letters to nature

Smart Race cDNA amplification kit (Clontech). Expressed sequence tags were amplified 29. Koopman, W. J. H. et al. Membrane-initiated Ca2þ signals are reshaped during propagation to
by PCR with the universal adapter primer provided with the kit and the various, specific subcellular regions. Biophys. J. 81, 57–65 (2001).
internal primers. 30. Curtis, E. A. & Landweber, L. F. Evolution of gene scrambling in ciliate micronuclear genes. Ann. NY
Acad. Sci. 870, 349–350 (1999).
Complete macronuclear gene-sized chromosomes
Telomere-specific primers in combination with internal gene sequences allow a Supplementary Information accompanies the paper on www.nature.com/nature.
straightforward recovery of the complete gene30. The specific (internal) primers were
based on the DNA sequences of internal fragments of the various genes, which were Acknowledgements We thank L. Landweber, J. Wong and W.-J. Chang for advice on the cloning
recovered previously by PCR with degenerated primers for conserved parts of the various of complete minichromosomes and for sharing the first sequence of a PDH gene in N. ovalis; S. van
genes. Weelden and H. de Roock for help in the metabolic studies; J. Brouwers for analysis of the
quinones; G. Cremers, L. de Brouwer, A. Ederveen, A. Grootemaat, M. Hachmang, S. Huver,
Phylogenetic analysis S. Jannink, N. Jansse, R. Janssen, M. Kwantes, B. Penders, G. Schilders, R. Talens, D. van Maassen,
Protein sequences were aligned with ClustalW and Muscle; unequivocally aligned H. van Zoggel, M. Veugelink and P. Wijnhoven for help with the isolation of various N. ovalis
positions were selected with Gblocks or manually. Phylogenies were inferred with sequences; and K. Sjollema for electron microscopy. G.W.M.v.d.S., S.Y.M.-v.d.S. and G.R. were
maximum likelihood by using a discrete gamma-distribution model with four rate supported by the European Union 5th framework grant ‘CIMES’. This work was also supported by
categories plus invariant positions and the Poisson amino acid similarity matrix, and equipment grants from ZON (Netherlands Organisation for Health Research and Development),
NWO (Netherlands Organisation for Scientific Research), and the European Union 6th
neighbour joining as implemented in ClustalW, correcting for multiple substitutions with
framework programme for research, priority 1 “Life sciences, genomics and biotechnology for
the Gonnet amino acids identity matrix, and bootstrapping with 100 samples.
ORFs with a lower size limit of 100 nucleotides were identified with ORF Finder health” to W.J.H.K..
(http://www.ncbi.nlm.nih.gov/gorf/gorf.html). tRNAs were identified with tRNAscan-SE
(http://www.genetics.wustl.edu/eddy/tRNAscan-SE). Potential mitochondrial import Competing interests statement The authors declare that they have no competing financial
signals were detected with MITOP (http://mips.gsf.de/cgi-bin/proj/medgen/mitofilter). interests.
Sequence searches were performed with BLASTX (http://www.ncbi.nlm.nih.gov/BLAST),
BLASTN and FASTA. For references on phylogenetic analysis see Supplementary Correspondence and requests for materials should be addressed to J.H.P.H.
Information. (j.hackstein@science.ru.nl). Sequences have been deposited at the EMBL database under
accession numbers AF480921, AJ871267, AJ871313–AJ871361, AJ871573–AJ871576, AY608627,
Received 18 August 2004; accepted 7 January 2005; doi:10.1038/nature03343.
AY608632–AY608634, AY616150–AY616152, AY619980, AY619981, AY623917, AY623919,
1. Müller, M. The hydrogenosome. J. Gen. Microbiol. 39, 2879–2889 (1993). AY623925, AY623926, AY628683, AY628684, AY628688.
2. Roger, A. J. Reconstructing early events in eukaryotic evolution. Am. Nat. 154, S146–S163 (1999).
3. Tielens, A. G. M., Rotte, C., van Hellemond, J. J. & Martin, W. Mitochondria as we don’t know them.
Trends Biochem. Sci. 27, 564–572 (2002).
