The Role of
Cytotoxicity in Allograft Rejection In Vivo
Delayed hypersensitivity, tissue
transplantation and the discovery of CTL
Initiation
of allograft rejection in vivo
General requirements for primary CTL
activation
The role of dendritic cells in triggering
vascularized allograft rejection
Initiation of allograft responses in
free-cell suspensions
Mechanisms of allograft rejection in
vivo
The histopathology of
allograft rejection
The role of vascular damage in allograft
rejection
DTH reactions and allograft rejection
The role of CD8-mediated cytotoxicity in
allograft rejection
Delayed
hypersensitivity, tissue transplantation, and the discovery of CTL
The discovery of cell-mediated immunity, and
ultimately of cytotoxic T lymphocytes, emerged from attempts to understand the
immunological basis and mechanism of two phenomena discovered early in the
history of immunology: delayed hypersensitivity reactions, and allograft
rejection. Both of these
phenomena were presumed by many
investigators to be immunological in nature, showing the properties of
specificity and memory, but for many years were misunderstood and open to other
interpretations.
Hypersensitivity
reactions were an unwelcome conundrum in the beginning days of immunology.
Early studies of immunity, at the end of the nineteenth century, were
understandably focused on its protective aspects, the ability of the immune
system to quickly and efficiently clear infections by potentially
disease-causing microorganisms. But shortly after the turn of the century,
reports from the French scientists Paul Portier and Charles Richet suggested
this might not always be the case. In a few cases, which were shown to be
entirely reproducible, initial exposure to an antigen resulted in a state in
which subsequent exposure to the same antigen resulted not in protection, but
in morbidity or even death. For
example, when a dog was given a sub-lethal but symptom-inducing injection of a
particular bacterial toxin from which it fully recovered, and then several
weeks later was given a second injection of the same toxin, the dog immediately
went into shock and was dead within the hour (Portier and Richet, 1902). It was
previous exposure to the toxin, in a regimen that in most other instances
conferred immunity, that appeared to have created the problem. As can be
imagined, the proposal that the immune system could also cause disease was not greeted with enthusiasm.
Nevertheless,
numerous other examples of these negative reactions were uncovered, and came to
be known as anaphylactic or hypersensitivity reactions. Moreover, a close study
of such reactions in the ensuing years led to the realization that there are
two different types of hypersensitivity, distinguished most notably by the
kinetics of development of the response. Immediate hypersensitivity (IH)
reactions develop within minutes of re-exposure to the provoking antigen, and
reach a peak within a few hours at most. Portier and Richetﾕs dog had clearly
experienced an IH reaction. DTH reactions, on the other hand, may not be
apparent for 12-24 hours, and may require two to three days to reach maximum
intensity.
IH
reactions, which came to include such readily recognizable maladies as allergy,
and potentially lethal secondary reactions to bee stings or drugs such as
penicillin, although not understood at first, seemed to fall within the
developing paradigms of immunology generally. Initial exposure to a provoking
antigen resulted in the production of serum substances (in humans, these turned
out to be antibodies of the IgE subclass), which could be passively transferred
from a hypersensitive to a normal individual, rendering the recipient
selectively hypersensitive to the provoking antigen immediately after transfer.
A similar phenomenon had been repeatedly observed in the passive transfer of
positive (protective) immunity with serum, and was considered a hallmark of
immune responses generally.
DTH reactions were found to underlie such well-known
phenomena as contact sensitivity, certain fungal infections and the tuberculin
reaction. The latter had been described already in 1891 (Koch, 1891) with the
finding that tubercular guinea pigs developed inflammatory skin lesions when
injected with substances obtained from cultures of M. tuberculosis. Animals made hypersensitive to a particular antigen
were not hypersensitive in general but only to the provoking antigen. The
problem posed by DTH reactions was that, although displaying the same
specificity and memory properties as IH reactions, they could not be
transferred with immune serum. Since serum antibodies were the only known
immune mechanism in the first half of the twentieth century, many remained
skeptical that DTH reactions were immunological in nature.
This
puzzle was not solved until the 1940s, through experiments by Karl Landsteiner
and Merrill Chase, who showed that both delayed cutaneous hypersensitivity to
chemicals and the tuberculin reaction in guinea pigs could be transferred
between animals using peritoneal lymphocytes, but not serum (Landsteiner and
Chase, 1942; Chase, 1945). These were the first convincing reports that cells,
as well as antibodies, could mediate a presumed immune phenomenon, but because
of the ambiguous nature of DTH reactions it would be a number of years before
the concept of cell-mediated immunity was rationalized and accepted by the
immunology community at large.
Allograft rejection reactions posed a
similar dilemma. From its inception, early in the twentieth century, the study
of rejection of allogeneic tumors or normal tissue transplants in animals
showed that this process displayed many of the characteristics of immune
reactions, such as memory and specificity. For example, an inbred mouse that
had previously rejected a tumor or skin from an allogeneic inbred mouse strain
would reject a second graft from the same strain in a greatly accelerated
manner. This accelerated ﾒsecondary responseﾓ phenomenon was well know to
immunologists, and was typical of virtually every immune response that had been
studied. In keeping with what was known about immune reactions generally, the
secondary response was highly specific. The mouse just described, if
secondarily transplanted with a tumor from an allogeneic strain unrelated to
the first, rejected that graft with the slower kinetics typical of a primary
reaction in an antigenically naive mouse. Gorer and Medawar would show some
years later that an exchange of tumors between allogeneic animals was no
different than the exchange of normal tissues between the same animals, an
observation that set the field of tumor immunology back for several decades. We
explore that story in Chapter 10.
So
allograft rejection looked very much like an immune phenomenon. But as with DTH
reactions, it was not possible to show that specific immunity to a tumor or
skin allografts could be transferred from an immune animal to a non-immune
animal with serum or any fraction thereof. The ability to carry out such
ﾒpassive transferﾓ of protective immunity was one of the founding doctrines of
immunology, and the failure to demonstrate it in tissue rejection led many to
challenge the notion that rejection was immunological in nature. N. A.
