|
The staging of cancer
and its treatment are dependent on the presence or absence of
metastases. The specific clinical details and the biology of metastases
vary with the site and extent of target organ involvement and with the
nature of the primary cancer. The picture is complicated by individual
variations in host response and anatomy and by the natural history of
the metastatic process as modified by treatment.
The metastatic cascade
consists of a number of exceptionally complex, overlapping, repetitive
steps. The first phase is concerned with invasion and ultimately
involves the entry of cancer cells into blood, lymph, and cerebrospinal
fluid as well as various body spaces and cavities. The second phase of
dissemination involves the transport of cells that have gained access to
the various disseminative channels and their arrest at target sites.
Most of the arrested cells are killed. The comparatively few viable
cells retained in the target organs develop into micrometastases, which
in themselves are relatively harmless. However, as a result of
cancer-host interactions, including neovascularisation, the
micrometastases may enter a growth phase and develop into metastases. A
major clinical problem is the detection of micrometastases and the
prevention of their development. The final step in the cascade is the
metastasis of the metastases.
A major problem in
determining the mechanisms involved in metastasis is obtaining suitable
animal models. Although the genetics and immunology of mouse tumor
systems are well documented, the short life span and hemodynamics of the
mouse make it in some respects an unsuitable model. In larger animals,
although their life spans and hemodynamics are more relevant to the
human situation, the genetic uncertainties, problems in obtaining
transplantable cancers, and practical difficulties associated with
procuring large animals with spontaneous tumors make it difficult to
perform reproducible experiments. No experimental system provides a
total model for human metastasis, but judicious studies with these
systems provide useful information on parts of this disease process. The
development of noninvasive techniques will, it may be hoped, permit the
ethical study of metastasis in humans.
Cell Detachment from the
Primary Cancer
Detachment is an
essential part of the metastatic process, because by definition, a
metastasis is a cancer that has lost contiguity with the tumor
generating it. The suggestion that the undeniable tendency of cancer
cells to detach from their parent tumors is an inherent part of the
malignant phenotype is a misleading simplification, because it focuses
exclusively on hypothetical stable cancer specific properties of the
cancer cells themselves. Cell detachment is not cancer-specific, and
although normal tissues do not metastasize, it must be emphasized that
detachment is not synonymous with metastasis. The relative ease with
which cancer cells detach from tumors varies not only with the
pathophysiologic status of the malignant cells themselves but also with
the dynamic properties of the whole tumor, which contains cancer and
noncancer cells, both modified by host response.
Factors Affecting
Detachment
Many tumors exhibit
considerable heterogeneity, not only with respect to the cancer cells
that they contain, but also with respect to regions of proliferation and
necrosis, host cell infiltration, fibrosis, encapsulation, and blood
supply. The effects of some of these variables on cell detachment have
been quantitated by various techniques in which cultured cells are
detached from their substrata by known hydrodynamic forces or are shaken
free of small blocks of tissues under carefully standardized conditions.
Growth Rate
A number of experiments
made with cancer and noncancer cells have revealed that the higher the
growth rate the more easily the cells are detached from one another and
from artificial substrates. As cell detachment is only an initial step
in a complex series, it is unlikely that correlations could be
demonstrated between tumor growth rate, detachment, and metastasis over
the whole range of cancers, where so many other variables come into
play. Attempts to correlate growth rate with metastasis reveal many
apparent contradictions, such as rapidly growing tumors that do not
metastasize and slow-growing tumors that do; the problems of occult
primary cancers are well known. However, in a number of palpable human
epitheliomas in which growth rate could be assessed by size change
measured with callipers, Gliicksmann reported a striking correlation
between growth coefficient and the incidence of local lymph node
metastasis.
Necrosis
Necrosis is a common
feature of solid tumors, as a consequence of growth-associated vascular
insufficiency, host-defence reactions, and therapeutic intervention. In
studies on W256 tumors, regions of central necrosis developed in tumors
exceeding 1 cm in diameter. It was shown that viable cancer cells were
more readily detached from each other in the juxtanecrotic regions than
in the more peripheral regions of tumors. This effect was mimicked by
exposing samples from the tumor peripheries to saline extracts of the
necrotic core. When W256 tumors grow in the rat liver, the parenchymal
cells close to the tumor edges may be more easily separated from each
other than those located 0.5 or 1 cm from the tumor periphery. The
evidence suggested that the necrotic material acted as a pool of free
lysosomal enzymes and other material that acted on the various cells
present in the tumors and facilitated their detachment by direct or
indirect mechanisms.
