Feature Story

Eyeing a new path for tumor growth

April 2000
Karen Titus

The letters and e-mails have been streaming in from pathologists all over the world.

Some are highly encouraging, reports Mary J.C. Hendrix, PhD, a recipient of the missives. "They say, ’Oh, I’ve seen this before and I just never knew what to make of it,’" says Dr. Hendrix, deputy director of the University of Iowa Cancer Center and Kay Daum Professor and head of anatomy and cell biology, University of Iowa College of Medicine, Iowa City.

Others are less supportive. "They tell us we’re full of hooey," she admits.

The root of such witness and wrath is a term coined by Dr. Hendrix and two colleagues to describe what was previously unthinkable: a means by which certain aggressive tumors create their own patterned vascular channels.

This so-called vasculogenic mimicry, they contend, is characterized by a network of back-to-back closed loops of matrix that appear to carry red blood cells but lack endothelial cells. The matrix loops encircle islands of tumor cells "like the beltway around Washington, DC," says Robert Folberg, MD, a member of the research team and currently professor and head of the Department of Pathology at the University of Illinois-Chicago. The patterning reflects tumor behavior and continues in the sites of metastatic disease, he adds, and its presence is highly predictive of poor outcome.

The researchers’ work to date has focused primarily on uveal melanoma and cutaneous melanoma; the implications are enormously exciting for ophthalmology in particular, Dr. Folberg says, because the patterning could be used as a noninvasive substitute for biopsy. "You can’t take a biopsy of the inside of the eye without blinding the patient," he points out.

That point is hardly debatable. But, as the team learned not long after publishing its findings last fall in The American Journal of Pathology (1999;155:739-752), other aspects of their work were. Critics accused the researchers of everything from shoddy science to attacking angiogenesis.

"This is controversial," Dr. Folberg acknowledges. "The pathology community will either hate this, or they’ll be puzzled by it—I don’t know that anyone will love it," he says, although any theory that likens Washington, DC, to a tumor is bound to have at least some appeal.

Some are intrigued. "It’s fascinating work for several reasons," says Charles Platz, MD, professor of pathology, University of Iowa Hospitals, Iowa City. Though he and Dr. Folberg occasionally consulted cases back and forth when the latter was at Iowa, Dr. Platz says, his only tie to the team’s research has been as an interested observer. "Bob showed me some of his early observations, so I’ve been aware of what they’re doing," he says.

Here’s how Dr. Platz—and presumably others who have written the Iowa team to express their figurative "aha’s"—explains the appeal of their theory:

"As I look at neoplastic disease, there are two or three broad categories of growth patterns, particularly for epithelial neoplasms. We were taught that the pattern of a carcinoma was an epithelioid, an organoid growth pattern. That pattern in its classical sense is the pattern they’re describing in their uveal tumors," he says. "Other cancers, such as an infiltrating carcinoma of the breast, grow in a much more disorganized fashion.

"As one looks at and compares these two groups of tumors, those that have a distinct, uniform structured pattern and those that grow in seemingly random, haphazard arrangement look very different," Dr. Platz continues. "And it’s always puzzled me, as I have read the angiogenesis literature, that no one pays any attention to how the blood supplied to these two different kinds of tumors would appear to be different because of their structural pattern.

"What they’ve done, as far as I’m concerned, is offer an explanation for a neoplasm growth pattern that I have looked at under the microscope for many years, one that’s made me say to myself, ’I don’t understand where this neoplasm gets the blood supply to support the kind of growth that it’s exhibiting,’" he suggests.

"In the course of day-to-day work and trying to decide what kind of tumor I’m dealing with, some tumors turn up with strikingly greater number of vessels as identified by the usual CD31, CD34, factor VIII markers for endothelium," Dr. Platz adds. "And some markers turn up with remarkably few vascular structures as identified by those markers. As I think back over what I’ve seen, I suspect what’s going on is those tumors are doing what [the Iowa researchers are] suggesting they can do, which is they’re building their own perfusion channels."

Or not. The researchers’ explanation differs so radically from prevailing thought about tumor growth that many pathologists may not even be familiar with the concept, but some who are have dismissed it as "nothing but artifact."