Mitchison unraveled this dilemma by showing that graft immunity, like DTH sensitivity,
could be transferred between mice with immune lymphocytes rather than immune
serum (Mitchison, 1953).
General requirements for
primary CTL activation.
In Chapter 2 we discussed some of the requirements for
activation of CTL in virgin, resting CD8 T cells. The first step (signal 1)
involves interaction of the CD8 T cell via its antigen receptor complex with
cognate antigen. In the case of an allograft reaction, cognate antigen consists
of a peptide-modified allogeneic class I molecule. Complete development of
signal 1 requires the participation of an array of costimulatory interactions
mediated through the immunological synapse. The result of signal 1 is up-regulation
by the resting CD8 cell of receptors for lymphokines needed to drive it to the
fully mature, cytolytic state. These lymphokines (signal 2) are supplied
largely by CD4 cells activated at the same time by class II MHC antigens of the
graft.
There is evidence that the requirements for activation of CD8 cells by
cognate antigen are considerably less rigorous than for CD4 cells. For example,
the number of CD8 TCR-MHC interactions required for activation is much less
than for CD4 cells, approaching the level of 1-10 TCR contacts per CD8 cell
(Brower et al., 1994; Kageyama et al., 1995; Sykulev et al., 1996). This may be
because the CD8 molecule itself, a crucial costimulatory molecule, binds to
class I MHC with a much greater affinity than that displayed in CD4-class II
interactions (Garcia et al., 1996; Gao et al., 1997; Delon et al., 1998; Kern
et al., 1998). CD8 cells also require a shorter contact time with antigen to
achieve full activation than do CD4 cells (Mercado et al., 2000; Kaech and
Ahmed, 2001; van Stipdonk et al., 2001), and yet in some situations they
proliferate more vigorously (Foulds et al., 2002) and longer (Homann et al.,
2001) than CD4 cells.
The CD28-B7 interaction so crucial in CD4 T-cell activation is
apparently also less critical in activation of CD8 T cells, since CD28-knockout
mice can still generate CTL, mount antiviral responses and reject skin grafts
(Shahinian et al., 1993; Kawai et al., 1996), although the rejection of heart
allografts may be impaired (Turka et al., 1992; Baliga et al., 1994). Even in
cases where interference with CD28/B7 engagement shows an effect on development
of allograft rejection or killer lymphocytes, the effect is rarely greater than
fifty percent (Guerder et al., 1995; McAdam et al., 2000), and appears to be
acting more on CD4 than CD8 cells (Newell et al., 1999). Although undoubtedly
CD28 engagement normally enhances CD8-cell activation, it provides but one of
several costimulatory pathways that can do so. CD40-CD40L interactions are also
less crucial in CD8 activation (Whitmore et al., 1999). These properties of CD8
cells as they undergo functional maturation will be important as we attempt to
understand how allograft reactions are initiated in vivo, and the role of
cytotoxicity in allograft rejection.
Transplanted
solid tumors rely on rapid establishment of a vascular connection with the host
for survival. This connection is made differently depending on the transplant.
Transplanted organs such as kidney or liver come with their own intact vascular
network, and are surgically connected to the host circulatory system at the
time of transplant. Skin transplants also come with a resident vasculature, but
for practical reasons cannot be surgically connected to the host circulation.
In that case, as we will see, the
resulting graft vasculature consists of host vessels that rapidly infiltrate
the transplant (most likely in response to signals such as VEGF emanating from
ischemic transplant cells), most of which quickly anastamose with transplant
vessels, resulting in a hybrid vasculature within the transplant. In the case
of free tumor cells transplanted subcutaneously or elsewhere in the body,
vascularization is entirely dependent on infiltration of host blood
vessels.
The cells contained in allografted tissues of these
types are almost uniformly class I MHC-positive, and thus potentially
recognizable by recipient CTL. However, in vitro studies had shown that na夫e T
cells (both CD4 and CD8) are essentially non-reactive to most parenchymal
cells, whereas (as evidenced by MLC reactions) they respond vigorously to
leukocytes. In vitro studies also demonstrated the need, within a stimulatory
leukocyte population, for an adherent accessory cell to achieve full
sensitization (Davidson, 1977). George Snell had proposed many years earlier,
based on in vivo experiments, that leukocytes entrapped in graft vasculature
(ﾒpassenger leukocytesﾓ) were likely to be the provoking cells in the rejection
of skin allografts (Snell, 1957). Steinmuller showed that leukocytes alone
could induce a state of allograft sensitization leading to subsequent
accelerated rejection of solid-tissue allografts (Steinmuller, 1967). Some
researchers thought passenger leukocytes would be necessary to provide a class
II stimulus for the production of CD4 helper T cells (Bach et al., 1976), but
others thought the situation would be more complex (Lafferty, 1983).
In fact we now know that the critical accessory cell
for optimal T-cell activation, in vitro or in vivo, is the dendritic cell,
which provides not only class I and class II MHC signals, but also numerous
ligands interacting with co-stimulatory receptors that are part of the T-cell
receptor complex (in the immunological synapse) discussed in Chapter 1.
Dendritic cells are what is generally referred to when the term ﾒprofessional
APCﾓ is used. DC encompass a
loosely related and broadly distributed lineage of bone marrow-derived
leukocytes that play a unique stimulating role in both innate and adaptive
immune responses. Dendritic cells are found in virtually every tissue in the
body, including skin (where they are known as Langerhanﾕs cells) and lymphoid
tissue, and they circulate in blood and particularly in lymph (Tew et al.,
1982; Austyn and Larsen, 1990; Shortman and Liu, 2002). Some dendritic cells
are not phagocytic, but all are vigorously endocytic and display elevated
levels of both class I and class II MHC proteins. There is compelling evidence
to support the notion that dendritic cells (DC) are the critical passenger
leukocytes brought in with tissue transplants, and stimulate the host immune
system to mount a rejection response. For example, if DC are destroyed in
grafts prior to transplantation, rejection can be circumvented, but if donor DC
are introduced into an animal onto
which such an allograft has been placed, rejection is swift and certain
(Lafferty et al., 1976; Lechler et al., 1982; Faustman et al., 1984; Benson et
al., 1987; Iwai et al., 1989). Recently attention has been focused on the role
of dendritic cells in inducing tolerance, and a possible role for this in
managing allograft rejection (Thompson and Lu, 1999).