Whether the demonstrable
necrosis-associated facilitation of detachment actually promotes
metastasis depends on whether the released cancer cells are free to
disseminate. On the one hand, the facilitated detachment of cells in the
tissues surrounding a tumor is reasonably expected to promote its
invasion, particularly as necrotic extracts promote the active movements
of some types of cancer cells. On the other hand, if necrosis results
from vascular insufficiency, immediate dissemination of detached cancer
cells may not occur. Depending on the accessibility of cancer cell'
'escape routes," there is the possibility that by causing necrosis in
the absence of a 100 percent kill, local therapy may actually promote
dissemination. In this connection it is of interest that exposure of
cancer cells to a variety of antimetabolites promotes their detachment
in vitro.
Enzymes
Many pathologic
processes associated with inflammation, immune response, therapy, etc.
that result from the release of free radicals are ultimately expressed
in enzyme release; in events related to metastasis, the enzymes may be
derived from the cancer cells themselves, from vascular endothelial
cells, from other cells of the reticuloendothelial system, and from
other types of cell. It has been known for many years that viable cells
may be liberated from a variety of tissues by the action of enzymes
combined with the application of distractive forces generated by muscle
movements, surgical trauma, etc. In essence, free enzymes may facilitate
cell detachment by lysis of intercellular material. In addition, the
release of endogenous enzymes by both cancer cells and nonmalignant
"bystanders" also facilitates detachment. The Iysosomes constitute an
obvious and well-documented nonexclusive source of such enzymes, and a
large number of factors activating Iysosomes promote detachment, whereas
a number of agents stabilizing Iysosomes inhibit the process, as do
enzyme inhibitors, including the tissue inhibitors of metalloproteinases
(TlMPs).
Stress
Although the effects of
surgical manipulation on cancer cell release are well documented, it is
not generally appreciated that the appearance of malignant cells in the
bloodstream and the release of cells from tissues can be promoted by
stress and anaesthesia. These disquieting phenomena are worthy of
further study.
Cell Movement and
Invasion
Cell movement may be
active or passive. In invasive processes, cancer cells may crawl through
tissues and breach basement membranes during intravasation and
extravasation. Alternatively. invasion may result from the expansive
forces generated by growing tumors (vis a tergo), and movement will
occur along the paths of least mechanical resistance. Expansion by
growth can result in arrested tumor emboli bursting out of blood
vessels. Although active movement of cancer cells was recognized by
Virchow in 1863, for various technical reasons the relative importance
of crawling and growth-associated expansive movements were never
quantitatively assessed until recently. A technique has been developed
in which, from statistical analyses of cancer cell density counts made
on tumor sections, the diffusive density patterns associated with active
cell movement may be discriminated from the more abrupt patterns
associated with growth. Tests of the technique made on sections of
malignant melanomas in human skin confirm the validity of the technique
and indicate that actively moving melanoma cells invade the dermis to a
depth of up to 500µm in advance of the main tumor body. Within this
zone. it appears that the cancer cells stop migrating and proliferate.
The whole process is then repeated. This progression of invasion in
500 µm steps suggests reappraisal of
wide-margin excision protocols for cutaneous melanomas.
For active locomotion to
occur, cancer cells must first make contact with and adhere to tissues
through which they move; locomotor energy generated by the cell acts on
these adhesive regions, and finally, localized detachment must occur
to permit translatory movements. Therefore, agents promoting initial
cell adhesion and cell detachment and stimulating the actin-containing
contractile microfilaments in cancer cells are expected to promote
active movements and vice versa. Enzymes are expected to have
paradoxical effects, because they may inhibit the formation of focal
adhesions or destroy them and may also promote detachment. The action of
any agent will probably depend on its effects on a rather delicate
balance of the three basic components of active movement.