Among the most public criticisms appeared in the February issue of AJP (2000;156: 383-388). The commentary was written by Donald M. McDonald, MD, PhD, Lance Munn, PhD, and Rakesh K. Jain, PhD; its genesis, Dr. Mc-Don-ald says, was discussions he and Dr. Jain had about the scientific merits of the Iowa researchers’ arguments.

"We came up with this because each of us independently felt there were problems with the paper published in September," says Dr. McDonald, professor of anatomy at the University of California, San Francisco. "We came to the conclusion independently, and then happened to be discussing it, and we found that we had the same sentiments. We thought it might be most useful for the scientific community at large if we would put together a joint commentary."

Here the controversy thickens a bit. Dr. McDonald acknowledges that a more typical response for him would be to deal directly with researchers whose work he questioned; he was spurred to the more visible response of a published critique because the September AJP paper represented "more than a standard research contribution," he says. "This immediately was in a more public view."

The spotlights shone from several directions. The article on vasculogenic mimicry was accompanied by a commentary by Mina J. Bissell, PhD, of Lawrence Berkeley National Laboratory at the University of California, Berkeley (1999;155: 675-679), which suggested the work shed "new light on our understanding of tumor perfusion." The headline of an article in Science (1999;285:1475) announced "New type of blood vessel found in tumors." In reiterating the researchers’ contention that the vascular channels were distinct from active tumor angiogenesis, the story raised the issue of whether current antiangiogenesis agents would be effective against the matrix patterned vascular channels. And the Chicago Tribune ran an inflammatory page-one story (Sept. 17, 1999) that described vasculogenic mimicry as a "potential chink" in angiogenesis theory.

With the topic now placed firmly in the public forum, Dr. McDonald and his co-authors decided to make their concerns public as well. "Our response wasn’t just for the authors; it was more for the scientific community that had been exposed to a new idea."

An idea, he adds, "that we felt was wrong."

The idea is actually a handful of ideas, which turn several conventions on their head. "Part of our problem in getting our paper published was, How many paradigms were we willing to break in one manuscript?" Dr. Folberg says.

The concept of vasculogenic mimicry grew, in part, out of observations made by Dr. Folberg and others in the late 1980s and early 1990s, during the search for a noninvasive substitute for biopsy for uveal melanomas. "That’s when we came up with this interesting observation that there were all these interesting channels forming inside the tumor," Dr. Folberg says.

The channels differ from normal blood vessels in several ways, says Andrew J. Maniotis, PhD, a member of the research team. One difference is their tiny size; another is their organization. "They’re patterned as though they’re ribbons wrapped around a bunch of grapes, with the grapes representing the clusters of tumor cells," he says. Moreover, the contents of the vascular channels do not clot yet appear not to contain endothelial cells, he says.

That latter observation was made by Dr. Maniotis, a research scientist at the University of Iowa College of Medicine, when he teamed with Drs. Folberg and Hendrix (whose research focuses on the metastatic properties of melanoma cells) to puzzle out the link between the patterns associated with aggressive tumors but not with the nonaggressive tumors.

Drawing on research experience honed in the laboratory of Harvard’s Don Ingber, MD, PhD—which is in the research division headed by angiogenesis pioneer Judah Folkman, MD—Dr. Maniotis compared cells from aggressive and nonaggressive tumors, trying to discern differences in their abilities to form blood vessels. "My expectation was that I would see the angiogenic switch for the first time," he recalls. "I wanted to be the first to see it in a dish, in three dimensions." Under Dr. Hendrix’s direction, Iowa had recently begun a large initiative in three-dimensional culturing—"tissue engineering of cancers," Dr. Maniotis calls it.

Dr. Maniotis placed aggressive cells with endothelial cells and, separately, nonaggressive cells with endothelial cells. The aggressive cell cultures made what appeared to be tubes and the nonaggressive cultures did not.

"Then I started staining the channels for endothelial markers to prove that the tubes were made of endos," he recalls. "And they didn’t label—not very much, anyway. It wasn’t a consistent lumen that got labeled. I got suspicious, so I deleted the endothelial cells from the experiment, just as a control.

"What I saw blew me away."

What he saw was that even sans endothelial cells, the aggressive cells continued making the patterned matrix.

Dr. Maniotis, who says his initial reaction was one of pure disbelief, continued to experiment. Using different uveal melanoma cell lines in a variety of culture conditions confirmed that aggressive melanoma cells, by themselves, can form networks of matrix.

