For Olfaction, a Hypothesis is Felled

Peter MombaertsCourtesy of Peter MombaertsRarely do scientific studies claim that something is not the case. Rarer still do negative results appear in top-tier journals. Yet two recent papers in Nature describe what olfactory sensory neurons do not do.12The olfactory system is often compared to the immune system. The key cell types in each system, the olfactory sensory neuron (OSN) and the lymphocyte, can detect a wide variety of chemicals, both natural and synthetic. Chemical recognition by the

Peter Mombaerts
Oct 24, 2004
<p>Peter Mombaerts</p>

Courtesy of Peter Mombaerts

Rarely do scientific studies claim that something is not the case. Rarer still do negative results appear in top-tier journals. Yet two recent papers in Nature describe what olfactory sensory neurons do not do.12

The olfactory system is often compared to the immune system. The key cell types in each system, the olfactory sensory neuron (OSN) and the lymphocyte, can detect a wide variety of chemicals, both natural and synthetic. Chemical recognition by these cells is specific and sensitive, so many odorant receptors (ORs) 3 and antigen receptors are needed to cover the spectrum of chemical structures.

It has been well established that each B or T lymphocyte expresses a single antigen receptor – an antibody or T-cell receptor. Analogously, each vertebrate OSN is thought to express a single OR gene. Lymphocytes restrict antigen receptor gene expression by genome modification, and many believed...


With ~1,200 in mouse and ~1,500 in rat, OR genes make up the largest family in mammalian genomes.1 From the time of their discovery,4 it was clear that the coding region of an OR gene comes in one piece and does not undergo sequence diversification in OSNs. By contrast, antibody and T-cell receptor genes come in pieces that are shuffled, combined, and modified during lymphocyte differentiation. These sequence alterations generate a repertoire of millions of antigen receptor variants. DNA rearrangements in antigen receptor genes have a dual purpose: to produce a wealth of new receptor variants, and to restrict expression to one receptor variant per cell. If the OSN genome were modified to restrict expression to one OR gene per cell, only the noncoding sequences of OR genes would be affected.

But determining whether or not this is the case proved difficult. What makes lymphocytes so much easier to study than neurons is their ability to propagate in culture, as normal or tumor cells. Clonal expansion reflects the normal behavior of B and T lymphocytes in the body upon stimulation, but neurons do not divide when stimulated and (with a few exceptions) do not make tumors. A mature neuron is postmitotic, and as a consequence few cell lines with mature neuronal properties are available. OSNs are no different.

Genomic analysis of single OSNs as a means to search for DNA rearrangements would always be regarded as inconclusive and unconvincing. So we had to devise another way to clone the genome of a single OSN. In 1999, I proposed to test the hypothesis of DNA rearrangements at OR loci by nuclear transfer from OSNs.5


Ideally cloning ought to be done with OSNs expressing a predefined OR. We had developed a gene-targeting technique to tag such cells by expression of the green fluorescent protein (GFP).6 But OSNs expressing a given OR are exceedingly rare: as few as 1 in 10,000 cells from dissociated preparations of olfactory mucosa. If no sequence alterations were found, skeptics would argue that there was no proof that nuclear transfer was carried out with those rare, green-fluorescent cells. Indeed, accidental nuclear transfer with any other cell in the preparation would give the same, negative result. Similar doubts about the identity of the occasional cell that gives rise to a cloned animal has troubled the cloning literature ever since the controversy about Dolly's origin.

We reasoned that we did not need cloned mice, but that a cell line carrying the genome of a particular OSN would be sufficient. We thus worked out a method of cloning the genome of a single cell by deriving an embryonic stem (ES) cell line from an embryo generated by nuclear transfer (nt).7

To demonstrate retrospectively that the rare cell that gave rise to an ntES line expressed a predefined OR, we took a lineage-tracing approach using 'reporter' transgenic mice. When the predefined OR is chosen for expression, the Cre-recombinase is expressed concomitantly; this enzyme catalyzes the excision of a DNA sequence that inhibits GFP expression from the transgenic reporter. If an ntES line has the specific excision and expresses GFP, its genome must have come from an OSN that expressed the predefined OR.

