An eggsample of why it is hard to prove a negative

Earlier this year, a paper claimed to have found cells, called ovarian stem cells (OSCs), in the adult ovaries of both mice and humans. These cells could apparently be grown in large numbers in the lab and could retain the ability to give rise to eggs. This paper met with, in my opinion, rather too much uncritical media coverage, with commentators making over-enthusiastic suggestions for OSC use in fertility treatment (1).

This study, from the lab of Professor Jonathan Tilly (2), was the latest to challenge the dogma that women (and other female mammals) have a limited supply of eggs that are generated during embryonic development, without the possibility of making new ones after birth.

However, this paper lacked robust evidence for some of the conclusions made, and alternative explanations were not given due weight. It is therefore perhaps unsurprising that this, together with previous work by Professor Tilly in the USA and that of another group in China led by Dr Ji Wu, has now been challenged.

A new study from Dr Hiroo Ueno and Dr Kui Liu's labs in, respectively, Japan and Sweden (3) makes use of elegant genetic techniques in the mouse, and finds no evidence for the existence of germ-line progenitors able to produce oocytes in postnatal ovaries. But is a lack of evidence sufficient to win the argument?

Before I address this, what does the new paper show? A quick recap - both Tilly and Wu made use of antibodies to isolate rare cells from adult ovaries. A small proportion of these cells could then be grown in vitro, and it is these cells that are claimed to be able to make eggs, either in vitro or after transplantation back into ovarian tissue in mice. However, neither lab provided evidence that this original rare cell population actually contributes to the oocyte pool in the adult ovary if left in place. Furthermore, it was unclear if the supposed OSCs were even cells of the germ-line rather than an in vitro artifact — that is, their properties are the result of the way they were isolated and cultured.

Moreover, there is a mystery surrounding the nature of the protein against which the antibodies were raised. Ddx4 is a protein thought from many other studies to be present only in the cytoplasm of germ-line cells and this fits with its function as an RNA-binding protein. But for cells to be isolated by an anti-Ddx4 antibody, the Ddx4 protein, or at least part of it, would have to be present on the surface of the cells. Still, Ddx4 should be an ideal marker, because its expression is known to be restricted to germ-line cells still able to undergo normal cell divisions (mitosis). It is not expressed in oocytes which have entered meiosis, the special type of cell division typical of oogenesis (and spermatogenesis).

The new study also uses Ddx4 as a marker of germ-line cells in the ovary, but rather than relying on antibodies to detect this protein, the researchers used the gene encoding it to drive the expression of Cre-recombinase. This is an enzyme that triggers recombination between specific DNA sequences, termed loxP sites.

The mice also carried a 'rainbow' multiple fluorescent reporter gene, which normally expresses green fluorescent protein (GFP), unless Cre-mediated recombination between several strategically located loxP sites triggers instead the activity of one of the other fluorescent marker proteins: red (RFP); orange (OFP); or cyan (CFP). Importantly, these would not just be active in dividing (or mitotically active) germ-line cells known to express Ddx4, but in all their descendants, including oocytes in meiosis.

The researchers used imaging techniques to see if any RFP- (or OFP- or CFP-) expressing cells in the postnatal ovary underwent mitosis in culture, which would be the expectation from Tilly's work. They failed to find any that did so, nor could they establish any stable RFP-positive cell cultures. Some GFP-expressing cell colonies did grow, but these seemed able to only give rise to follicular cell types, and not to any oocytes either in vitro or in vivo. They did observe some oocyte-like cells in long-term cultures of ovarian cells, similar to those described by Tilly's lab, but these expressed GFP, so were unlikely to have ever been Ddx4-Cre-expressing germ-line cells. Moreover, they soon degenerated.

Ueno and Liu's teams were therefore unable to replicate any of Tilly or Wu's data using techniques that seem to rely on the activity of the same gene, Ddx4. But is this proof that the latter are wrong? Well, even though I consider these new experiments to be clear and rigorous, the answer is not quite - neither on technical nor philosophical grounds.

First, it is possible that the activity of Ddx4-cre is not sufficient to trigger recombination in all Ddx4-expressing cell types. Alternatively, its prolonged activity beginning at embryonic stages could lead to some abnormal recombination events and to the loss of all fluorescent markers from adult Ddx4-expressing cells.

Second, excessive Cre-activity is known to affect function or survival of some cell types, so the strategy may have accidentally lead to all the OSCs being killed.

Third, it is conceivable that the antibodies used by Tilly and Wu do not in fact recognise any part of Ddx4, but a protein made by another gene that happens to share a similar structure (epitope). This would also explain its cell surface location, which does not make sense for Ddx4. It is therefore conceivable that 'OSCs' do not actually have a direct germ-line origin. In other words, they will have never expressed either Ddx4 or Ddx4-cre.

This is not a possibility recognised by Tilly, but it is not so far-fetched to think that OSCs could arise in culture by reprogramming from a somatic cell type. Against this, the recent study suggests that GFP-positive (Ddx4-negative) ovarian cells could not give rise to oocytes under any circumstances tested, in vivo or in vitro. However, the latter may not perfectly replicate those used by Tilly or Wu. Which brings us to the difficulty of proving a negative.

It is common to set up experiments to test a specific hypothesis by trying to disprove it. But science is rarely about certainty, and something could always be different between the conditions used by different scientists. Alternatively, there could be explanations that none of us have yet considered that could reconcile the different results. It is not just over (ova?) to Tilly to find a riposte, if he can, but for others to consider the latest data and to design additional tests.

The only certainty is that more experiments are required to build up evidence either for or against the existence of OSCs. Nonetheless, considering the unnecessary hype surrounding Tilly's work, which rather prematurely suggested cures for infertility, and the likely false expectations this engendered, I hope this latest work will at least provide some balance.