In invasion, the
degradation of tissue matrix and basement membranes is effected by four
types of proteases, namely (1) metalloproteases (e.g., collagenases);
(2) cysteine proteinases (e.g., cathepsins B,D,L); (3) serine protease
plasmin; and (4) plasminogen activators. In general, these enzymes are
produced by leukocytes, fibroblasts, endothelial cells, and cancer
cells.
The urokinase-type
plasminogen activator converts plasminogen into plasmin, which brings
about pericellular proteolysis. Although initially secreted as inactive proenzyme (pro-uPA), it is activated by binding to a specific receptor (uPA-R)
at the surfaces of cancer and noncancer cells. In addition to direct
proteolysis of tissue matrix elements (e.g., fibrin, laminin,
fibronectin, proteoglycans), plasmin converts procollagenase type IV
into active collagenase IV, which degrades collagen IV, the major
structural component of basement membranes. Degraded collagen
stimulates cancer cell migration. Naturally occurring plasminogen
activator inhibitors (PAIs) fall into two groups; PAI-1 and PAI-2 belong
to the serpin superfamily, and PAI-3 is the protease nexin. Preliminary
results indicate that in breast cancer, expression of uPA and PAl-1
antigens are independent prognostic factors for relapse-free survival,
and multivariate analysis reveals that breast cancers exhibiting high uPA and high PAl-1 levels are associated with high risk for relapse
regardless of lymph node involvement. This interesting observation will
doubtless be tested further. Receptormediated internalization of the
uPA/PAI-1 complex may also trigger cell proliferation, illustrating
coupling between the metastatic process per se, and cancer cell
proliferation.
In addition to these
effects, enzymes, by acting on the noncancerous tissues surrounding a
tumor, may facilitate movement by expansion. In contrast, by destroying
the substrate through which a cell could move and thus preventing
adhesion, enzyme-related tissue lysis could also prevent or inhibit
active movements of cancer cells and make inactive movement by tumor
expansion (vis a tergo) more important in this phase of metastasis.
Circulating Cancer Cells
Cancer cells may gain
direct access to the venous circulation, or alternatively may gain
access first to the lymphatic system and then to the bloodstream. There
are so many communications between the lymphatic and venous systems that
except in the earliest stages of metastatic cancer, it seems
unrealistic to consider metastasis as being limited to one system or the
other. This view should not be confused with the question of the
desirability of lymph node removal in the treatment of early cancers,
because this has the double function of permitting accurate TNM staging
and the eradication of potential generalizing sites.
The observation that in
the early phases of clinical disease, sarcomas tend to give rise to
haematogenous metastases whereas carcinomas generate lymphogenous
metastases, has given rise to the commonly accepted concept that
carcinomas disseminate via the lymphatics and sarcomas via the blood.
The evidence is that lymph node metastases are approximately three times
as common in carcinomas as in sarcomas, which does not make lymphogenous
metastasis of sarcomas rare, and the postregional lymph node pattern of
metastasis from carcinomas is frequently haematogenous. The initial
disseminative route is to some extent influenced by the site of cell
detachment from the primary lesion. If detachment occurs prior to
contact between the tumor and blood channels, as in the case of early
cutaneous melanomas, for example, where cancer cells actively locomote
through the dermis, they will have the opportunity to gain early access
to the lymphatic system. If contact with blood channels occurs prior to
detachment, venous dissemination will be the preferred initial route. If
parts of an invasive tumor project into the lumen of a blood vessel,
cells can be detached by a combination of factors described above. An
alternative mode of entry takes place in some sarcomas, where vascular
clefts lined by cancer cells are present and shedding is directly into
the bloodstream.
Comparisons between
sarcomas and carcinomas, disregarding disease status, are incorrect
because carcinomas metastasize via both the blood and lymphatics, and
lymph node involvement may be detected at an earlier stage than small
haematogenous metastases in other organs. Alternatively, carcinoma cells
may initially disseminate synchronously via both routes, but the
posthaematogenous delivery phase in the microvascular beds of organs may
take place in a more hostile environment for carcinoma cells than lymph
nodes; the converse may be true for sarcomas. Therefore, differential
metastatic patterns between the two classes of tumors depend not only on
factors governing entry into, and delivery via, the different routes,
but also on differential interactions in the post delivery phase of
disease.