The aggressive melanoma growth pattern could permit the perfusion of a tumor, says Dr. Hendrix, and she and some of her colleagues suspect the patterned vascular channels may allow a route for dissemination of metastases. "The most direct evidence will be with some in vivo microscopy, and we’re trying to design those experiments right now," she says.

The researchers also began collaborating with two cancer geneticists at the National Human Genome Research Institute in Bethesda, Md., trying to learn whether a molecular analysis of the aggressive tumor cells could provide some clues regarding their function. Microarray analysis of 5,000 genes comparing highly invasive melanoma cells with poorly invasive ones revealed 210 genetic differences between the two types of cells, which suggests "There is a lot of disregulation going on in the aggressive phenotype," says Dr. Maniotis.

This differential gene expression, the researchers noted in their September paper, was associated with a combination of phenotypes, including endothelium (TIE-1) and epithelium (keratin 8). "Each tumor cell is expressing genes it shouldn’t express," says Dr. Folberg, who calls this phenomenon "the most exciting aspect of our work. What you’re basically looking at is an embryonic genotype in the highly malignant cells. And what do embryonic cells do? They migrate—the analog here is metastasis."

This offers a conceivable therapeutic target, Dr. Folberg proposes. "If we can figure out how to shut this down, we’re not only shutting off the primary tumor, but we can shut off the metastases as well. Or, we could use the patterning to try to detect the metastases."

The genes may eventually provide other diagnostic targets, Dr. Hendrix adds. "There are at least a dozen genes that might allow a pathologist to predict the potential of a certain tumor to be highly aggressive versus nonaggressive," she says.

Should further research confirm this notion of vasculogenic mimicry, pathologists may find it has implications for vessel counting as well. In a review paper published in AJP (2000; 156:361-381)—though not as a response to the McDonald commentary published in the same issue—the team raises the matter of compartmentalization. Vasculogenic mimicry may be restricted to a tumor’s cellular compartment, they hypothesize, while vessels associated with angiogenesis are seen in the stromal compartment. "If you’re not aware that vasculogenic mimicry may be present in a tumor, your vessel counts may be off," Dr. Folberg submits. The possibility of multiple perfusion channels might also explain why some therapeutics are more effective than others, he adds. "Which compartment is being perfused—the stromal compartment, the connective tissue compartment, or the cellular compartment? Maybe that explains why certain chemotherapeutic agents get inside the tumor and others don’t."

The team is now looking at a variety of tumor systems. "We’re really looking at everything in which the tumor forms a solid compartment of tumor cells," says Dr. Folberg. "If tumor cells permeate but one cell at a time through connective tissue, they’re probably not going to exhibit this phenomenon. But if they form a cluster of tumor cells, a big nodule, then they’re candidates for this to create a perfusion in the cellular compartment."

The team now faces "a huge agenda," he continues. "We have to do more work on topology to understand what this looks like in three dimensions, and that’s what we’re working on furiously now. We need to figure out how many other tumors do this. We need to look at molecular mechanisms by which we can shut this process down. And we are working on techniques to better image these patterns, because if they are linked to the behavior of the tumor, then we can come up with a number of noninvasive imaging strategies that can be used elsewhere, not just in the eye."

They also face a chorus of critics who suggest their theory needs to be taken back to the bench.

The existence of the looping patterns is not the issue—"Those things are there," says Dr. McDonald. What is being questioned is what the networks represent, and how significant they may be.

Dr. McDonald, who says his primary research focus in recent years has been endothelial cell biology and tumor microcirculation, questions whether the networks—defined by the team as three or more PAS-stained back-to-back loops—give some indication of the architecture of the vasculature. "The Folberg group says yes, and our statement is very clearly no." Instead, he says, the stained networks are likely connective tissue patterns that are a feature of the tumor.

"I think a fundamental problem here is that the PAS stain is not the right tool to use," he says. "It is not a specific stain for blood vessels, it never has been, and it never will be. Sometimes it catches blood vessels, and other times it doesn’t. So don’t use it to try to make some deduction about the geometry in the vasculature."

