1958).
Different mechanisms could bring about mitotic quiescence in cancer cells present as disseminated micro-disease: (1) the cells, perhaps because of their anatomical situation, are not supplied with essential growth factors, or (2) the cells undergo a process akin to differentiation which is associated with reversible cessation of division. There are several examples (cf. Rudland & Warburton, 1982;Yoda & Fujimura, 1979) where, in response to stimuli such as dimethylsulphoxide (DMSO) or prostaglandins, malignant cells (e.g. mammary carcinomas, neuroblastomas, and erythroid and myeloid leukaemias) differentiate in vitro; yet when the stimulus to differentiation is withdrawn this process is reversed. A reason why the cells in clinically evident disease do not become quiescent, when those in microscopic disease do, could be that the tumour itself releases a diffusible product which either prevents differentiation or produces a growth factor without which the tumour cells do not proliferate. As a result, the tumour cells within small lesions would either differentiate or not divide, because the local concentration of the tumourproduced inhibitor of differentiation or the growth factor is too low. On the other hand, in large lesions, the concentration of the putative diffusible tumour product would be sufficient to prevent mitotic quiescence.

Differentiation in vitro and as xenografts of AML cells taken from the blood of patients
The hypothesis that in some instances the malignant cells in remission are in a different state from clinically evident cancer was derived in part from our studies on differentiation of AML cells in culture and as xenografts, and also from observations made by Clarkson et al. (1977) on the basis of continuous in vivo labelling with 3H thymidine, which indicated that in patients presenting with AML a few leukaemic cells were out of cycle. There has been a considerable amount of published work with established AML lines on induction in vitro of differentiation by agents such as DMSO. We found that without adding colony-stimulating factor AML cells taken directly from the peripheral blood of patients with a blood-cell sorter will proliferate for a few weeks in suspension culture, after which they stop increasing in number, stop synthesizing DNA and eventually the culture dies out (Chapuis et al., 1977) (see Fig. 2). With some populations, there are clear indications of differentiation under these culture conditions to either neutrophils or macrophages, on the basis of morphology, histochemistry and surface markers (Palu' et al., 1979a). Using these tests, however, there are some populations (Forbes et al., 1981) for which no evidence of differentiation can be detected in the in vitro culture, but none the less there are indications with these cells of differentiation in vivo as xenografts.
When AML cells are transplanted s.c. into mice which have been immunosuppressed by adult thymectomy and whole-body irradiation, they grow for 2-3 weeks and then the tumours, which are made up of < 90o of human cells almost invariably regress (Palu et al., ]979b). This regression is not due to an immune rejection, since if mice in which tumours consisting of AMIL cells have regressed are re-inoculated with the same population of AML cells, the latter grow again and give rise to tumours which then agaiin regress. The regression in vivo appears to be associated with differentiation, since the AML cells, as they grow in the mice, progressively acquire Fc receptors and esterases characteristic of neutrophils or macrophages. Regression occurs after the AML cells in the xenograft express markers indicating differentiation (Table I).
Initially (PalW' et al., 1979a, b) these investigations were carried out in considerable detail, using 2 populations which differentiated respectively to neutrophils and macrophages. We have now, however, extended this study to 36 different populations, but find no correlation between prognosis and the behaviour of   (Table II). The patients whose cells did not differentiate in culture did not fare detectably worse than those whose cells did differentiate. However, it must be borne in mind that the cells when transplanted as xenografts regress even when they do not show obvious differentiation in culture. With a population of AML which did not differentiate significantly in the normal culture medium, Forbes et al. (1981) found that the rate of differentiation was accelerated by adding prostaglandin A (Fig. 3). This suggests that differentiation of at least some AML cell populations is capable of being modified by physiological factors.

Dormant mnetastasis
The interpretation I have offered for resistance of some subclinical cancer to antiproliferative drugs is similar to one of the mechanisms advanced for tumour dormancy. There are many clinical instances, particularly in cancer of the breast and melanoma, in which tumours locally removed recur locally or as metastases after a long disease-free interval. Tumour dormancy can be studied in experimental U) Ii U, I-. U, 2 z a systems. Fisher & Fisher (1959) showed that a few tumour cells inoculated intraportally remained dormant within the liver until their growth was stimulated by unknown factors associated with hepatic trauma. With oestrogen-dependent tumours, dormancy extending for very long periods can be detected by transplanting the tumours into noirmal recipients, in which no growth occurs, and then admininistering oestrogen, when macroscopic growth sets in. Noble & Hoover (1975), Eccles & Alexander (1975) and Eccles et al. (1980) found that dormant metastases were common in chemically induced rodent sarcomas and lymphomas which did not metastasize frequently when transplanted to normal syngeneic animals, in which they can be "cured" by surgical removal of the primary implant and its draining node. When such tumours were allowed to grow for a set period before being excised surgically, dormant metastases, seeded from the primary implant before it was excised, were revealed as lung or distant lymph-node metastases after immunosuppression, especially that caused by Cyclosporin A (Eccles et al., 1980), long after surgery (Fig. 4). Similarly, cell sus- Cyclosporin A was administered either before inoculation of the tumour, (luring the 14 (lays of tumour growtlh, or at different times after excision of the tumour transplant. The metastases which arise as a result of immtunosuppression after iremoval of the primary are (lorloorlat rnoetastases (Eccles et al., 1980). Aletastases: L=lutng; LN=Iymph no(le(s).