4. Embley, T. M. et al. Hydrogenosomes, mitochondria and early eukaryotic evolution. IUBMB Life 55,
387–395 (2003).
5. Dyall, S. D., Brown, M. T. & Johnson, P. J. Ancient invasions: From endosymbionts to organelles. ..............................................................
Science 304, 253–257 (2004).
6. van der Giezen, M., Sjollema, K. A., Artz, R. R., Alkema, W. & Prins, R. A. Hydrogenosomes in the
anaerobic fungus Neocallimastix frontalis have a double membrane but lack an associated organelle
Image segmentation and lightness
genome. FEBS Lett. 408, 147–150 (1997).
7. Clemens, D. L. & Johnson, P. J. Failure to detect DNA in hydrogenosomes of Trichomonas vaginalis by
perception
nick translation and immunomicroscopy. Mol. Biochem. Parasitol. 106, 307–313 (2000).
8. Leon-Avila, G. & Tovar, J. Mitosomes of Entamoeba histolytica are abundant mitochondrion-related Barton L. Anderson1 & Jonathan Winawer2
remnant organelles that lack a detectable organellar genome. Microbiol. 150, 1245–1250 (2004).
9. Fenchel, T. & Finlay, B. J. Ecology and Evolution in Anoxic Worlds (Oxford University Press, Oxford, 1
University of New South Wales, School of Psychology, Sydney, New South Wales
UK, 1995).
10. Embley, T. M., Horner, D. A. & Hirt, R. P. Anaerobic eukaryote evolution: hydrogenosomes as
2052, Australia
2
biochemically modified mitochondria? Trends Ecol. Evol. 12, 437–441 (1997). Massachusetts Institute of Technology, Brain and Cognitive Sciences, Cambridge,
11. Voncken, F. et al. Multiple origins of hydrogenosomes: functional and phylogenetic evidence from Massachusetts 02139, USA
the ADP/ATP carrier of the anaerobic chytrid Neocallimastix sp. Mol. Microbiol. 44, 1441–1454 .............................................................................................................................................................................
(2002). The perception of surface albedo (lightness) is one of the most
12. van der Giezen, M. et al. Conserved properties of hydrogenosomal and mitochondrial ADP/ATP
carriers: a common origin for both organelles. EMBO J. 21, 572–579 (2002).
basic aspects of visual awareness. It is well known that the
13. Martin, W., Hoffmeister, M., Rotte, C. & Henze, K. An overview of endosymbiotic models for the apparent lightness of a target depends on the context in which
origins of eukaryotes, their ATP-producing organelles (mitochondria and hydrogenosomes), and it is embedded1–6, but there is extensive debate about the
their heterotrophic lifestyle. Biol. Chem. 382, 1521–1539 (2001).
14. Martin, W. & Müller, M. The hydrogen hypothesis for the first eukaryote. Nature 392, 37–41 (1998).
computations and representations underlying perceived light-
15. Akhmanova, A. et al. A hydrogenosome with a genome. Nature 396, 527–528 (1998). ness. One view asserts that the visual system explicitly separates
16. Brunk, C. F., Lee, L. C., Tran, A. B. & Li, J. Complete sequence of the mitochondrial genome of surface reflectance from the prevailing illumination and atmos-
Tetrahymena thermophila and comparative methods for identifying highly divergent genes. Nucleic pheric conditions in which it is embedded7–10, generating layered
Acids Res. 31, 1673–1682 (2003).
17. Burger, G., Gray, M. W. & Lang, B. F. Mitochondrial genomes: anything goes. Trends Genet. 19,
image representations. Some recent theory has challenged this
709–716 (2003). view and asserted that the human visual system derives surface
18. Dyall, S. D. et al. Non-mitochondrial complex I proteins in a hydrogenosomal oxidoreductase lightness without explicitly segmenting images into multiple
complex. Nature 431, 1103–1107 (2004).
