Though all cells within an organism carry the same set of genes, they display a wide range of characteristics, depending on where they are in the body and where the organism is in its life cycle. Howard Chang is interested in how large-scale gene regulatory programs control a cell's behavior in space and time.

A practicing dermatologist, Chang studies many of these questions in the skin. After investigating apoptotic signaling mechanisms for his PhD with David Baltimore at MIT (1), Chang joined Pat Brown's laboratory for his postdoc, where he discovered that dermal fibroblasts have markedly different gene expression patterns depending on their anatomical position (2). In his own lab at Stanford University, Chang has found that this positional identity partly relies upon long, noncoding RNAs that direct chromatin-modifying proteins to control the expression of HOX genes (3, 4). Inappropriate expression of these RNAs can lead to diseases such as breast cancer (5). Chang's lab has also used gene expression profiling to reveal some of the molecular mechanisms underlying tumorigenesis (6) and aging (7).

In a recent interview, Chang discussed how he arrived at his current position and the future program of his research.

EARLY IDENTITY

Where did you grow up?

I was born in Taipei, Taiwan, but my family moved to southern California when I was 12. During high school, my biology teacher helped me get involved with a research project on organ transplantation at the local university, UC Irvine, and I became really interested in how immunosuppressant drugs work. So from fairly early on, I was pretty sure that I wanted to be involved in research. But I was always interested in questions that relate to human health, so an MD-PhD program seemed like the best plan for me.

“[This is] broadening the repertoire of how lncRNAs can regulate genes.”

Why did you choose David Baltimore's lab for your PhD?

The Baltimore lab was a really exciting place to be because people were studying lots of different questions using a wide variety of techniques. The philosophy there was to make a multipronged attack on whatever biological problem people wanted to solve. I was really attracted to that style. At the time, the mechanisms of programmed cell death were just beginning to be deciphered. So I got involved in a project to understand how certain cell surface “death” receptors could trigger apoptosis.

After finishing your MD, why did you join Pat Brown's lab for your postdoc?

Pat was a pioneer of microarray technology and functional genomics. That was very different from what I'd done before. But I'd read one of his early papers in a journal club class during graduate school, and I thought it was a really exciting technology. I really appreciate the fact that Pat took me into his lab, because I wanted to do my clinical training and my postdoc at the same time. He wasn't fazed when I told him that I'd be working very irregular hours or that he might not see me for months at a time.

One of the early questions we thought about was whether we could investigate the mechanisms of aging by looking at gene expression patterns in tissues from people of different ages. Because I was training as a dermatology resident, I decided to do this in skin. In the literature, a frequent source of primary human fibroblasts is neonatal foreskin, whereas cells from older persons are usually from the arm or somewhere other than the foreskin. So I thought I should do a control experiment to show that the anatomic site didn't matter, and then I could go on to compare the different ages. I arranged to get fibroblasts from different anatomical sites of the same person and the results were the exact opposite of what I was expecting—fibroblasts from different body sites showed different gene expression patterns. It was so interesting that working on the positional identity of cells within tissues quickly became the main focus of my research.

POSITIONING SYSTEMS

How do cells know where they are in the body, and why investigate this question in skin?

Skin is a fantastic system because, in many animals, including humans, there are lots of site-specific features that have important functions. For example, many of us have long hairs on our scalps but no hair on our palms and soles. Skin is also constantly regenerating—in humans, the epidermis turns over every 28 days—but the new cells carry information about their location in the body. We believe that much of this positional information can be traced back to the embryonic patterning mechanisms—the HOX genes—which tell cells where they are along the different axes of the body. Although primary human fibroblasts look very generic, we've found that they are actually quite different, depending on where they come from in the body. These differences are strongly associated with the specific HOX genes that they continue to express, helping the cells to maintain a positional memory over time.

How did you end up working on noncoding RNAs?

HOX genes are organized into clusters of nine to eleven genes in a row. We wanted to examine these loci at high resolution, looking at all the expressed RNAs and at the different chromatin states associated with this expression, so we built a device called a tiling array. We were surprised to notice that, in addition to the HOX genes themselves, there were a lot of additional intergenic transcripts, many of which turned out to be what people now call long noncoding RNAs, or lncRNAs. They're actually the dominant transcriptional output of the HOX loci, and they also show very nice position-specific patterns of expression.