J56
pensions prepared from clinically diseasefree lungs from animals in which the tumour had been excised gave rise to tumours when injected into immunosuppressed recipients. The cells comprising the dormant metastases are not a genetically different subpopulation with unusual properties, because the cells in the metastases induced by immunosuppression have the same biological properties (e.g. TD50, immunogenicity, growth rate and metastatic potential in the normal nonimmuinosuppressed host) as those in the primary implant.
We have not so far been able to induce the outgrowth of dormant metastases of carcinomas by immunosuppression, but the existence of quiescent tumour deposits is indicated in a number of models bv the late appearance of metastases after surgery The tumour cells in these late metastases, when transplanted, grow rapidly and do not differ from those in the primaryx implant; their delayed appearance indicates a host-induced quiescent state. A link between dormant metastases and the resistance to drugs of residual cancer is indicated bv the fact that the dormant sarcoma and lymphoma cells are obviouslv in an unusual state in which they are invulnerable to immune attack. The hosts from which the tumour had been excised (and which harboured the dormant metastases) rejected a second inoculation (i.m. or s.c.) of the same tumour cells. The cells inoculated into the host, immunized by a growing tumour which was excised, were killed and did not become dormant. We have the anomaly that dormant tumour cells persist in an animal in which the same tumour cells are killed when deliberately injected. A clear demonstration of a state of resistance to T-mediated immunity. The phenomenon that disseminated micro-cancer can be resistant to antiproliferative cytotoxic agents when "bulk disease" is responsive is demonstrated most clearly in AML, but may apply also to other malignant diseases. It is possible that the refractory disseminated cells in these clinical cancers, and the dormant metastases in experimental animals, are present in a mitotically quiescent state that is reversible. For AML there are good grounds for attributing the mitotic arrest to differentiation.
What are the possible strategies for destroying such quiescent cells; i.e. what treatments could prolong remission? An approach which may be particularly applicable when the dormant state is due to hormone dependence would be deliberately to bring the cells into cycle by a hormonal manoeuvre and to follow this with conventional antimitotics. A second means of dealing with such differentiated cells would be agents promoting differentiation. These can cause cancer cells to undergo irreversible differentiation in vitro, but we are very much in the dark on how to proceed along such lines in vivo. r find myself most attracted to the idea of looking, for agents which kill cells which are not dividing, but which yet show a degree of selectivity based on cell type. That such an approach is not total "pie in the sky" comes from findings made in experimental animals and in man that a component of the graft-vs-host (GvH) reaction can have an antileukaemic effect. GvH disease) is conventionally induced by first ablating the marrow, usually by total-body irradiation, and then transplanting genetically dissimilar marrow cells. Immunocvtes derived from the graft attack cells of host phenotype. The nature of the cytotoxic effector mechanism of GvH disease is not known, but it shows a remarkable tissue selectivity. The normal tissues at high risk to the conventional antimitotic chemotherapy are those of high turnover (i.e. marrow and mucosae of the GI tract) but these are unaffected in GvH disease, which produces its symptoms and can kill by damaging the skin at the dermal-epidermal junction, the liver at the level of the bile-duct epithelium and the gut by processes which do not primarily involve the mucosae. That GvH disease also attacks leukaemia has been known in experimental animals for a long time (cf. Okunewick et al., 1981) but it has only very recently been recognized that it may have a role in preventing recurrence of AMIL in man. The Seattle group (Weiden et al., 1981) have found that recurrence of AML can be greatly reduced if patients in remission are given 10 Gy of total-body irradiation followed by marrow graft. This procedure was initiated on the assumption that the irradiation would eradicate residual AML cells left after intensive chemotherapy, and the marrow graft was given to reverse an otherwise lethal ablation of marrow function. The severity of the ensuing GvH disease varies, and depends on the genetic disparity between patient and marrow donor. On analysing their results (Fig. 5) the Seattle group found an impressive inverse correlation between the likelihood of leukaemic recurrence in patients with AML who had received irradiation and marrow grafting and the severity of the GvH disease. The inference I draw is that the cytotoxic effector of GvH disease kills residual leukaemia cells which are refractory to the conventional antimitotic agents used in cancer chemotherapy.
A major difficulty in searching for agents which kill quiescent cancer cells is that their activity will probably only be revealed when tested in an adjuvant setting. Response of bulk disease is unlikely to provide a guide, and one may have to abandon the precept of cancer chemotherapists that only agents that are effective in inducing remission are justified in protocols designed to eradicate residual disease.