Any quantitative
assessment of haematogenous metastasis requires knowledge of the numbers
of cancer cells entering the bloodstream. Much of the earlier literature
on the occurrence of circulating cancer cells in people is suspect
because in a number of papers, megakaryocytes and degenerate cells were
erroneously described as cancer cells. However, reports of positive
identification of cancer cells in the venous effluents of many tumors
clearly indicate that large numbers enter the bloodstream, although
quantitation is sparse.
The rates of entry of
cancer cells into the bloodstream from primary cancers of known size was
determined in patients with renal carcinomas, by collecting blood
directly from the renal vein just prior to nephrectomy. Cells were
released at rates of millions per day, probably for some time before nephrectomy. Two patients in this study were of particular interest: the
first had a primary renal tumor, 10 cm in diameter, that was releasing
cancer cells at a rate of 5 x 109 per 24 h; and the second had a
6-cm-diameter tumor that released cells at rate of 2.3 x 108 per 24 h.
Neither had detectable metastases after 66 and 31 months, respectively,
indicating a level of metastatic inefficiency of substantially less
than 109; that is, more than 109 renal carcinoma cells have to be
released into the renal vein to generate one haematogenous metastasis in
the lungs.
On the reasonable
assumption that millions of viable cancer cells are liberated into the
blood of patients with cancer, then because overt metastases usually
occur with frequencies that are orders of magnitude less, one is forced
to conclude that in terms of the cancer cells involved, metastasis is a
remarkably inefficient process. The mechanisms and implications of
"metastatic inefficiency," which is the driving force underlying metachronous (sequential) metastatic pattern formation, is reviewed
elsewhere. It is paradoxical that an inefficient process should
result in the deaths of so many patients! However, even an inefficient
process, will succeed if repeated often enough.
By all accounts, the
circulation is a hostile environment for cancer cells; the vast majority
perish around the time of arrest in the microcirculation. Part of the
circulatory trauma appears to be due to the mechanical deformation
imposed on cancer cells in passing through the microcirculation, where
they undergo shape transformation from spheres to cylindroids, resulting
in stretching and lethal rupture of their external membranes.
Arrest of Circulating
Cancer Cells
Without arrest,
metastasis cannot occur. Cancer cells are arrested in the
microvasculature; single cancer cells in capillaries, and clumps of
cells in larger vessels. From an analytic approach, it is important to
discriminate between mechanical trapping of cells, which is in a sense
independent of local chemistry, and adhesion, which involves chemical
bonding between .cancer cells and luminal endothelium and/or subendothelium. It is useful to regard trapping, which brings the
arrested cancer cells into close proximity with vessel walls, as the
initial event followed by bond formation involving so-called cell
adhesion molecules (CAMs).
The surfaces of all
types of human cells carry a net negative electric charge. This results
in an average electrostatic repulsion between cancer cells and vascular
endothelial cells, which tends to prevent their contact and adhesion.
The charged groups on a number of cells tend to be arranged in
clusters, and in some cells contact is made via fine probes or
macromolecules and those surface regions between the clusters, where the
charge density is low.
Another factor
inhibiting contact relates to fluid displacement from between the
vascular endothelium and approaching cancer cells. When a cancer cell
moves in the blood, hydrodynamic forces are generated that cause
pressure changes and plasma movements close to its surface. This
hydrodynamic field is significantly perturbed when the distance
separating the cancer cell from the vascular endothelium is less than
the cancer cell radius, leading to retardation in contact and hence
cell arrest. If the cancer cells are deformed in the contact-making
process, so that the cancer cell is flattened to match the contours of
the opposing endothelium, the average distance between the two surfaces
will be decreased and contact and arrest will take longer.
Another interaction of
cancer cell emboli and the vascular endothelium involves embolic size.
Single cancer cells are temporarily arrested at the level of
capillaries and postcapillary venules, whereas multicellular emboli tend
to be shunted through larger vessels where they are either eventually
arrested or die in the circulation. In arrested multicellular emboli,
the outer cells tend to protect the innermost cells. These and possibly
other factors lead to a greater metastatic efficiency of multicellular
than of unicellular emboli.
