Nature © Macmillan Publishers Ltd 1998

8The Amazon basin harbours the most

diverse assemblage of freshwater fishesin the world, including a disproportionatelylarge number of marine-derived groups,such as stingrays, flatfishes, pufferfishes, andanchovies1. On the basis of our molecularphylogenetic analysis of South Americanfreshwater stingrays (Potamotrygonidae),coupled with reconstructions of Amazonianpalaeogeography, we propose that somemarine-derived freshwater fish species origi-nated as a by-product of massive move-ments of marine waters into the upperAmazon region during the Early Mioceneepoch, 15–23 million years ago.

Miocene South America experiencedprofound changes of topography, environ-ment and river drainage patterns2,3. A com-bination of sea-level changes and tectonicloading of the foreland basin in the upperAmazon produced significant ingressions ofsea water, as indicated by the presence ofmarine and brackish mollusc, copepod andmangrove fossils4,5. Nuttall4 says the upperAmazon during the Miocene was “up to 500km wide, occupied by a continually shiftingpattern of streams, swamps, and lakes ofvarying salinity and offering intermittentconnections with the Caribbean”. Theseconditions would have been ideal for theisolation of marine fishes in progressivelydesalinized habitats. We suggest, therefore:first, that the divergence between pota-motrygonid stingrays and their closestmarine relative should have occurred dur-ing the Miocene; and second, that the dis-tribution of the closest marine relativeshould include the Caribbean, a proposedsource of the marine incursion.

To test the marine-incursion hypothesis,we investigated the evolutionary history ofstingrays using DNA sequences from the geneencoding mitochondrial cytochrome b (Fig.1a). We analysed sequences from ten SouthAmerican freshwater stingrays, includingrepresentatives of all three potamotrygonidgenera, and from all the supposed marine relatives of potamotrygonids. We calibrated a rate for cytochrome b evolution in pota-motrygonids based on the amount ofsequence divergence between species sepa-rated by a known geological event, the upliftof both the eastern cordillera of the Andesand the Merida Andes, which isolated theMaracaibo stingray (Potamotrygon yepezi)approximately 8 million years ago2,3.

Based on our cytochrome b calibrationand phylogenetic tree, the estimated diver-gence time between potamotrygonids andtheir closest marine relative indicates thatneotropical freshwater stingrays originatedin the Early Miocene. Moreover, one of thetwo species that are the marine relatives of

even though South American rivers havebeen open to the oceans for more than 100million years, endemic freshwater stingrayshave a single, unique origin. We believe thatthis pattern argues against invasions bymarine rays, and suggests instead that a sin-gle geological event, such as a marine incur-sion, was crucial.

It has also been proposed that uplift ofthe Andes blocked a previously westward-draining proto-Amazon, and isolated acomponent of the Pacific marine fauna inprogressively desalinized inland lakes orseas. This would mean a Late Cretaceous orPalaeocene origin for Potamotrygonidae, asthis is when Andean uplift probably severedthe last connection between the PacificOcean and the upper Amazon9. But we esti-mate that potamotrygonids originatedmore recently, and the marine relative of

NATURE | VOL 396 | 3 DECEMBER 1998 | www.nature.com 421

scientific correspondence

the freshwater rays is distributed along thenorthern coast of South America (Fig. 1b).Both the estimated origination time of thefreshwater species and the biogeography ofthe closest marine relative are therefore consistent with predictions of the marine-incursion hypothesis.

Alternative explanations have beenadvanced to account for the presence ofmarine-derived taxa in South America. Forinstance, it has been suggested1 that theextensive interface between marine andfreshwater habitats in the lower Amazonserved as a gateway for marine invaders ofneotropical rivers. Such invasion hypothesesare difficult to refute because they are com-patible with many different biogeographicalpatterns. Invasions are essentially oppor-tunistic, and we might expect fresh water tohave been colonized several times; however,

Marine incursion into South America

}

PliMyr10

20

30

40

50

60

70

80

90

*Mio

Oli

Eoc

Pal

Cre

* = first Potamotrygon fossils

Africa and South America separate

Birth of the Andes (ongoingAndeandevelopment for next 90 Myr)

Marine incursions

North-flowing palaeo-Amazonas-Orinoco’ drainage established

Marine incursions

Marine incursions

Marine incursions

Parana captures headwaters of palaeo-Amazonas-Orinoco’

Maracaibo isolated, Amazon andOrinoco attain present configurations

South Americanpalaeogeological events

Freshwater

Himantura pacifica (Paci

fic)

Himantura schmardae (Carib

bean)

Paratrygon aiereba (Amazo

n)

Potamotrygon castexi (Amazo

n)

Potamotrygon motoro (Orino

co)