19. Hrdy, I. et al. Trichomonas hydrogenosomes contain the NADH dehydrogenase module of
layers11,12. Here we present new lightness illusions—the largest
mitochondrial complex I. Nature 432, 618–622 (2004). reported to date—that unequivocally demonstrate the effect that
20. van Hoek, A. H. A. M. et al. Multiple acquisition of methanogenic archaeal symbionts by anaerobic layered image representations can have in lightness perception.
ciliates. Mol. Biol. Evol. 17, 251–258 (2000). We show that the computations that underlie the decomposition
21. Degli Esposti, M. Inhibitors of NADH-ubiquinone reductase: an overview. Biochim. Biophys. Acta
1364, 222–235 (1998).
of luminance into multiple layers under conditions of transpar-
22. Akhmanova, A. et al. A hydrogenosome with pyruvate formate-lyase: anaerobic chytrid fungi use an ency can induce dramatic lightness illusions, causing identical
alternative route for pyruvate catabolism. Mol. Microbiol. 32, 1103–1114 (1999). texture patches to appear either black or white. These results
23. Boxma, B. et al. The anaerobic chytridiomycete fungus Piromyces sp. E2 produces ethanol via
pyruvate:formate lyase and an alcohol dehydrogenase E. Mol. Microbiol. 51, 1389–1399 (2004).
indicate that mechanisms involved in decomposing images into
24. van Hellemond, J. J., Klockiewicz, M., Gaasenbeek, C. P. H., Roos, M. H. & Tielens, A. G. M. layered representations can play a decisive role in the perception
Rhodoquinone and complex II of the electron transport chain in anaerobically functioning of surface lightness.
eukaryotes. J. Biol. Chem. 270, 31065–31070 (1995). The amount of light projected to the eyes (luminance) is
25. Sickmann, A. et al. The proteome of Saccharomyces cerevisiae mitochondria. Proc. Natl Acad. Sci. USA
100, 13207–13212 (2003).
determined by a number of factors: the illumination that strikes
26. Cotter, D., Guda, P., Fahy, E. & Subramaniam, S. MitoProteome: mitochondrial protein sequence visible surfaces, the proportion of light reflected from the surface
database and annotation system. Nucleic Acids Res. 32, D463–D467 (2004). and the amount of light absorbed, reflected or deflected by the
27. Voncken, F. G. J. et al. A hydrogenosomal [Fe]-hydrogenase from the anaerobic chytrid Neocallimastix
prevailing atmospheric conditions (such as haze or other partially
sp L2. Gene 284, 103–112 (2002).
28. van Hoek, A. H. A. M. et al. Voltage-dependent reversal of anodic galvanotaxis in Nyctotherus ovalis. transparent media). Only one of these factors, the proportion of
J. Eukaryotic Microbiol. 46, 427–433 (1999). light reflected (lightness), is associated with an intrinsic property of
NATURE | VOL 434 | 3 MARCH 2005 | www.nature.com/nature 79
© 2005 Nature Publishing Group
letters to nature
surfaces, and hence is of special interest to the visual system. To into their constituent physical sources, and to suggest that the visual
accurately recover lightness, the visual system must somehow system uses computational ‘short cuts’ to generate representations
disentangle the contributions of surface reflectance from the illu- of surface lightness. Such theories have suggested that the visual
mination and atmospheric conditions in which it is embedded. One system divides an image into two-dimensional regions (or ‘frame-
theoretical view asserts that the visual system explicitly decomposes works’) rather than layers. In such models, lightness is derived
images into a set of separate maps or layers, corresponding to the through processes that bias the highest luminance to appear white11,
separate physical contributions to retinal luminance9,10. However, and/or by using statistical estimation techniques within local image
there is a growing body of data showing that the visual system can regions to compute reflectance12; no explicit decomposition of the
make systematic errors in estimating surface reflectance11, the image into separate layers occurs.