Dr. Ingber, professor of pathology at Harvard Medical School and Children’s Hospital in Boston, seconds that concern. "I would like to see additional staining done for antibodies for known connective tissue components, which would help define the presence of a basement membrane or connective tissue collagens, or for other evidence of endothelial cells." Likely candidates would be basement membrane molecules, such as laminin and type IV collagen, rather than interstitial collagens.

Dr. McDonald offers an alternative explanation for what the Iowa group has observed: The red blood cells purportedly contained in the vasculogenic channels could well be extravasated red blood cells. "That’s a very common feature of tumors," he says.

It’s an idea that Dr. Folberg quickly rejects. "It’s not leakage," Dr. Folberg insists, noting that the red blood cells under discussion are confined to thin, tube-like channels. "When blood leaks, you see it all over the place."

Counters Dr. McDonald, drawing on his own research in tumor leakage: "Blood leakage is a fascinating and underexamined area of tumor biology. Blood vessels of tumors are known to be leaky, but I think there’s more to the story than that." The distribution pattern of leaked red blood cells is determined by a tumor’s structure, he says. Each tumor contains relative barriers that restrict distribution of material from its blood vessels. "It’s not just random," he says.

"What I think happens is, depending on the type of tumor, the red cells can be confined to a specific site and even look like a blood vessel. They can appear as spots that, for whatever reason, tend to aggregate based on the connective tissue barriers. I think it would vary from tumor to tumor as to how this would all play out."

Hemorrhage in uveal melanomas is common, Dr. McDonald adds. "The literature always makes the point that it’s unclear whether that hemorrhage occurs in the tumor spontaneously or is a product of the surgical removal of the tumor."

Dr. Ingber suggests another interpretation: that the networks are remnants of pre-existing vessels that have necrosed. He uses the analogy of scaffolding that remains even after the structure it once helped support has crumbled or decayed. Perhaps the tumor, in its early stages, featured normal capillaries whose endothelial cells at some point died or became degraded, leaving behind the basement membrane or fragments of extracellular matrix. "What remains is basically a remnant space," he says.

The complex nature of tumor microvasculature defies any simple explanation, he adds. "The connective tissues that encapsulate tumor cells are usually very vascular—but not always," he says. "Some of them aren’t—it’s mostly just loose connective tissue. If I were to force in a fluorescent dye under pressure, it would fill the spaces between the lobules. But that doesn’t mean that’s where the flow normally is."

Moreover, he notes, tumor vessels can be quite fragile, resulting in "red blood cells showing up in places where you wouldn’t expect them to be, including connective tissue."

At times the discussion between the proponents of vasculogenic mimicry and those who question its validity resembles an argument between spouses who circle round and round in agreeing the marriage has soured—but for completely different reasons.

In their February AJP commentary, Dr. McDonald and his colleagues stated they would not address the in vitro or microarray data presented in the Iowa team’s September AJP paper; instead, they focused on the histological, immunohistochemical, and electron microscope evidence. Explains Dr. McDonald: "Everything is 100 percent dependent on the primary data of tumors. There’s no point in going any further. That was the whole thrust of our commentary—not to judge the in vitro or microarray data, because you can interpret them one way or another depending on whether the first IHC observations are correct. If the first observations are not correct, it doesn’t make any difference what the rest says. So we didn’t go down that path."

Big mistake, counters Dr. Hendrix. "We would never have been able to reach the conclusions we did without the three-dimensional in vitro work and the molecular analysis. The in vitro work showed the tumor cells could create their own networks, and the molecular work told us these cells were capable of expressing the genes needed to undergo this vasculogenic mimicry."

Adds Dr. Folberg: "So far as we’re concerned, they’re setting up a straw argument."

Or take the matter of whether endothelial cells exist in the channels. "That’s not the principal focus of our work," says Dr. Folberg. "The fact channels appear not to be lined by endothelial cells is incidental."

Hardly, says Dr. McDonald. "It’s a very important question. It’s a fundamental part of angiogenesis."

Ahhh yes, angiogenesis. Like the pink elephant in the room, no one quite wants to acknowledge it, but eventually everyone does.

The Iowa team is careful to point out its work is not meant to be interpreted as undermining angiogenesis theory; attempts to view it as such, they say, stem largely from simplistic coverage in the lay press. Instead, they suggest their theory offers one of several alternative explanations for tumor growth where angiogenesis may not be actively occurring.

