Einstein was celebrated for proclaiming that God is subtle, but not malicious. The proof that, despite His predilection for games of chance, God does not attempt to change the rules mid-game certainly would have delighted Einstein. Would it have convinced him of the validity of quantum mechanics? My intuition tells me not. ■

Seth Lloyd is in the Department of Mechanical Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139, USA.e-mail: slloyd@mit.edu

1. Zurek, W. H. Phys. Rev. A 76, 052110 (2007).2. Wootters, W. K. & Zurek, W. H. Nature 299, 802–803

(1982).

CANCER DIAGNOSTICS

One-stop shop Jonathan W. Uhr

Detecting cancer early and monitoring its progress non-invasively are high on oncologists’ agenda. So the design of a neat device that detects and counts cancer cells shed into the blood by tumours is a welcome advance.

A hallmark of carcinomas — malignant tumours of epithelial tissues such as breast and colon — is the invasion of neighbouring tissues at an early stage of tumour growth. This pro cess is associated with the shedding of tumour cells into the bloodstream. So counting and characterizing such circulating tumour cells to monitor the progress of the disease might be the most viable alternative to the unacceptable process of repeated, invasive biopsies. On page 1235 of this issue, Nagrath et al.1 report the development of an efficient and selective method for isolating these rare cells using microchip technology.

It might be considered too late to diagnose and characterize carcinomas if tumour cells have already entered the circulation, but this is not always the case. For example, although the presence of circulating tumour cells in patients with breast cancer usually indicates a poor prognosis, it does not necessarily indicate cancer metastasis. This is because only a small fraction of these cells have metastatic poten-tial2, and soon after entering the bloodstream, many tumour cells become committed to a suicide programme3. Also, circulating tumour cells are found in a significant proportion of breast-cancer patients long after such patients have undergone a mastectomy, despite being clinically disease-free and at a very low risk of their cancer recurring4.

Identifying and counting these cells can also be useful for monitoring carcinomas. For example, patients who have more than a certain level of circulating tumour cells have a poorer prognosis and so require more aggressive therapy5. Counting the cells after a course of treatment has been started can show whether it is effective, thereby allowing the treatment regimen to be modified much earlier than if physicians had to rely solely on changes in tumour size5. Furthermore, analysis of particular genes or proteins in tumour cells has revealed that they are continually altering genetically. So genetic changes in themselves

could be a reason for modifying a patient’s course of treatment6, and could be useful for selecting drugs that target particular cellular pathways involved in the malignancy. Indeed, such patient-specific approaches are deemed the therapies of the future.

The usual criteria for identifying circulating tumour cells are their microscopic appearance (they are larger than most white blood cells, with a large nucleus and little cytoplasm), a characteristic expression profile of the cyto-keratin protein, and the absence of a protein marker for normal blood cells. Although not definitive, these criteria are probably sufficient identification if there are large numbers of tumour cells in the patient’s blood.

Various techniques have been used to cap-ture these rare cells, and advances include the development of an automated instrument for counting them known as CellSearch, which is currently in clinical use7. The detection unit of this instrument uses metal particles coated with an antibody to the EpCAM protein, which is present on virtually all carcinoma cells8. A magnetic field is then used to capture antibody-bound tumour cells. However, such techniques generally detect circulating tumour cells in only 20–60% of patients with metastatic disease9.

The work of Nagrath et al.1 is the latest advance. The authors have developed a microfluidic (‘lab-on-a-chip’) device that can separate circulating tumour cells from whole-blood samples. The surface of this chip houses 78,000 microposts, each coated with antibod-ies to EpCAM. Fresh, unprocessed blood is pumped across the chip under controlled flow conditions to avoid damaging the fragile tumour cells. (This is a great improvement on conventional immunomagnetic enrichment procedures, which often damage the cells.) The cells bind to the microposts and are identified by a camera that can detect their morphology, viability and the presence of tumour markers on their surface (see Fig. 1 on page 1235). Dif-ferent tumour markers can be identified by staining the cells with specific antibody–dye conjugates.

The chip seems to be highly sensitive. The authors could detect circulating tumour cells in almost all of the patients they examined who had recurrent carcinomas, regardless of the tumour’s organ of origin, and in all of seven patients with early-stage tumours. Moreover, the purity of the cells obtained with this device was far higher

Red and white blood cells in an arteriole. The chip designed by Nagrath et al.1 can readily detect tumour cells among these.

DR

KES

SEL

& D

R K

ARD

ON

/TIS

SUES

& O

RGA

NS/

GET

TY

1168

NATURE|Vol 450|20/27 December 2007NEWS & VIEWS

0 50 100 150 200 250Time (million years)

Giant impact

LMO start

Earliest crust LMO final?

New

Old

New

Old

Core formation

Silicate differentiation

Earth

Moon

than can be achieved with other methods. But Nagrath and colleagues’ one-step method does have its limitations. For example, it cannot examine individual cells for genetic changes associated with progress of the tumour.