opacity of transparent surfaces or media13 and the amount of One of the most widely used techniques to explore context effects
illumination striking a surface14. These errors have led some to in lightness perception is to embed identical target patches in
question whether the visual system explicitly decomposes images different surrounds. Most studies with this method have used

Figure 1 Static versions of the lightness illusions studied in our experiment (see also surrounds are identical. In both cases, the figures on the dark surround appear as light
Supplementary Video 1). In a, the corresponding textured disks on the dark and light objects visible through dark haze, whereas the figures on the light surround appear as
surrounds are physically identical, and in b the corresponding chess pieces on the two dark objects visible through light haze.
80 NATURE | VOL 434 | 3 MARCH 2005 | www.nature.com/nature
© 2005 Nature Publishing Group
 letters to nature
untextured patches containing a uniform reflectance or luminance. determine the contribution of simple contrast enhancement mecha-
In such images, it is usually difficult if not impossible to determine nisms of the surrounds on uniform grey patches (see Methods and
whether a target is decomposed into multiple sources; the target Supplementary Video 2). Results of these experiments are shown in
region simply appears to be a particular shade of grey. To assess the Fig. 2. The solid lines depict the lightest and darkest pixels in the
role of layered image representations in perceived lightness, we target patches (Fig. 2a). Lightness matches by the observers corre-
devised a new set of stimuli that would make a layered decompo- spond closely to these lines, but there is also a slight overestimation
sition perceptually apparent if it was occurring (Fig. 1). We of the lightness of the targets on both surrounds (the apparent
generated textured images that contain a continuous distribution saturation of the matches to the light target reflects the limited
of luminance values and we also manipulated geometric and luminance range of the monitors; the luminance setting in these
luminance relationships known to play a role in the segmentation regions is simply the maximal available). This overestimation plays
of surfaces into multiple layers13,15–17. A common ‘seed’ texture was only a small part in the magnitude of the lightness transformation
used to create both the targets and the surrounds (see Methods). reported here, but it is consistent with data showing that the visual
The target regions in the two images were identical; only the system normalizes luminance in a manner that generates a bias for
surrounds differed. One surround was made lighter than the seed observers to perceive the highest luminance as white11. The contrast
image, and the other darker. The target regions were placed in the control experiment (Fig. 2b) showed that simple contrast enhance-
same relative position (compared to the seed image) on each of the ment processes produce a much smaller illusion (only 11% as large
two surrounds. The critical image properties manipulated using this as the largest lightness difference with the textured targets), and
technique were the polarity and magnitude of contrast between the therefore cannot account for the illusions in Fig. 1.
textures and their surrounds. In the image with the dark surround, These results are consistent with the hypothesis that the trans-
the polarity of the surround–target border was dark–light along its formation in lightness observed in Fig. 1 arises from segmentation
entire length of the border (respectively); in the light surround, the processes involved in the perception of transparency. If this is
surround–target border was light–dark. Contrast magnitude varied correct, then such lightness transformations should be abolished
continuously over both surround–target borders. As can be seen in if the conditions critical for inducing the perception of transparency
Fig. 1 (and even more dramatically in the moving versions in
Supplementary Video 1), this manipulation caused a striking
difference in appearance between the central targets. For the dark
surround, the target regions appeared white, visible through dark,
partially transparent clouds; for the light surround, the identical
targets appeared black, visible through light clouds. Note that the
fluctuations in contrast magnitude along the target–surround
border appear as variations in the opacity of the transparent layer;
this is in keeping with recent research demonstrating that the visual
system uses variations in contrast magnitude to compute the
opacity of transparent layers13.