How T cells encounter DC during the course of sensitization to
vascularized allografts has been a subject of intense investigation, driven by
the possibility that if this interaction could be manipulated it might have
profound effects on the course of transplant rejection. Key to this inquiry has
been a study of trafficking patterns of both host and donor DC to and from
lymphoid tissues (Austyn and Larsen, 1990). In general DC enter lymph nodes via
primary afferent lymphatic vessels, which originate in open tissue spaces
(lymphatic sinuses). Once inside the node, DC home to T-cell compartments such
as the paracortex. Host dendritic cells are mobilized and attracted to
inflammatory sites under the influence of locally produced inflammatory
cytokines (Kaplan et al., 1987; Sallusto and Lanzavecchia, 1999). Activated DC
drifting away from these sites and into the lymph fluid appear to have
MHC-bound antigenic peptides on their surface. Antigen-pulsed DC tend to remain
in regional lymph nodes for long periods of time, and only a few traffic onward
to the blood stream. Those that do reach the blood stream in turn home into the
spleen and liver (Kupiec-Weglinski et al., 1988).
In the specific case of allografts, the question of
whether recipient DC circulate from the blood to implanted allografts and back
to host lymphoid tissues is controversial and may depend on the grafted tissue
(Larsen et al., 1990a; Saiki et al., 2001b). On the other hand, donor DC do migrate from allografts, both those
revascularized by surgical anastamosis (cardiac, kidney; Larsen et al., 1990b),
or by infiltration of host vasculature into skin or tissue fragments implanted
e.g. under the kidney capsule (Hall, 1967; Tilney and Gowans, 1971). Both host
and donor DC migrate into draining lymph nodes and to the spleen.
Antigen-stimulated DC interact directly with and expand CD8 T cells
(Tsunetsugu-Yokata et al., 2002). Draining lymph nodes are the major site for
T-cell activation; little if any primary T-cell activation occurs at the graft
site.
The involvement of DC in the activation of both CD4 and CD8 T cells
poses interesting questions for allograft rejection that are not fully
resolved. Do host T cells respond directly to allogeneic donor DC MHC
alloantigens during a primary allograft response, or indirectly, to self (host)
DC that have scavenged donor antigenic material? In vitro, in the mixed
leukocyte culture (MLC) reaction, depletion of adherent accessory cells (now
presumed to be dendritic cells) from the responding population has little
effect on either the proliferative or cytotoxic phases of the reaction. Removal
of adherent cells from the stimulating population, on the other hand, results
in profound suppression of both phases. Moreover, as just discussed, when DC
are depleted from allografts prior to transplantation, rejection is also
greatly suppressed. These data have been interpreted to mean that responder T
cells directly recognize and are activated by MHC and other cell surface
products on allogeneic stimulator DC.
On the other hand, skin allografts taken from mice lacking functional
class II genes are rejected
just as rapidly as normal skin, implying that critical class II signals
required for recipient CD4 T cell activation can be provided by host class
II-positive cells (Auchincloss et al., 1993; Gould and Auchincloss, 1999). In
such cases it is believed that damaged donor cells sloughed off after
transplantation enter the host circulation, eventually reaching regional lymph
nodes where they are phagocytosed by recipient interdigitating reticular cells,
a form of dendritic cell that is phagocytic, which in turn present donor
information to recipient T cells
(Saikai et al., 2001a). Alternatively, donor material could be scavenged at the
transplant site by recipient DC, and carried back to regional lymphoid tissues.
The operation of an indirect activation pathway for CD4 T cells in allograft
rejection has in fact been observed (Lee et al., 1997).
Initiation
of allograft responses in free cell suspensions
The rejection of
allografts consisting of free cell suspensions (usually allogeneic hematopoieic
tumor cells introduced into the peritoneal cavity of an immunocompetent host),
also presents some interesting problems in understanding both initiation of the
response and ultimate rejection. In this situation there is no vascular
connection between graft and host; oxygen and nutrients are obtained from
ascites fluid accumulating in the peritoneum. The growth of ascites tumors in
allogeneic hosts is initially similar to the unimpeded expansion observed in
syngeneic hosts, but after several days a dramatic decrease in the number of
tumor cells occurs, and abundant, highly potent peritoneal exudate CTL (PEL)
specific for the allogeneic tumor are found within the peritoneal cavity
(Chapter 3).
A major question in such cases is, what signals on the
allografted cells control CTL generation? The vast majority of tumors used as
allografts display neither class II antigens, nor important co-stimulatory
ligands such as B7. PEL can be generated with tumor cells that have been
maintained in vitro for many generations, eliminating potentially stimulatory
contaminating cells such as dendritic cells or macrophages as the source of the
stimulus. So how does the reaction get off the ground? How are na夫e host CD4
cells, which are essential for CD8 cell maturation, stimulated to respond in
the absence of stimulatory class II antigens, and the ligands for activation
co-receptors on CD4 cells? The most likely possibility is the indirect route
for sensitization described above for solid transplants. Host dendritic cells
in the peritoneum could scavenge dead or dying graft cells, migrate to local
lymph nodes, and present donor peptides in association with class II to host
CD4 cells, which would then release the needed cytokines.
What is not entirely clear in this scheme is how CD8 cells receive
required primary and costimulatory signals ordinarily obtained from
CD4-activated dendritic cells. In allografts we think that CD8 cells are
normally triggered by allogeneic class I plus peptide on donor DC, and fed
costimulatory signals by the DC at the same time. There are no allogeneic class
I molecules on self DC. It is possible that the responding CD8 cell could pick
up this signal from the tumor cell itself if the tumor cell also displayed some of the required
accessory molecules such as CD40. While in general CD8 cells are thought not to
use their limited CD40L in activation sequences (Clarke, 2000), there have been
reports that under some circumstances they may do so (LeFrancois et al., 1999;
Eck and Turka, 1999). But clearly there are some things we do yet not fully
understand about sensitization against cell suspensions in vivo.