Separating the two theories is no mere intellectual exercise. One tenet of angiogenesis inhibition is that unlike chemotherapy, antiangiogenesis agents target nonmutating cells—including endothelium. If the vascular channels identified by the Iowa team do not contain endothelial cells, antiangiogenic agents targeting these cells may not be effective.

On the other hand, they might well be. If the tumor cells exhibit endothelial cell phenotypes, as the Iowa team suggests, "What’s to say they don’t have the same sensitivity as the endothelial cells to antiangiogenic therapy?" asks Dr. Ingber.

Dr. Folberg also notes that while angiogenesis is a physiologic response, occurring in events ranging from wound healing to ovulation, the development of the channels by tumor cells appears not to be. "Vasculogenic mimicry does not occur anywhere in the normal adult," he says. If vasculogenic mimicry does occur in tumors, this may open the door to therapies that would not interfere with phy-si-ologic events and could have fewer side effects than standard chemotherapies.

The angiogenesis link may have also forged a more antagonistic dialogue—as well as inspired more media coverage—than might normally characterize scientific debate. Antiangiogenesis research is on the developmental fast track these days, absorbing a tremendous amount of resources—time, money, effort—from all quarters. "Any alternative explanation could be perceived as threatening," suggests Dr. Platz. "There’s a lot at stake here, and scientists, like everyone else, don’t like to be wrong."

Closer to home, Dr. Platz offers other reasons why pathologists in particular may be resistant to considering the Iowa team’s theories.

Included in the panel of stains the researchers used in their search for endothelial cells was the notoriously demanding factor VIII-related antigen. "I hesitate to say some pathologists may not look at their factor VIII stains as critically as they should be, but that may be part of the issue," he suggests.

"I think another part of it may be that we’re taught one thing as residents, and then we go on seeing and interpreting things that same way," he continues. "If we suddenly have to say, ’Whoa!’ and question observations we’ve never questioned before, we may be reluctant to accept an alternate explanation simply because it doesn’t fit our preconceived notions."

For all the back and forth, the participants in this discussion agree on at least several points.

Everyone concedes the vascular channels need to link to normal blood vessels at some point to maintain a blood supply. The Iowa team says, in its September paper, that PAS-positive networks were traced in serial sections to connect with the vortex vein, which in turn drains the choroid; a reference (Eye. 1997;11: 227-238) containing a photograph of this anastomosis appears in both papers.

The illustration is not particularly persuasive, says Dr. McDonald, calling it "a wisp of connective tissue drifting up to a blood vessel."

But the Iowa team continues its investigations in this area. "We’ve taken pictures of anastomosis, and we’ve tried to reproduce it," says Dr. Maniotis. "But we don’t understand the process very well yet."

Quantitation is another area in need of data. "It’s not inconceivable that there are some vessels of the type Folberg and his colleagues have described," says Dr. McDonald. "But if they represent one percent of the total and a tiny fraction of blood flow, does that make any difference functionally? Ultimately the question is, What fraction of the total does this represent?"

Most also recognize that the national attention now being given to the Iowa team’s research has raised the discussion to another level.

"That means they have to clear a higher bar," observes Dr. McDonald.

"This matter has been debated in the clinical ophthalmology literature for several years, and now it’s being resurrected on a much different playing field," says Dr. Folberg. "In my view, this isn’t an issue of a higher level or a higher bar. People who haven’t shown an interest in this line of research previously are now focusing attention on it."

Has the Iowa team reached that standard? It depends on who you ask. Though the researchers acknowledge that many questions remain, they frame those unknowns in terms of clarifying the mechanism of vasculogenic mimicry. They’re clearly moving forward with new research and appear confident their findings are solid. "I think we have something new here," says Dr. Maniotis.

Others are less convinced.

"I don’t see any need for debate," says Dr. McDonald. "I see a need for more research in the laboratory. I think the questions that need to be answered are answerable; I just don’t think they’ve done it yet."

"What they’ve shown is intriguing, but we need more primary data," agrees Dr. Ingber. "What we need are more experiments. At this point, nothing constructive is going to come from more discussion."

Which seems like a good reason to end this story here.

Karen Titus is CAP TODAY contributing editor and co-managing editor.