We performed a lightness matching experiment to determine
what was responsible for the perceived lightness of the targets in
these images. The targets in Fig. 1 contain luminance values that
span the range from white to black. One explanation of the
perceived lightness difference in these images is that the two
surrounds cause the target regions to be decomposed in two very
different ways. For the targets on the light surround, the darkest
pixels appear to form an unobscured view of the distant surface, and
lighter pixels appear to be a combination of a light transparent layer
and a dark distant surface. For the targets on the dark surround, the
lightest pixels appear to form an unobscured view of the distant
surface, and darker pixels appear to be a combination of a dark
transparent layer and a light distant surface. This decomposition is
consistent with recent theory that asserts that the visual system
makes use of the sign and magnitude of image contrast to determine
those portions of surfaces that are in plain view and those surface
regions that are obscured by transparent media15. In this account,
the highest contrast regions are seen in plain view (the lightest and
darkest pixels on the dark and light surrounds respectively), and
lower contrast values are seen through a contrast-reducing medium
(where the opacity of the transparent layer is proportional to the
amount by which the highest contrast is reduced). According to this
view, the entire target regions in each figure are seen to have a single
lightness value as determined by the pixels in plain view. If this Figure 2 Lightness matching data. a, Data obtained using the moving version of the
analysis is correct, then the perceived lightness of the light target illusion. The light and dark surrounds were held constant, and the contrast (luminance
should correspond to the perceived lightness of the brightest pixels range) of the circular target patches were varied. Solid lines depict the luminance of the
in the target region, and the matches for the dark target should lightest (upper line) and darkest (lower line) pixels in the target patch. The data are close to
correspond to the perceived lightness of the darkest pixels in the these lines, but there is a bias for observers to report the targets on both light and dark
target region. surrounds as lighter than these values. b, The data in a are compared with the control
To test this hypothesis, we varied the range of intensities in the experiment using homogeneous targets that varied in luminance (plotted as a difference
target region (that is, its contrast), and observers adjusted the between the matches to the targets on the dark and light surrounds, respectively). The
luminance of a test patch until it appeared to match the lightness contrast effects are only 11% as large as the largest effects with the textured targets. Error
of the target patches. A control experiment was performed to bars depict the standard error of the mean for three subjects.
NATURE | VOL 434 | 3 MARCH 2005 | www.nature.com/nature 81
© 2005 Nature Publishing Group
letters to nature

Figure 3 Transparency control experiment. The same targets and surrounds are used as Video 3). This rotation destroys both the geometric and luminance conditions needed to
in Fig. 1a, except that the surrounds have been rotated by 908 (see also Supplementary evoke a percept of transparency, and also destroys the lightness illusion.

are removed but all other aspects of the display remain unchanged. induce percepts of transparency in stereoscopic displays17. However,
To test this hypothesis, the light and dark surrounds were simply the causal role of such segmentation processes in lightness percep-
rotated by 908, destroying both the geometric conditions (the tion has not been previously established for monocular images.
continuity of the textures in the targets and the surround) and The data presented here provide unequivocal evidence that seg-
the luminance relationships needed to induce the perception of mentation processes underlying the formation of layered image
transparency (the contrast polarity now reverses along the representations can play a critical and dramatic role in lightness
borders of both set of targets). As can be seen in Fig. 3 (and in perception. Theories of lightness perception that do not include
Supplementary Video 3), this manipulation destroys the lightness such processes are at best incomplete. A
difference observed in Fig. 1, demonstrating that transparency plays
a critical role in the lightness transformations that occur in these Methods
displays. Textures
The phenomena reported here provide new insights into the Textured ‘seed’ images were generated in Matlab as grey-scale noise with a specified
computations underlying lightness perception. The data are con- power spectrum that varied as (1/f 4), 512 £ 512 pixels. The different frequency
sistent with the view that lightness perception cannot be understood components were summed with random phases and orientations. The target and
background images were spatially identical to the seed image, but differed in the range
with low-level mechanisms such as lateral inhibition, as such of luminance values. The target image had 99% Michaelson contrast, with intensities
mechanisms produce a much smaller illusion2,5,11. Recent lightness ranging from 1 to 96 cd. For the surrounds, the luminance ranges were compressed and
models that omit computations that generate layered image either shifted up or down. For the light surround (which gives rise to the percept of a
representations also fail to account for the phenomena reported dark target seen through light clouds), the range was 36 to 96 cd (45% contrast) and for
the dark surround, the range was 1 to 77 cd (98% contrast). The illusions (static, Fig. 1;
here. Such models decompose images into a set of discrete two- moving, Supplementary Video 1) were created by aligning the target texture with one of
dimensional sub-regions, and estimate lightness within each sub- the surround textures and then showing the target through a circular aperture on either
region separately using principles of anchoring11 or statistical the light or dark surround. The multiple apertures in Fig. 2 represent the effect of
estimation12. Note, however, that the perceived transmittance of motion. For the control demonstrations (Fig. 3), the identical targets were used but the
surrounds were rotated by 908. This caused the polarity relationships between the target
the transparent layer appears to vary continuously over the entire patch and the surround to vary, destroying the percept of transparency and the lightness
image in Fig. 1. It is unclear how such models could account for illusion.