Nonetheless, Nagrath and colleagues’ results1 bring us closer to having a fully automated instrument that can detect circulating tumour cells with exquisite sensitivity. Such an instru-ment would allow routine monitoring of blood for tumour cells as part of a medical exami-nation, and could result in early detection and treatment, with the chance of obtaining higher cure rates. But this approach would also

necessitate minimizing false-positive results, particularly when only a few putative tumour cells are isolated. One way to solve this prob-lem would be to count sufficient numbers of chromosomes or genes in candidate cells to detect one of the hallmarks of a malignant cell — a significant deviation from the two-copies-per-cell norm. A simpler alternative might be to test for increased expression of the many known malignancy-associated proteins. In time, it might even be possible to detect and therapeutically target or remove the cells’ organ of origin at an early enough stage to prevent the cancer from metastasizing. ■

Jonathan W. Uhr is in the University of Texas Southwestern Medical Center, 6000 Harry Hines Boulevard, Dallas, Texas 75390-8576, USA.e-mail: jonathan.uhr@utsouthwestern.edu

1. Nagrath, S. et al. Nature 450, 1235–1239 (2007).2. Fidler, I. J. J. Natl Cancer Inst. 45, 773–782 (1970). 3. Méhes, G. et al. Am. J. Pathol. 159, 17–20 (2001). 4. Meng, S. et al. Clin. Cancer Res. 10, 8152–8162 (2004).5. Riethdorf, S. et al. Clin. Cancer Res. 13, 920–928 (2007).6. Meng, S. et al. Proc. Natl Acad. Sci. USA 101, 9393–9398

(2004). 7. Cristofanilli, M. et al. N. Engl. J. Med. 351, 781–791

(2004).8. Went, P. T. H. et al. Hum. Pathol. 35, 122–128 (2004).9. Allard, W. J. et al. Clin. Cancer Res. 10, 6897–6904 (2004).

PLANETARY SCIENCE

A younger MoonAlan Brandon

The most recent study of lunar rocks indicates that the Moon formed later than previously thought — a conclusion that requires our view of the early history of the inner Solar System to be revised.

It was almost 40 years ago that we found out what the Moon is made of: lunar samples col-lected during the Apollo missions turned out to consist of rocky material similar in compo-sition to that found on Earth. We also think we know that the Moon formed from the accre-tion of molten and vaporized ejecta produced by a collision between proto-Earth and a giant Mars-sized impactor1,2. But the answer to the ‘When?’ of Moon formation has remained elu-sive. On page 1206 of this issue, Touboul et al.3 examine this question with tungsten-isotope measurements of lunar rocks.

The mass of the giant impactor that led to the formation of the Moon was as much as 30% of Earth’s mass today, and this collision is thought to represent the last significant growth stage of our planet1. The post-collision disk of ejecta and the new bulk Earth were a combi-nation of proto-Earth material and that of the impactor. The tremendous release of energy probably melted the Earth, producing a glo-bal-scale magma ocean. Iron-rich metal from the proto-Earth and the impactor merged to form the present Earth’s core, with Earth’s mantle and crust forming from silicate and oxide (non-metal) material. The time intervals between the various events — the giant impact, the formation of Earth’s core, and the onset and solidification of magma oceans in Earth and the Moon — tell us when accretion was completed and constrain the early differentia-tion and cooling histories of the ‘terrestrial’ planets (Mercury, Venus, the Earth–Moon system and Mars).

Enter the evidence from the 182Hf (hafnium) and 182W (tungsten) radiometric clock. The isotope 182Hf decays to 182W with a half-life of 9 million years. This clock dates events that

occurred in the first 60 million years or so of Solar System history, before most 182Hf decayed away. It is useful for tracking metal–silicate dif-ferentiation times, because W has an affinity for iron-rich metal cores whereas Hf favours mantle silicate and oxide minerals. If core for-mation occurred when 182Hf was still actively decaying, then a record is preserved in the amount of 182W present.

The amount of 182W measured in Earth samples4,5 indicates that W was removed into Earth’s core about 30 million years after the Solar System began to accrete 4.567 billion years ago (Fig. 1). Before Touboul and col-leagues’ new study3, the lunar 182W data had yielded two conclusions. First, that formation of the Moon — and so the giant impact event — occurred around the same time as Earth’s core formed at about 30 million years4,5. Second, that the lunar magma ocean largely solidified 10 million years after that6. This chronology provides a well-ordered Earth–Moon system that behaved exactly as we expected — rapid accretion of Earth, core formation and global-scale melting were coincident with a giant impact, and were followed by rapid cooling and consequent solidification of the magma oceans of Earth and the Moon.

But some observations do not fit this simple lunar history (Fig. 1). First, radiometric ages for the lunar crust, which was made from light minerals that solidified and floated to the top of the magma ocean, show that it began to form no earlier than 70 million to 150 million years after accretion of the Solar System began7. Sec-ond, the 146Sm (samarium) and 142Nd (neodym-ium) radiometric clock shows that the lunar magma ocean largely began to solidify as late as 215 million years8. Third, as the new lunar data

Figure 1 | The earliest history of the Moon and Earth. Time 0 marks the onset of Solar System accretion at 4.567 billion years ago, with the rest of the scale denoting time after that event. The previously inferred ages for the giant impact and the onset of solidification of a lunar magma ocean (LMO)4–6 were based on older W-isotope data for Moon rocks. Touboul and colleagues’ W-isotope data3 provide a new and later time for the giant impact and for the start of solidification of the LMO. The final solidification time of the LMO is obtained from the 146Sm–142Nd chronology of Moon rocks8. The vertical band is the earliest time for the giant impact that formed the Moon; error bars represent 2σ uncertainty estimates for each event. Data points for events on Earth are given for comparison.

of Touboul et al.3 show, the previous W isotope data for lunar materials were not completely corrected for 182W produced by the decay of 182Ta (tantalum). The isotope 182Ta forms when cosmic rays bombard the Moon. The excess 182W from this process can result in spurious ages using the 182Hf–182W clock.

When correctly accounting for 182W pro-duced from 182Ta, Touboul et al.3 find that the amount of 182W in lunar samples indicates that the Moon could not have formed before 62+

− 9100 million years after the initiation of Solar

System accretion, at least 16 million years after Earth’s core formed and no later than forma-tion of the first lunar crust (Fig. 1). These new findings are in agreement with the slower rate of magma-ocean solidification obtained from ages for the lunar crust and 146Sm–142Nd data.

1169

NATURE|Vol 450|20/27 December 2007 NEWS & VIEWS