Cellular and molecular mechanisms of allograft
rejection worked out in vitro must ultimately be compatible with what we
actually observe in vivo. Long before such explanations were even attempted,
the classic and detailed histopathological analysis of allograft rejection
undertaken by Kidd provided a solid framework for mechanistic interpretations
of this process (Kidd and Toolan, 1950). He studied the fate of several types
of tumor cell suspensions implanted subcutaneously into susceptible (syngeneic)
and resistant (allogeneic) hosts.
When implanted into syngeneic mice, the inbred strain in which they had
originated, the cells of these cancers progressed and formed tumors that killed
the animals within a few weeks.
The cancer cells grew equally well for a time after implantation into
allogeneic mice, often forming palpable nodules within a week to ten days. After that the tumors stopped growing
and disappeared within about a week.
The mice in which the tumor allografts had regressed were highly resistant
to re-implantation with the same tumor cells, forming no visible or palpable
nodules.
Microscopic analysis revealed that five-seven days days after implantation
of allogeneic tumor cells, lymphocytes began to accumulate in blood vessels
serving the tumor nodules and surrounding tissues. Kidd observed that the number of lymphocytes increased
rapidly, extravasated, and began to penetrate among the proliferating cancer
cells, moving inward from the periphery of the nodules. The penetrating lymphocytes were highly
pleomorphic, typical of activated lymphocytes observed in the presence of
antigen in vitro. Not infrequently two or more of the lymphocytes were seen
engaged with a single tumor cell. Kidd observed that once the lymphocytes had
made contact with the tumor cells, the latter then began to die one after
another in rapid succession. Importantly, he noted that the tumor cells died in
a very different way from cells killed by heat or toxins. He was seeing
lymphocyte-induced apoptosis long before anyone else.
The cancer cells died
individually; in a given field, the microscope often disclosed
lymphocyte-associated cells in the process of dying, whereas nearby tumor cells
not yet contacted by lymphocytes, remained unchanged; some were even in
mitosis. Even when the bulk of the
cells at the periphery had been overcome, tumor cells in the central parts of
the regressing growths remained morphologically unchanged and viable.
When
the same allogeneic tumor cells were reimplanted into animals that had
previously overcome the tumor, lymphocytes infiltrated the tumor site much more
promptly. They quickly surrounded the site of implantation and became
intimately associated with the numerous tumor cells, which were nearly always
overcome in the immune hosts before they had formed palpable nodules. Although
the process of regression was greatly accelerated in the immune hosts compared
with naive animals, it was otherwise the same.
Only
lymphocytes seemed to be active in tumor allograft destruction. Neutrophils accumulated in abundance
around tumor cells within the first eight hours after implantation into mice
regardless of their immune status. However, they were notably less numerous at
24 hours and were essentially absent by 48 hours. A few macrophages were found
in areas in which regression was taking place, but they were no more numerous
here than in other areas of the tumor or in the surrounding tissue of the host,
and they never had a conspicuous relationship to the necrobiotic cancer cells
Kidd
mixed lymph node cells, harvested from mice immediately after rejection of a
tumor allograft, with fresh allograft cells in vitro for one-two hours, and
then injected the mixture into a na夫e host. No tumors developed in such cases.
Kiddﾕs
results, which confirmed and extended elements of an even earlier study
(Murphy, 1913), led him to postulate most of the rules of cell-mediated
immunity that would be ﾒrediscoveredﾓ over the next quarter century. For
example, here is a quote from his closing argument:
ﾒFrom the findings as a whole, it seems obvious
that the ﾒsensitizedﾓ
lymphocytes participate actively in the process
whereby cancer cells
are overcome in resistant and sensitized hosts.
Indeed, it might be
assumed at present, as a working hypothesis not
contradicted by any
of the available facts, that the sensitized
lymphocytes attach themselves
to the individual cancer cells and actually kill
them, thus inducing the
sequence of necrobiotic changes already described.ﾓ
In
this one passage, Kidd anticipates the work of Mitchison (1953), Govaertz
(1960) and Kerr et al (1972).
An
ingenious experiment carried out in the mid-1960s provided further insight into
the interaction of host immune cells and allograft cells (Strober and Gowans,
1965). ﾒAﾓ strain rats were connected by vascular anastamosis to a
semi-allogeneic (AxB)F1 kidney maintained outside the body
(extracorporeal transplant). After varying periods of exposure to the
semi-allogeneic kidney, the recipients were disconnected and a week later
grafted with (AxB)F1 skin. Six days later the skin graft showed
typical characteristics, as judged by histological examination, of accelerated
second-set rejection. The authors concluded that host lymphocytes had been
sensitized to ﾒBﾓ strain transplantation antigens at the graft site, and
traveled back to host lymph nodes and spleen where they completed their
activation sequence. This would seem at odds with what we currently think about
the sensitization process in vivo, and the role of dendritic cells. The
extracorporeal kidneys were perfused prior to anastamosis with the host to
remove donor blood elements, but it is possible that residual donor cells still
migrated into the host circulation during the experiment. This experiment has
never been refuted, and should be kept in mind by anyone thinking about
donor-host interactions during allograft rejection.
Yet
another approach to analyzing the involvement of cells in allograft rejection
came later, through the use of
ﾒsponge allograftsﾓ (Roberts and Hayry, 1976). A synthetic, sponge-like
material was used as a bed for growing adherent cells (fibroblasts or
dissociated solid tumor cells). Sponges containing these cells were then
implanted subcutaneously into allogeneic recipients, and after varying lengths
of time the sponges were removed and their cellular contents analyzed. When
implanted into a na夫e animal CTL activity could be detected within the sponge
matrix, peaking at day 8 ﾐ the same kinetics as development of cytotoxicity in
the spleen and lymph nodes. The sponge also contained numerous macrophages and
some granulocytes. The fact that activated CTL were found in regional lymph
tissue suggests the graft had been infiltrated by blood vessels, although this
was not directly determined. When sponges were implanted into a previously
immunized mouse, typical accelerated kinetics of infiltration and appearance of
reactivated CTL in local lymph nodes were observed (Roberts, 1977).