The observation that
cancer cells are arrested immediately on entering capillaries, in spite
of undoubted electrostatic and viscosity barriers, tends to favor a
mechanical trapping mechanism due to both the disparity between the
cancer cell and capillary diameters and the irregularities in their
surfaces-a Velcro-like phenomenon.
Following cancer cell
arrest, the vascular endothelium retracts, exposing the subendothelial
basement membrane, which is rich in adhesive molecules associated with
the extracellular matrix (ECM). These molecules bind to receptors on the
cancer cell surface; one major class of receptors is the integrin
superfamily of heterodimers, which bind various ECM proteins, including
fibronectin [and other proteins containing arginine-aspartate-glycine (RGD)
or leucine-aspartate-valine (LDV) sequences], laminin, collagen IV, and
fibrinogen. Other CAMs include the selectins, cadherins, and
immunoglobulin supergene family (which includes carcinoembryonic
antigen, CEA).
Although these CAMs play
an important role in the arrest of leukocytes in the microvasculature,
this may not be their primary role in metastasis. Many if not all of
these molecules have transmembrane domains, and their primary role may
be in signal transduction, utilizing G-proteins, in which signals from
the ECM stimulate the reproduction of trapped cancer cells.
Thrombosis and Cancer
Cell Arrest
Fibrin and platelet
deposition is often seen associated with arrested cancer cells,
particularly where there is a defect in the vascular endothelium
revealing the underlying basement membrane. Studies by the author and
his associates in mice revealed no changes in the arrest patterns of
several types of intravenously injected cancer cells when the platelet
release reaction was depressed by the administration of aspirin. In
addition, therapeutic doses of heparin and warfarin, which affect
different levels of the coagulation cascade, were also without effect
on cancer cell arrest patterns in both normal mice and tumor-bearing
animals with the coagulopathies typically associated with cancer. These
experiments suggest that coagulation factors do not playa key role in
cancer cell arrest.
From the foregoing, one
would expect that a fibrin cocoon provides temporary protection to
arrested cancer cell emboli. It therefore seems paradoxical that cancer
cells exhibit differing degrees of plasminogen activation; the resulting fibrinolysis would be expected to reduce their metastatic potential.
In view of the
association of platelets and fibrin with the metastatic process, there
have been many attempts to utilize anticoagulants in antimetastatic
therapy. Although the reports of these attempts are both conflicting
and confusing, it appears that under certain conditions anticoagulants
can reduce the incidence of metastases. However, this anti
metastatic effect may not be due to their anticoagulant activity. For
example, aspirin interferes with prostaglandin biosynthesis, and warfarin can inhibit cell motility. Another complicating factor is that
some agents, for example aspirin, seem to have differential activity on
platelets and on vascular endothelial cells.
A physiologic role of
the vascular endothelium is to maintain vascular integrity by means of
self-purging mechanisms. In situations where thrombi involving
platelets are concerned, there is a delicate intravascular balance
between the activities of thromboxane Az (TXAz), which aggregates
platelets as part of host defence, and prostacyclin (PGIz), which is
released by the vascular endothelium and which inhibits TXA2-induced
platelet aggregation. The PGIrTXA2 balance is disturbed in the presence of some cancer cells in favor of TXAr-induced platelet
aggregation, which is thought to promote metastasis. On this basis, Honn
and colleagues have administered agents either promoting PGI2 activity
or synthesis or inhibiting TXA2 synthesis to mice injected with cancer
cells or bearing tumors and have demonstrated antimetastatic activity.
Regardless of whether this form of therapy proves applicable to humans,
the observations serve to illustrate the importance in metastasis of
interactions involving the vascular endothelium.
The Role of Immune and
Inflammatory Responses
The retention of cancer
cells by the vascular endothelium may be modified by the immune response
associated with tumor bearing. It has been shown that following
sensitization, retention of injected cancer cells in the lungs is
increased in some tumor-host situations; in others it is decreased, and
in others not changed at all. Retention may also be modified by the
inflammatory response. For example, in mice, inflammatory reactions in
the lungs following bleomycin therapy increase pulmonary retention of
circulating cancer cells.