Potamotrygon yepezi (Mara

caibo)

Pleisotrygon iwamae (A

mazon)

Potamotrygon orbigyni (Orino

co)

H. schmardae

H. pacifica

a

b}

}

}

’

’

FFiigguurree 11 Phylogeny and biogeography of freshwater stingrays. a, Phylogenetic hypothesis for freshwaterstingrays and their closest marine relative, and timescale showing palaeogeological events in South Ameri-ca2,3. Relationships of Potamotrygonidae, amphi-American Himantura and seven other stingray genera wereinferred from parsimony analysis of 765 base pairs of cytochrome b sequence, with transitions at thirdcodon positions weighted one-fifth of all other changes10 (not all taxa shown; sequences available fromGenbank). The potamotrygonid species are a monophyletic group, so neotropical freshwater rays had a sin-gle origin. H. pacifica and H. schmardae were supported as the closest sister taxa to potamotrygonids, con-sistent with morphological analysis11. Ages of nodes were estimated from transversion distances at thirdpositions to minimize effects of saturation and selection. The estimated age of Potamotrygon agrees withputative fossil evidence for this taxon12. b, Distribution of H. pacifica and H. schmardae (black areas alongcoastlines), the closest marine relatives of the freshwater rays. The group is found in the Caribbean (the pro-posed source of Miocene marine incursions) but not the mouth of the Amazon (red arrow) or along thePacific coast of South America (other proposed points of colonization).

Nature © Macmillan Publishers Ltd 1998

8

freshwater stingrays does not occur alongthe Pacific coast of South America (Fig. 1b).

A diverse cross-section of the South Amer-ican river fauna, including dolphins, fishes,crabs and snails, appear to be derived frommarine ancestors. If these taxa originated contemporaneously with freshwater sting-rays, the Miocene marine transgressions ofSouth America will have had a profound effecton the diversification and structuring ofneotropical communities.Nathan R. Lovejoy*, Eldredge Bermingham†, Andrew P. Martin‡*Section of Ecology and Systematics, Corson Hall, Cornell University, Ithaca, New York 14853-2701, USA†Smithsonian Tropical Research Institute,Naos Marine Laboratory, PO Box 2072, Balboa, Republic of Panama‡Department of EPOB, CB 334, University of Colorado, Boulder, Colorado 80309, USAe-mail: am@stripe.colorado.edu

1. Roberts, T. R. Bull. Mus. Comp. Zool. 143, 117–147 (1972).

2. Hoorn, C. et al. Geology 23, 237–240 (1995).

3. Lundberg, J. G. et al. in Phylogeny and Classification of

Neotropical Fishes (eds Malabarba, L. R., Reis, R. E., Vari, R. P.,

Lucena, C. A. S. & Lucena, Z. M. S.) (Museu de Ciências e

Tecnologia, Porto Alegre, Brazil, in the press).

4. Nuttall, C. P. Bull. Br. Mus. Nat. Hist. 45, 165–371 (1990).

5. Hoorn, C. Palaeogeogr. Palaeoclimatol. Palaeoecol. 105, 267–309

(1993).

6. Croizat, L. Space, Time, Form: The Biological Synthesis

(Published by the author, Caracas, Venezuela, 1962).

7. Brooks, D. R., Thorson, T. B. & Mayes, M. A. in Advances in

Cladistics (eds Funk, V. A. & Brooks, D. R.) 147–175 (New York

Botanical Garden, 1981).

8. Grabert, H. Natur Mus. 113, 61–71 (1983).

9. Kroonenberg, S. B., Bakker, J. G. M. & van der Wiel, A. M. Geol.

Mijnb. 69, 279–290 (1990).

10.Swofford, D. L. PAUP: Phylogenetic Analysis Using Parsimony,

version 3.1 (Illinois Natural History Survey, Champaign, 1993).

11.Lovejoy, N. R. Zool. J. Linn. Soc. 117, 207–257 (1996).

12.Lundberg, J. G. in Phylogeny and Classification of Neotropical

Fishes (eds Malabarba, L. R., Reis, R. E., Vari, R. P., Lucena,

C. A. S. & Lucena Z. M. S.) (Museu de Ciências e Tecnologia,

Porto Alegre, Brazil, in the press).

scientific correspondence

422 NATURE | VOL 396 | 3 DECEMBER 1998 | www.nature.com

(a-factor and a-factor, respectively), whichare detected by cells of the opposite matingtype, triggering a series of events that lead tomating. In particular, the cell responseincludes directed growth towards the matingpartner, the position of which is estimatedfrom the direction of maximum pheromoneconcentration1–3. Because yeast cells do notmove, an appropriate choice of growth direc-tion is essential for efficient mating.