The role of vascular
damage in allograft rejection.
Initially
it was imagined that allograft rejection by lymphocytes would involve a
cell-by-cell destruction of the entire graft by infiltrating killer T cells,
essentially as described by Kidd. However, another possibility was put forward
early on. In those cases where the graft is implanted with its own vascular
system intact, killer cells might only have to attack and destroy the
vasculature itself; the rest of the graft would die quickly through ischemic
infarction. Evidence that this might be the case was gathered from a number of
different allografting situations in animals and in humans (Henry et al., 1962;
Waksman, 1963; Kountz et al., 1963; Flax and Barnes, 1966).
One of the earliest detailed analyses of the
interaction between host immune cells and donor blood vessels was carried out
by Dvorak in allografted human skin. Using a highly sophisticated microphotometric
technique, Dvorak had observed that in contact dermatitis (a form of DTH
reaction), infiltrating lymphocytes tended to cluster around capillaries and
other blood vessels, and that the vessels appeared to be seriously damaged. In
most cases, however, the reparative phase of inflammation took over and
prevented complete destruction of the vessels (Dvorak et al., 1976).
Subsequently applying his technique to an analysis of events accompanying skin
allograft rejection, he saw the same thing ﾐ lymphocytes clustering around
blood vessels, followed in this case by actual destruction of the vessels.
Destruction of blood vessels in every case could be observed before necrosis of
dermal cells and other nonvascular graft elements, which he concluded died from
infarction and subsequent necrosis (Dvorak et al., 1979). Interestingly, host
blood vessels in the immediately adjacent graft bed were not injured,
suggesting considerable specificity of the host immune response.
These basic observations were confirmed and extended
using a SCID mouse system in which the interaction between human lymphocytes
and allogeneic human skin were studied. Human skin was grafted onto a SCID
mouse, and allowed to heal in before intravenous infusion of allogeneic human
lymphocytes (Murray et al., 1994). One of the important contributions of this
paper concerns graft vascularization. It was long thought that in skin
allografts, donor vasculature died via necrosis and was replaced by
infiltrating host blood vessels. However, Murray et al showed there is also
considerable spontaneous anastomosis between donor and host vessels, even
across species barriers. He found that class II MHC molecules and adhesion
ligands were quickly upregulated on the human vascular endothelium, and human
lymphocytes were seen clustering around the human vascular elements in the
graft, but not mouse vascular elements. Many of the lymphocytes clustering
about blood vessels were perforin-positive. The graft human blood vessels as
well as the human/mouse hybrid vessels were ultimately destroyed, but the graft
survived as mouse blood vessels replenished the graft bed. The fact that host
vessels were also destroyed in the experiments of Dvorak et al., but not those
of Murray et al., is likely related to the fact that the latterﾕs observations
were made on grafts that had already healed in prior to onset of the allograft
reaction, whereas Dvorakﾕs observations were made on allografts still in the
process of healing in. Dvorakﾕs observations thus more closely approximate a
normal allograft situation.
The destruction of graft blood vessels as a
preliminary step in vascularized allograft rejection has been further
documented in heart (Forbes et al., 1983), liver (Matsumoto et al., 1993), and
kidney (Busch et al., 1971; Leszcyznski et al., 1987) allografts. In these cases, host and donor blood
vessels are surgically anastomosed, and vascular damage to the graft is rapid
and lethal. Perivascular accumulation of lymphocytes and monocytes/macrophages
is followed by vascular degeneration, which is in turn followed shortly by
graft failure. Thus in all solid tissue and organ allografts, it seems clear
that allograft rejection is not caused by cell-by-cell destruction of the graft
parenchyma, but rather by vascular damage, ischemia, and global tissue
necrosis.
Vascular endothelial cells can in fact be killed in
vitro by appropriately sensitized CTL (Collins et al., 1984). There has also
been a suggestion that endothelial cells may themselves be stimulatory for CD8
CTL, although this notion has come largely from a single laboratory. Although
CTL can be generated that have a degree of endothelial cell specificity, the
CTL are rather atypical and it is not obvious from the data presented that
endothelial cells are a major target in primary allograft sensitization and
rejection (Biederman and Pober, 1998; Dengler and Pober, 2000).
DTH reactions and
allograft rejection
DTH
reactions are inflammatory in nature, and as such are often accompanied by
significant ﾒcollateral damageﾓ ﾐ destruction of nearby healthy tissues
antigenically unrelated to the provoking antigen. But inflammatory reactions
are remarkably efficient in clearing a wide range of potentially lethal
infections, and moreover have built-in mechanisms that limit collateral damage
and initiate rapid healing. The establishment of a delayed hypersensitive state
to biological or chemical antigens can be brought about in numerous ways, which
are beyond the scope of the present discussion. Establishment of a DTH state in
a na夫e animal is absolutely T-cell dependent, and depending on the particular
antigen, its mode of administration, and the species involved, either CD4 T
cells, CD8 T cells, or both, may be involved. Memory T cells resulting from the
primary response initiate an enhanced inflammatory reaction upon re-exposure of
hypersensitive animals to the original antigen.