The reticuloendothelial
system (RES), which includes the mobile and sessile mononuclear
phagocytes, leukocytes, and vascular endothelial cells, also
participates in the loss of arrested cancer cells. This is indicated by
studies in which stimulation of the RES by glucan, BCG, Corynebacterium
parvum, endotoxin, and zymosan was associated with increased clearance,
whereas inhibition of the RES by silica and trypan blue was associated
with reduced clearance. It seems on balance that although the
inflammatory response kills tumor cells, the tissue degradation
resulting from it facilitates the invasion of the survivors.
Further Development of
Metastases
Cells retained in the
vasculature of an organ extravasate at an early stage in metastasis
development. Extravasation can occur if cancer cells actively migrate
through vessel walls, or alternatively, growing emboli can burst out.
Both these processes would presumably be aided by enzyme-mediated lysis
of basement membranes and the weakening associated with increases in
vascular permeability. In addition, the interaction of arrested cancer
cells with microvessel endothelium and basement membrane, mediated by
adhesion molecules and signal transduction, probably results in
expansive intravascular growth of the cancer emboli, thereby promoting a
"bursting-out" mechanism of extravasation.
Although the developing
metastasis is less than approximately 2 mm in diameter, it can obtain
its nutrition by diffusion processes. This provides a functional
definition of a micrometastasis, because the increase in size of a micrometastasis to a metastasis requires
vascularisation. This is achieved by
diffusion of angiogenic factors from cancer cells and other involved
tissues and cells to local host capillaries; the endothelial cells are
stimulated into mitosis and grow toward and invade the tumor, ensuring
its nutrition and growth. The role of neovascularisation is emphasized
by experiments showing that agents such as cartilage inhibit tumor
growth by inhibition of angiogenesis. Angiogenic factors are not specific for cancer cells but may also be
produced by inflammatory cells and fibroblasts. It seems likely that an
important component in the neovascularisation cascade is prostaglandin E1. It is therefore of interest that prostaglandin E (PGE) production is
one of a number of factors involved in the failure of the immune system
to eliminate tumor growth; although immune stimulation results in
enhanced PGE production, the prostaglandins produced inhibit function by
a negative feedback mechanism.
Failure of
micrometastases to grow is associated with the so-called dormant state,
which is well recognized by clinicians. In this condition, patients with
removed primary cancers survive for long periods with no overt
metastases. In apparent response to apparently trivial stimuli they
then develop a metastatic "explosion." When recurrence or new cancers
are ruled out, the dormant state may represent true dormancy, in which
micrometastases behave in an inert manner, with their constituent cancer
cells apparently in the nondividing (Go) state. Alternatively, the
situation may represent a pseudodormant state in which cancer cell
multiplication is balanced by loss. Either way, if by definition
micrometastases are not vascularised, it is not difficult to understand
their unresponsiveness to systemic chemotherapy. The various complex
interactions of cancer cells and the microvasculature are pivotal to the
process of metastasis.
Metastatic Patterns
A final step in the
development of the metastatic process is the metastasis of metastases.
It appears that most human cancers metastasize to so-called
generalizing sites, determined by anatomical considerations, the
commonest being lymph nodes, lung, and liver. From these secondary
sites, tertiary metastases occur. The consequent metastatic patterns
depend partially on "mechanical" factors, including target organ blood
flow, and partially on as yet undefined "seed and soil" factors. The
elucidation of the mechanisms of pattern formation in humans is
rendered exceptionally difficult by the perturbations induced by
therapy.
Anatomic considerations
dictate disseminative routes for cancer cells, and constitute one
important factor in determining metastatic patterns. Thus, the initial
sites of metastasis are for the most part determined by venous drainage;
gastrointestinal tumors metastasize first to the liver via the portal
vein, cancers draining into systemic veins metastasize first to the
lungs via the venae cavae, and pelvic tumors disseminate in part via the
paravertebral venous plexus.
Taking colorectal
carcinoma as an example, the clinically important question is whether
or not a proportion of the cancer cells entering the liver via the
portal vein, leave via the hepatic vein to seed the lungs and a
proportion of those pass through the lungs to seed other organs via the
arterial route On the one hand, if this synchronous seeding sequence
occurs, then providing that enough cancer cells pass through the liver
to generate lung metastases, and enough pass through the lungs to
generate arterial metastases, the presence of detectable liver
metastases would invariably indicate the presence of metastases or
micrometastases in the lungs and other organs. This would rule out
curative local therapy.