Cells of mating type a secrete a proteasethat hydrolyses a-factor4,5. Mutants defi-cient in this ‘barrier activity’ (bar1mutants6,7) are highly sensitive to phero-mones7, so the protease was thought to beinvolved in the recovery of yeast cells fromthe pheromone-induced cell-cycle arrestthat is part of the pre-mating response. Butbar1 mutants also mate less efficiently witha-cells in a mass mating mixture1. A role forthe protease in mating-partner detectionhas also been suggested4.

To understand how the protease mayincrease mating efficiency, consider an a-cell surrounded by several a-cells, whichsecrete a-factor (Fig. 1a,b). These cells arepart of a larger population of a- and a-cells.The steady-state profile of pheromone con-centration between cells, [P(r)], where r

represents a position in three-dimensionalspace, is given by solving the diffusionequation ![P(r)]/!t 4D?2[P(r)]40, whereD is the diffusion coefficient.

It is useful to consider an analogy withan electrostatic system, as the electrostaticpotential satisfies the above equation. Thea-cells can be thought of as electrostaticpoint charges, the pheromone concentra-tion as an electrostatic potential, and thepheromone gradient as an electric field. Thepheromone profile is given by:

[P(r)]4/i =1N f (r1ri), f(ärä)41/ärä

where f(r1ri) is the contribution of theith cell, located at ri, to the concentration.Because 1/r is a slowly decaying functionof the radial distance, r, the pheromoneconcentration at any point is determinedby the contribution of many a-cells. Ingeneral, therefore, the pheromone gradientat the location of an a-cell will not point inthe direction of the nearest mating part-ner, but at an angle, u, from it (Fig. 1a).

But what is the role of the protease?Consider first-order degradation:![P(r)]/!t 4D?2[P(r)]1k[B][P(r)]40

where [B] is the concentration of the barprotease, and k is a kinetic rate constant. Interms of the electrostatic analogy, the new

θ<20°grad>1%

bar++

bar+

bar —

0 2 4 6 8 10

1

0.9

0.8

0.7

0.6

0.5

1

0.8

0.6

0.4

0.2

0

0 50 100

Gradient strength

Detection accuracy

Inverse screening length

Ave

rage

det

ectio

n ac

cura

cy

Ave

rage

gra

dien

t st

reng

th

a

b

c

FFiigguurree 11 Protease secretion improves detection of thenearest mating partner. a, b, An aa-cell is surroundedby a-cells, which secrete a-factor matingpheromones. The steady-state pheromone concen-tration ranges from high (red) to low (dark blue). Theaa-cell grows up the pheromone gradient. a, In theabsence of protease, the pheromone gradient isdetermined by many a-cells, leading to a failure tolocate the nearest mating partner: the angle betweenthe direction to the nearest mating cell and the direc-tion of the pheromone gradient (and thus the direc-tion of growth), u, is not zero. b, Widely diffusedprotease hydrolyses the a-factor, limiting the range ofpheromone diffusion and improving the detection ofthe nearest mating partner. c, Average detectionaccuracy <cos u>, and the average magnitude of thepheromone gradient, <ä?[P]ä>, as functions of theinverse screening length, 1/l. We placed 10,000 cellsrandomly in a three-dimensional cube and calculatedthe pheromone gradient at the centre. The screeninglength, l, was normalized by the mean distancebetween cells; the average strength of the gradientwas normalized by its maximum (obtained for zeroscreening, with l4÷); θ is the angle between thedirection of the gradient and the direction to the cellnearest the centre. Detection accuracy increasesrapidly for l<1. Minimum accuracy occurs at l4÷,which corresponds to an error of u450º. Inset, effectof the bar protease on the detection of the nearestmating partner. Blue columns show percentage ofcells exposed to a pheromone gradient with normal-ized values of more than 1%. Green columns showthe probability that this gradient will point to within 20º

of the nearest a-cell. Strain names are arbitrary: bar1

indicates cells that do not secrete protease; bar&,protease-secreting cells with 1/l44; and bar&&, cellsthat secrete a higher level of protease, with 1/l410.

Protease helps yeastfind mating partners

The choice of mating partner by the yeastSaccharomyces cerevisiae involves the detec-tion of mating pheromones produced byother yeast cells. A cell that is capable ofmating deduces the position of its nearestmating partner from the spatial gradient ofpheromone. While studying this process,we realized that, in the presence of manypotential mating partners, the gradientmight not point in the direction of thenearest partner. Here we show that degra-dation of some of the mating pheromoneby protease enzymes helps to align the gra-dient in the direction of the nearest partner,which increases mating efficiency.

Haploid yeast cells of the two matingtypes, a and a, signal to each other by thesecretion of cell-type-specific pheromones