The role of T cells in DTH is generally thought to be
restricted to antigen recognition and the release of numerous inflammatory
cytokines. The antigen-presenting cells in DTH are the same as in other T-cell
activation reactions, namely dendritic cells and probably macrophages. Among
the cytokines important in the development of DTH are IL-3, GM-CSF, IFNg, TNFb, MAF and MIF. These cytokines play a role in attracting
monocytes and other nonspecific cells to the site of the inflammation, and in
inducing changes in local vascular endothelium that favor extravasation of
these cells into surrounding tissues spaces. Perivenous accumulation of
monocytes is one of the early hallmarks of DTH reactions. Although full
development of a DTH lesion may take up to 48 hours, an accumulation of
macrophages may be evident in as little as 4 hours. T-cell cytokines are also
important in activation of inflammatory cells. Particularly important in this
latter regard is the role of MAF (macrophage activating factor), which promotes
maturation of monocytes into macrophages, and stimulates enzyme production and
phagocytosis. Once activated, macrophages and other nonspecific cells release cytokines
of their own which further promote the inflammatory response. The damage caused
by activated macrophages, which are considered the principal inflammatory
mediator in DTH, is due to release of numerous hydrolytic enzymes and other
degradative molecules during phagocytosis of nearby damaged cells. This damage
makes no distinction between the compromised cells provoking the initial
reaction and surrounding normal cells. However, once the provoking antigen is
removed, the reaction rapidly diminishes and the macrophages release a spectrum
of mediators that promote healing of residual tissue.
It would seem not unreasonable that some portion of
the tissue destruction that occurs during allograft rejection, and other in
vivo immune phenomena we will
discuss in subsequent chapters, could
be caused by the inflammation accompanying DTH. In fact, for many years
immunology textbooks have classified allograft rejection as a ﾒType IVﾓ
hypersensitivity reaction. Other Type IV (DTH) reactions include the tuberculin
reaction, contact reactivity to certain substances such as poison ivy or picryl
chloride, and inflammatory responses to various fungal substances (Gell and
Coombs, 1968). The case for involvement of DTH in allograft rejection, or at
least its concomitant occurrence, was first made by Brent et al. (1962) and
Brent and Medawar (1966), who showed that allografting is accompanied by a
hypersensitivity state demonstrable by intradermal injection of graft donor
cell extracts. This results in a typical ﾒwheal and flareﾓ reaction (erythema
and induration) at the injection site within 1-3 days. It gradually became
accepted that transplantation immunity, along with previously defined DTH
reactions, were different expressions of a newfound ﾒcellular immunityﾓ, or ﾒcell-mediated immune responseﾓ (Brent et al., 1962;
Waksman, 1962a; Uhr, 1966; Turk and Stone, 1963; Cooper et al., 1968; Good et
al., 1969).
Thus a major question that must be addressed is, to
what extent is allograft rejection a result of inflammatory DTH reactivity, and
to what extent is it caused by cell-mediated cytotoxicity ﾐ by killer
lymphocytes? Although triggering of a DTH inflammatory response requires
specific T-cell recognition of allografted cells, the subsequent attack
mechanism could be independent of target cell recognition, and basically any
cell in the region of the reaction may be killed. This is the in vivo
equivalent of the ﾒinnocent
bystanderﾓ issue raised with in vitro systems evaluating CTL function
(Chapter 4). On the other hand, if allograft rejection in vivo is caused by CTL
directly engaged with specifically recognized target cells, there should be
little or no collateral damage to innocent bystanders.
Early
insight into this question in vivo was provided by a study carried out in Sweden
(Klein and Klein, 1972). In one variant of the experimental scheme, varying
numbers of sarcoma cells syngeneic to the host were mixed with allogeneic tumor
cells to see if the former were killed as part of a generalized response to the
allogeneic tumor cells. In fact, even when the syngeneic tumor cells were mixed
at a frequency of only 10-4 with the allogeneic tumor cells, they
were not killed; they grew out as syngeneic tumors that eventually killed their
host. In this subcutaneous environment the dissociated tumor cells very likely
reassociate, but certainly remain closely associated once injected. Yet the
syngeneic cells were clearly not killed. These results argue for a highly
discriminatory immune attack on the implanted cells.
Additional insight into this question came from an ingenious
experimental system using allophenic mice (Mintz and Silvers, 1970). Allophenic
or tetraparental mice are produced by mixing cells of allogeneic embryos at
preimplantation stages of development. Individual cells in allophenic mice
express (for example) either H-2a or H-2b MHC antigens,
with no cells co expressing both H-2a and H-2b. Mintz and
Silvers found that when allophenic skin was grafted onto one of the parental strains, only melanoblasts and hair
follicle cells expressing the MHC of the opposite parental strain were
rejected, whereas those expressing the hostﾕs MHC type did not suffer
irreversible damage.
Rosenberg and Singer refined this basic experiment.
When skin from H-2a-b allophenic mice was grafted onto
immunoincompetent H-2b nude mice, the resulting graft was accepted
and produced patches of both white (from the H-2a skin cells) and
black (from the H-2b skin cells) fur. Subsequently, T cells from a
normal immunocompetent H-2b mouse were infused into the engrafted
H-2b nude mouse. Initially, there was a vigorous inflammatory
response at the graft site that caused extensive damage to epidermal cells of
both H-2a and H-2b origin. However, once the inflammatory
response subsided, the underlying dermal cells returned to a healthy state, and
regenerated fur - but only black (H-2b) fur. The H-2a
cells had been killed by the post-inflammatory immune response of the infused
H-2b T cells, but the H-2b-expressing cells were not
killed (Rosenberg and Singer, 1988). Other experiments illustrating the
exquisite specificity of allograft rejection in vivo involved placing syngeneic
skin grafts adjacent to allogeneic grafts on a single graft bed, with the
rejection showing a clear demarcation at the division between the two grafts.
The researchers involved in all of these experiments
drew two conclusions. Allograft reactions are indeed accompanied by an
inflammatory response (DTH), and this response can cause considerable
collateral damage that is independent of graft-cell MHC expression. However,
this damage is not sufficient to lead to allograft rejection; only those graft
cells specifically recognized by effector T cells appear to be killed. If graft
rejection were due to an inflammatory process, they reasoned, we would expect
both to be killed, since the inflammatory mediators would spread readily
throughout the entire graft. The data strongly suggested that allograft
rejection in vivo is caused by antigen-restricted CTL bound to specific target
cells, and not by antigen-nonspecific inflammatory mediators. Moreover, as
shown subsequently, in mice in which the genes for interferon, or for
interferon plus IL-2 ﾐ the two principal inflammation-promoting cytokines ﾐ are
deleted, allograft rejection is unimpeded (Saleem et al., 1996; Zand et al.,
2000).