On the other hand,
metastatic inefficiency favours metachronous seeding patterns. That is,
it predicts that the majority of cancer cells entering the liver are
arrested and killed there, and that the development of liver metastases
is a relatively slow process. Only small, nontumorigenic numbers of
cancer cells therefore reach the lungs directly from the primary cancer;
this direct seeding stops when the primary cancer is resected. It is
therefore expected that only cancer cells released from the secondary
liver metastases will successfully seed the lungs and that the majority
of these will be arrested and killed there. Thus, arterial metastases
are not expected to be seeded directly from liver metastases, but
seeding is predicted from cancer cells released from tertiary lung
metastases.
Evidence in favor of
metachronous seeding comes from the analysis of autopsy reports, where
groups were identified with liver metastases and none elsewhere; with
liver and lung metastases and none elsewhere, and liver, lung, and
arterial metastases.
Conversely, lung
metastases seldom occurred in the absence of liver metastases, and
arterial metastases seldom occurred in the absence of lung metastases.
Recognition of this sequential pattern of haematogenous metastasis could
advantageously be incorporated in staging systems.
Metachronous seeding can
be accounted for in terms of anatomy and general metastatic
inefficiency. However, specificity with respect to arterial metastatic
patterns requires cellmicroenvironmental interactions in the
postdelivery phase of metastasis. As predicted by the "seed and soil"
hypothesis of Paget in 1889, these interactions can either promote or
inhibit metastatic development.
Analysis of human
metastatic patterns has not been overly successful in the past, because
of the difficulty in disentangling cancer cell delivery from the
subsequent interactions at the target site. A method of discriminating
between the delivery and postdelivery events has now been described:
In essence, cancer cell delivery in the case of arterial metastases is
considered to be proportional to target organ blood flow; this
assumption is based on actual measurement made on laboratory animals.
Values of human organ blood flow are obtained from recent measurements
reported in the physiologic literature, and the incidence of arterial
metastases in these sites is obtained from large autopsy series on
patients dying with metastatic cancer. The ratio of blood flow (ml/min)
to incidence (%) is termed the Metastatic Efficiency Index (MEI), which
falls into three groups. Most sites are within the range of 0.01 to
0.09; values higher than 0.09 indicate postdelivery interactions
favourable to metastasis development, whereas values of less than 0.01
indicate unfavourable interactions.
The brain as a whole
falls into the middle range for most primary cancers, but into the "unfavourable"
range «0.01) for osteosarcoma and ovarian, gastric, and bladder
carcinomas. Unfortunately, it was not possible to match the incidence
of metastases in different regions of the brain with differential blood
flow on the bases of presently available data. These values for the
whole brain are in contrast to the generally unfavourable range for
skeletal muscle, and the highly favourable range for the adrenals. The
posterior choroid of the eye is by far the most favourable site for
metastasis development per unit of cancer cells delivered, from each of
three types of primary cancers (breast 8.5 percent; colon, 4.2 percent;
bronchus, 6.7 percent) although the actual incidence of intraocular
metastasis, which partially reflects cancer cell delivery, is low. It is
hoped that this type of numerical documentation will lead to studies of
specific site-associated mechanisms of favourable and unfavourable
interactions.
In vitro experiments, in
which preferential adhesion of cancer cells to tissue sections has been
reported, reflect neither differential arrest in the microvasculature
under hemodynamic conditions, nor the differential growth of
arrested/adherent cancer emboli. In vivo experiments have also been
reported, in which cancer cells growing in colonies at different
anatomic sites following intravenous injection, are "selected" by
repetitive injections and harvesting procedures, for preferential
growth in different sites following injection. Any interpretation of
experiments of this type must take into account site-associated changes
occurring in cancer cells after delivery, possibly indicated through
interactions with the extracellular matrix. In addition, autopsy data
fail to reveal preferential metastasis to one of paired organs from
primary cancers in the other.