That led us to ask what the function of these RNAs might be. lncRNAs involved in dosage compensation or genomic imprinting tend to silence nearby genes. We depleted a HOX lncRNA called HOTAIR and saw that certain HOX genes were derepressed, but those genes were actually on other chromosomes. More recently, we've looked at an RNA called HOTTIP. This lncRNA creates a chromatin state that activates gene expression. So these examples are broadening the repertoire of how lncRNAs can regulate genes.

How do these lncRNAs contribute to human disease?

That's something we're really interested in. For example, we found that HOTAIR silences HOX genes by targeting the chromatin-remodeling Polycomb complex to them. Overexpression of Polycomb complex proteins can drive cancer progression, so we wondered whether HOTAIR itself could also be involved in tumorigenesis. Several years ago we looked at HOX noncoding RNAs in human cancer samples and found that HOTAIR is very highly expressed in breast cancers that go on to metastasize or kill the patient. HOTAIR is a powerful predictor of clinical outcome, and it actually drives cancer progression. So a project that started with a basic curiosity about skin patterning may have some important medical implications. More recently, we found a family of cell cycle–regulated lncRNAs, which are also dysregulated in human cancers.

GROUP DYNAMICS

You've done a lot of gene expression profiling on tumor samples. How does that approach aid cancer research?

Many people in the field use expression profiling to create molecular snapshots of cancer or other disease processes. When I started my lab, I wanted to take these descriptions and turn them into an understanding of causality. If hundreds of genes are changing, are there key individual switches that cause those genes to go up or down?

We developed computational methods and other strategies to define factors coordinating the activities of large groups of genes. That information can be valuable in several ways. It provides insight into disease mechanisms, and it's possible that you can track these regulators as markers of the disease rather than having to measure the activity of hundreds of genes. More importantly, maybe we can manipulate these master switches to create better disease models or even block the disease process.

“In the clinic, one sees many phenomena that trigger interesting questions.”

You've also returned to the study of aging. What have you found out?

We started by examining changes in gene expression patterns that occur with aging and then looked for regulators that might be causing these changes. We realized that aging is associated with overactivity of the stress-inducible transcription factor NF-kB. We removed NF-kB from the skin of old mice and saw something very interesting: the skin cells started behaving as if they were young. As I mentioned before, the skin is constantly turning over, so it's always been curious that new skin cells born in an old person behave as if they're old. So we think that NF-kB is part of an active mechanism that tells cells to behave “old.” This project got a lot of attention in the lay press!

What is your lab working on now?

The questions we're working on have the broad theme of how large groups of genes work together to carry out a biological program. One current focus is on lncRNAs. Work from many groups, including our own, suggests that there are thousands of these long RNAs in the genome, but how they perform their functions is still mysterious—it remains to be seen if there are any systematic rules.

The biological contexts of our studies remain cancer, aging, and lineage identity. We're currently collaborating with Marius Wernig's group to look at what happens when fibroblasts are reprogrammed into other cell types. We started with the question of how fibroblasts maintain positional information over time. Studying reprogramming is an interesting way to probe the same question: what aspects of positional memory have to be overcome to change cell fate?

How does being a practicing physician influence your research?

In the clinic, one sees many phenomena that trigger interesting questions. Even my original postdoc project was partly inspired by what I was learning in my clinical training. In dermatology, where a lesion occurs on the body makes a huge difference, but I realized that the molecular basis of this wasn't understood. So I felt that investigating whether fibroblasts from different locations were identical was an important problem. I think these kinds of opportunities will continue to arise in the future.

References

References
1.
Chang
H.Y.
et al
.
1998
.
Science.
281
:
1860
1863
.
2.
Chang
H.Y.
et al
.
2002
.
Proc. Natl. Acad. Sci. USA.
99
:
12877
12882
.
3.
Rinn
J.L.
et al
.
2007
.
Cell.
129
:
1311
1323
.
4.
Wang
K.C.
et al
.
2011
.
Nature.
472
:
120
124
.
5.
Gupta
R.A.
et al
.
2010
.
Nature.
464
:
1071
1076
.
6.
Adler
A.S.
et al
.
2006
.
Nat. Genet.
38
:
421
430
.
7.
Adler
A.S.
et al
.
2007
.
Genes Dev.
21
:
3244
3257
.

Author notes

Text and Interview by Ben Short