While
the notion that inflammation is not an exclusive or even general mechanism of
allograft rejection may be correct, it is based on assumptions about the
functional radius of DTH reactions that are difficult to assess. One of the T-cell cytokines released at
the site of DTH reactions is macrophage inhibitory factor (MIF), which exerts a
potent cytostatic effect on macrophages and would be expected to inhibit their
migration much beyond the immediate site of the initial response (David et al.,
1964). Moreover, given that the principal cellular targets in allograft
rejection in most cases are vascular endothelial cells, inflammatory damage to
the blood vessels around which monocytes and macrophages are clustered could
appear to be highly specific in the allophenic system described above, and
sufficient for allograft rejection.
In the case of ascites
tumor allografts, the early stages of rejection are marked by a massive influx
of macrophages (and possibly dendritic cells), the number of peritoneal
macrophages being proportional to the number of tumor cells injected. This
correlation suggested initially that macrophages are responsible for the tumor
rejection. At the peak of allogeneic ascites tumor growth, peritoneal macrophage
populations consist primarily of small macrophages with compact cytoplasm and
rounded nuclei; they also exhibit only slight phagocytic activity for the tumor
cells. As the number of tumor cells decreases, larger macrophages, which
contain irregular lipid granules and tumor cell fragments (Amos, 1962) and
which adhere to the tumor cells in vivo (Baker et al., 1962), are observed. In
the system studied by Baker and co-workers, the intraperitoneal injection of
tumor cells elicited tumor cell-macrophage clusters within 30 minutes, and the
generation of cellular debris and large macrophage-tumor cell clumps within
several hours. Given the prime
role of macrophages in mediating DTH reactions, it is hard to imagine that they
are not causing inflammmatory damage to ascites tumor allografts. However, as
we have seen, PEL invading the peritoneal cavity are very efficient CTL and
specifically kill cognate tumor target cells in vitro. In addition, mixing
experiments of the Klein and Klein type just discussed have also been carried
out in the peritoneal cavity. Even at large allogeneic to syngeneic tumor cell
ratios, only the allogeneic tumor is cleared from the peritoneum, leaving the
syngeneic tumor to grow out and ultimately kill the host (G. Berke, unpublished
observations). Still, given the rampant macrophage-mediated inflammatory
reaction that develops in the peritoneum, it seems unlikely that some tumor
cells ﾐ perhaps both syngeneic and allogeneic - are not killed as a result of
inflammatory damage.
Although
antigen-specific lymphocytotoxicity was discovered in connection with allograft
rejection, it must be admitted that some 40-plus years later the evidence for a
role of cytotoxicity in allograft rejection remains indirect at best. That
cytotoxic T cells are generated in allograft reactions is beyond question, nor
is the primacy of CD8 T cells in causing allograft rejection at issue. If the
development of CD8 cells is prevented in mice, by disruption of the gene for b-2 microglobulin, the ability to reject allografts is lost[1].
What is unclear is the extent to which CD8 T-cell-mediated cytotoxicity is a factor in allograft rejection in vivo.
Antigen-specific, in vitro-cytotoxic T cells can be recovered from within rejecting
allografts (Bradley et al., 1985). Immune CTL populations, as well as CTL
clones, can cause accelerated graft rejection in immunoincompetent hosts.
Perforin- and granzyme-containing cells can be identified by in situ
hybridization at graft rejection sites, and local levels of these agents seems
to correlate with the vigor of rejection (Griffiths et al., 1991; Clement et
al., 1994). Given what we know about CTL, perforin and granzymes it is hard to
imagine that the cytotoxic processes we observe in vitro with these cells are
not taking place in vivo. But most of the data gathered over the past four
decades, while consistent with a
role for cytotoxicity in allograft rejection, do not demand that cytotoxicity be a factor in rejection.
Important insights into involvement of the two known mechanisms of
CTL-mediated lysis have come from gene knockout mice (perforin) and the natural
mutant mouse strains lpr and gld (Fas and FasL mutations, respectively). In
non-vascularized islet cell or tumor allografts (Walsh et al., 1996; Ahmed et
al., 1997), in skin allografts with mixed donor-recipient vascularization
(Selvaggi et all, 1996), and in revascularized heart allografts (Schultz et
al., 1995), the same striking results were obtained. When donor cells or tissues
from an lpr mouse, which lacks Fas
expression, are transplanted into an allogeneic perforin knockout mouse, the
rate of rejection of the allografts is not different from the rate of rejection
of Fas-expressing tissues transplanted into allogeneic perforin-expressing
mice. In the former cases, neither perforin nor Fas can be involved in
allograft rejection, yet the rejection rate is unaffected. As expected,
allograft rejection in such cases is not accompanied by generation of CTL
capable of lysing target cells in vitro. These results ﾐ particularly those
involving transplantation of hematopoietic tumor cells like those used as
targets in vitro CTL assays - provide powerful evidence that cytotoxicity
mediated by CTL is not an absolute requirement for allograft rejection in
vivo.
In all the above instances, mononuclear and
lymphocytic infiltration of the allografts was the same, whether functional CTL
(as measured by in vitro assays) were generated or not. In the case of
hematopoietic tumor allografts (Walsh et al., 1996), some of the Fas-negative
tumors were resistant to TNFmediated cytotoxicity, yet were rejected with the
same kinetics as TNFsensitive tumors, suggesting TNFmediated killing (Chapter
5) is also not a factor in allograft rejection. Most (17/21 tested)
Fas-deficient tumors placed in vivo undergo upregulation of Fas expression
(Rosen et al., 2000). Although it is technically possible that in the case of
Fas-negative tumor allografts, some tumors may have expressed Fas after
transplantation under the influence of inflammatory cytokines (Rosen et al.,
2000), the results just cited with lpr tissue transplants reinforce a lack of requirement for Fas in
allograft rejection.