Molecular Biology of
Metastasis: A Brief Overview
Some of the major recent
advances in oncology have generally been in the area of molecular
biology. and have focused on the role of oncogene and suppressor gene
products in the poorly restricted growth of cancer cells. The proto-oncogenes
encode proteins which transduce extracellular signals from the cell
membrane to the nucleus which stimulate cell proliferation. Thus,
mutations activating the oncogenic potential of cell proto-oncogenes
can lead to loss of control of proliferation seen in normal cells.
Mutations of this type, and those inactivating genes controlling growth
suppression account for the development of cancer. although many of the
detailed mechanisms presently require clarification. It is therefore not
surprising that a great deal of recent research in metastasis has been
directed at determining the molecular biology of metastasis.
A good example of the
potential value of such research is provided by carcinoma of the
breast, in patients without lymph node involvement. Approximately 15
percent of these patients subsequently develop metastases that warrant
treatment by systemic chemotherapy. However, this 15 percent patient
subpopulation cannot be identified on the basis of routine histologic
examination, and the question arises whether all patients in this
apparent NoMo category should be subjected to treatment. in order to
"trap" the 15 percent minority. The problem could be resolved if probes
were available to predict the biologic behaviour of individual tumors,
over and above the group statistical evaluations obtained from standard
histologic examination. A number of promising leads are currently under
investigation, particularly in relation to carcinomas of the breast and
colon.
From a clinical
viewpoint, the salient features of metastasis concern the growth of
primary cancers initially, and the subsequent growth of surviving
cancer emboli into metastases. The potential for growth can be assessed
by a number of histologic and immunohistologic procedures including
mitotic index, BUdRincorporation, Ki-67, and proliferating cell nuclear
antigen (PCNA) expression. Proliferation potential can also be assessed
with molecular biologic probes capable of detecting oncogene (e.g.,
C-erb-B2) and suppressor gene (e.g., NM23 and RB) products; expression,
loss of expression, and mutations in these' gene products have been
associated with metastatic status. However, at present there is no
universal indicator for metastasis prediction that is applicable to all
types of cancers, or even for breast and colonic cancers. The complexity
of the metastatic process virtually precludes the existence of a
universal metastatic gene, and from a mechanistic viewpoint it is
important to discriminate between causal relationships and associations
between up- or downregulation of gene products and metastasis. An
example of multifunctional gene products has been hypothesized in
connection with invasion, which is regarded as due to an imbalance
between the activation of two sets of genes, an invasion promoter and an
invasion suppressor such as TIMP. A cell adhesion molecule (E-cadherin,
L-CAM, Uvomorulin) has been implicated as an invasion suppressor gene
product.
Foulds' concept of tumor
progression, originally described at the level of individual cells,
involves irreversible steps leading from premalignant states to cancer.
Confusion has been introduced into terminology by the clinical aphorism
that "cancer goes from bad to worse," which is a reflection of the
progressive deterioration in the patient's condition as the disease
progresses, over time, as distinct from specific changes in cancer cells
themselves.
Credit is due to Fidler
for suggesting that subpopulations of cancer cells evolve that express a
so-called metastatic phenotype, so that metastasis would be a nonrandom,
selective process. In contrast, Weiss noted that even among cell lines
that were cloned on the basis of metastasis-related properties the vast
majority were destroyed during or after delivery, and that at the cancer
cell level, metastasis is essentially a random process. Weiss favours
the concept of a "transient metastatic compartment," in which at anyone
time, the whole heterogeneous population of cancer cells within a tumor
each possess different metastatic potential depending on metabolic and
cycle status, host interactions, and topography. In other words,
virtually all of the cancer cells in a tumor are potentially
metastatic, but the probability of each cancer cell actually generating
a metastasis varies from time to time, and is very low.
From a practical
viewpoint, regardless of underlying random or nonrandom elements, the
important conclusion is that at any time, some cancer cells in a
surgical specimen of a potentially metastasizing tumor may express
specific metastasis-related properties that, if identifiable and
quantifiable, may have predictive prognostic value. At present, it
seems likely that attempts to assess and predict the biologic behaviour
of cancer will not depend on the use of a single metastasis-specific
probe, but may well depend on the use of panels of probes, markers, and
standard histologic assessments, not only of the cancer cells, but also
of tumor/host interactions in, adjacent to, and distant from the tumor
itself. The information obtained must be useful to the individual
patient as distinct from cohorts of patients.
|