To cut a long story short: We did not find evidence for sequence alterations at the locus encoding M71, the OR gene that we had selected. In a complimentary set of experiments, we injected the ntES cells into wild-type embryos, thereby generating chimeric mice. We found that OSNs that differentiated from the injected ntES cells (as recognized by GFP expression) did not invariably express M71, but could express other ORs.

The labs of Rudolf Jaenisch, Richard Axel, and Andrew Chess came to a similarly negative conclusion.2 They chose OSNs that may have been expressing the P2 OR at the time of the nuclear transfer, and they provided evidence that cells that were likely at the origin of the ntES lines were postmitotic.



Courtesy Peter Mombaerts

Harvey, the mouse, was created when nuclear transfer embryonic stem (ntES) cells carrying a genome cloned from an OSN that expressed the M71 OR gene were injected into a tetraploid blastocyst. The ntES cells outcompete the tetraploid cells, and the ensuing chimeric mouse is nearly entirely derived from ntES cells. Harvey may harbor a minor contamination of tetraploid cells, however, and is therefore considered a 'clonal' mouse.

Publication of these results is remarkable because they describe negative evidence. Scientific journals, particularly top-tier journals, want research articles to provide positive evidence for a hypothesis. Indeed, failure to obtain positive evidence for the tested hypothesis may be due to a failure of the test itself: trivial errors, technical inadequacies, or flawed design. But it is worthwhile to report negative evidence because it stimulates alternative thinking and can free scientists from misguided assumptions. During the course of this project, from 1999 to 2004, our experimental strategy was dominated by the concern that, after all, the sequence of OR loci is not modified in OSNs. Both Nature papers are a series of control experiments, and controls for control experiments, in an attempt to rule out that the favorite hypothesis, alas, is not correct.

But the experiments were carried out with only two of 1,200 OR genes. Perhaps we and our colleagues were unlucky, and selected OR genes that are not representative and do not undergo modifications. Perhaps nuclear transfer reverted the sequence modifications back to the germline configuration: The extensive nuclear reprogramming that occurs in the oocyte upon nuclear transfer may also erase the sequence alteration at the OR locus. Perhaps ... perhaps ...


Nonetheless, I believe that the hypothesis of DNA rearrangements in OR loci, which was already raised in the 1991 paper describing the discovery of OR genes,4 has been laid to rest with a sufficient degree of certainty. This was a pet hypothesis for many scientists working on this problem, including myself (many of us were immunologists once upon a time). OSNs and lymphocytes have solved the issue of gene choice in radically different ways. As to alternative hypotheses, I perceive a deafening silence. The enigma of OR gene choice appears to have receded even further.

But how sure are we that each OSN expresses just one OR gene? Perhaps the hypothesis of DNA rearrangements has been inspired erroneously by the one receptor-one neuron rule. Could evolution really come up with a mechanism that flawlessly chooses just one OR gene in each OSN during differentiation? I find such a perfect scenario unlikely. Elsewhere I have analyzed the available evidence and concluded that the one receptor-one neuron rule is far from being proven.8 Initially a hypothesis, it evolved to dogma, skipping the stage of theory. I suggest that conceiving and testing models of OR gene choice may be more productive if the one receptor-one neuron rule is not taken too strictly.

One possible scenario: During OSN differentiation a low probability event may underlie the choice of (often) zero, (usually) one, but (sometimes) a few OR genes. Cells that express no or more than one OR would be rejected by selective processes. Shaping a repertoire of cells by selecting the useful and rejecting the undesired members is well described for lymphocytes, but is not sufficiently considered for OSNs. The outcome of the process – the expression of one OR per mature OSN – must not be confused with the process itself: choice of OR gene(s) in an immature OSN.

Peter Mombaerts is a professor at the Rockefeller University in New York. He studies odorant receptor gene choice and axonal wiring of olfactory sensory neurons in the mouse, using gene targeting, transgenesis, and cloning by nuclear transfer. Mombaerts did his PhD thesis on antigen receptor gene rearrangements with Nobel laureate Susumu Tonegawa, and a postdoc on odorant receptor genes with Richard Axel, who won the 2004 Nobel Prize.

He can be contacted at

Interested in reading more?

Become a Member of

Receive full access to digital editions of The Scientist, as well as TS Digest, feature stories, more than 35 years of archives, and much more!
Already a member?