Cytotoxicity may thus simply be a marker for the
sensitization of CD8 T cells that utilize an as-yet unidentified,
non-cytotoxicity-based mechanism to effect allograft rejection. There have been hints of the
involvement of perforin or Fas in a few cases, under special circumstances. For
example, in the study of Schultz et al. cited above, if the allograft donor
differed from the recipient at only a single class I MHC gene (Sub-locus), then
Fas-deficient hearts were rejected more slowly in perforin-negative
recipients (88 days survival) than
in perforin-positive recipients (31 days survival), suggesting a possible role
in this situation for perforin. And when gld mice, which lack the Fas ligand, are used as
recipients for fully allogeneic cardiac transplants, rejection is prolonged
compared to normal recipients. However, when cardiac tissue from lpr mice, which lack Fas antigen, is transplanted into
fully allogeneic normal controls, there is no impact on the rate of rejection
(Seino et al., 1996). Thus the general immune disorder in the gld mice may account for the observed differences in
graft rejection, rather than a defect in the FasL/Fas pathway of
lymphocyte-mediated cytotoxicity per se.
Interestingly, both perforin and Fas appear to be involved in an
allograft-related situation in vivo, namely graft-vs. -host disease (GVHD).
GVHD is a major complication of bone-marrow transplantation, wherein mature
donor T cells present in bone marrow preparations (in part from ruptured bone
marrow blood vessels) react against host MHC antigens. In immunocompromised
recipients, this can result in skin and gastrointestinal lesions,
lymphadenopathy, wasting and death. In a mouse model wherein recipients are
irradiated prior to infusion of donor T cells, CTL specific for recipient class
I MHC antigens develop in lymphoid organs, and the recipients invariably die
after 6-8 days. When either the perforin or Fas pathways are blocked in this
model, mortality is significantly delayed; when both pathways are blocked,
mortality is eliminated. (Braun et al., 1996; Baker et al., 1996; Baker et al.,
1997). Whether disruption of these pathways affects other manifestations of
GVHD was not addressed in these studies. A critical role for granzyme B has
also been discerned in GVHD mediated by CD8 T cells (Graubert et al., 1996).
Another surprising situation in which the perforin and
Fas pathways play a role involves cutaneous DTH. Contact skin sensitivity to
the chemical di-nitrofluorobenzene (DNFB) in mice is mediated by CD8 T cells
that recognize DNFB-modified peptides on self class I MHC molecules. The
resultant inflammatory reaction manifests itself as an obvious dermatosis upon
secondary exposure to DNFB, and CTL specific for DNFB-modified target cells can
be isolated from adjacent lymphoid tissues. When mice doubly deficient in the
Fas and perforin lytic pathways were painted with DNFB, they developed neither
DTH reactivity nor actively lytic CTL. Mice deficient in either pathway alone
developed both DTH and CTL, suggesting that either pathway is sufficient to
support both reactivities (Kehren et al., 1999).
So where does all this leave us in terms of defining
the mechanism of allograft rejection in vivo? As discussed earlier, the
exquisite specificity of in vivo rejection would seem to argue against a
generalized inflammatory attack on allografts as a mechanism. The Klein and
Klein tumor mixing experiments in particular are difficult to reconcile with
inflammation as an exclusive means of allograft rejection. In that experiment,
recall, the syngeneic and allogeneic tumor cells were confined in a physically
constricted space. Yet at a concentration of one syngeneic tumor cell per
thousand or less, only the allogeneic tumor cells, and not the syngeneic cells,
were rejected.
The specificity of CTL killing (as measured in vitro)
seems more consistent with the specificity of allograft rejection we observe in
vivo, and has since its first description been the presumed effector mechanism
of choice in explaining allograft rejection. Cytotoxic T cells arise in vivo
with the onset of allograft rejection; they recognize, bind to and kill
specific targets in vitro; and passive transfer of CTL to immunocompromised
hosts results in rapid and specific graft rejection. Could we imagine that CTL
do not conjugate with cognate target cells in vivo? And once conjugated, could
we imagine that they do not release cytotoxic granule contents, or engage with
Fas where it is expressed? Yet as
we have just seen, mice deprived of the use of these pathways of CTL-mediated
cytotoxicity, as well as membrane TNF, are unimpaired in their ability to
reject allografts. As counter-intuitive as it seems, we may have to admit that
the vigorous cell-mediated cytotoxicity we measure using 51Cr
release assays in vitro is not required for allograft rejection in vivo. Of
course, failure to disrupt allograft rejection in the absence of any one, or
even all, of these cytotoxic mechanisms doesnﾕt mean that cytotoxicity has no
hand in allograft rejection ﾐ only that we must assign it a more modest role.
Other mechanisms are clearly able to compensate for its absence.
We are left then with the vague and somewhat unsatisfying view that
multiple, overlapping effector mechanisms must exist for allograft rejection,
some involving cytotoxicity and some not, some with specificity for donor
antigens, and others involving indirect effector processes. We began this
chapter with a brief discussion of the historical entanglement of DTH reactions
and allograft rejection, and took note of the fact that for many years
cell-mediated cytotoxicity was listed as simply one of several manifestations
of DTH. But the in vitro evidence for a swift, powerful, highly specific
cytolytic mechanism that could cause the complete destruction of allogeneic
cells in a matter of minutes led much of the field away from the view of CTL acting
through a mechanism notable for its lack of discrimination. It seems that in
the end we may have come full turn; it may be time to put CTL back where they
started out, as one of the mechanisms of Type IV hypersensitivity.
Perhaps this should not be surprising: given that
allograft rejection can never be regarded as anything other than an accident of
nature, why should there be a distinct, highly focused mechanism to deal with
it? The rejection of organ and tissue transplants is clearly secondary to some
other, more natural process, and as best we understand it, that process is
elimination of cells compromised by intracellular pathogens, and perhaps
oncological transformation. We turn to these more orthodox immunological
challenges in the following chapters.
[1] It should be noted that such mice are not absolutely class I-negative. Some class I, especially at the D locus in mice, does get through to the surface in the absence of b-2 microglobulin (Bix and Raulet, 1992).