these phenomena.
It should be noted that layered image representations underlying Matching experiment
the illusions reported here are conceptually related to figure–ground To quantify the perceived lightness in Fig. 1, subjects adjusted a test patch to match the
perceived lightness of the targets. Using the stimuli described above, observers were
reversals16. Note, however, that traditional figure–ground reversals
presented with a circular target (38 in diameter) moving back and forth horizontally (one
involve shifts between image regions that occupy different regions of cycle every 5 s) on either a light or a dark square surround (158 per side). They adjusted the
space, whereas the phenomena reported here involve image regions luminance of a uniform, square test patch (28) on a black and white checkered background
along the same visual directions, and hence involve the same set of (38) until the test patch appeared to be the same lightness as the moving target. Subjects
pixels. To see this, consider the segmentation processes involved in had unlimited time to make the matches. Stimuli consisted of either the light or dark
surrounds depicted in Fig. 1 along with one of six target stimuli similar to the target in
viewing a surface through an occluding mesh or screen. In such Fig. 1, but ranging in contrast from 0.43 to 0.82. To measure the contribution of simple
contexts, the visual system must determine which image regions contrast enhancement processes to the illusion, homogenous grey disks that were identical
correspond to the occluding screen and which regions correspond in size to the textured targets were also presented (ranging in luminance from 10 to 89 cd),
to the underlying surface visible through the holes in the screen. and the same matching task was used. All combinations of backgrounds (one light and one
dark) and targets (six patterned and five grey) were presented five times each in random
This conception of transparency is readily generalized to continu- order to each subject, for a total of 55 trials.
ous media by simply allowing the holes to become infinitesimally Stimuli were generated using Vision Shell software and were presented on an Apple
small. If the perceived depth order of the surfaces is reversed, then Macintosh G4 computer using a Lacie (electron22blue) monitor that was calibrated and
the perceived lightness of the two layers will shift as well, as can be linearized before testing observers. Viewing distance was 57 cm. One subject, CB, was
naı̈ve and the other two observers were the authors.
experienced in Fig. 1.
There is a growing body of data demonstrating that a variety of Received 13 October; accepted 13 December 2004; doi:10.1038/nature03271.

factors influence perceived lightness, including surface curvature4, 1. Koffka, K. Principles of Gestalt Psychology (Harcourt, Brace and World, New York, 1935).
2. Gilchrist, A. L. Perceived lightness depends on spatial arrangement. Science 195, 185–187
surface orientation18, depth2,10,11,17 and simply the number of (1977).
different surfaces in a scene11. Previous research has shown that 3. Gilchrist, A. L. When does perceived lightness depend on perceived spatial arrangement? Percept.
transformations in perceived lightness can occur in images that Psychophys. 28, 527–538 (1980).

82 NATURE | VOL 434 | 3 MARCH 2005 | www.nature.com/nature

© 2005 Nature Publishing Group
 letters to nature
4. Knill, D. C. & Kersten, D. Apparent surface curvature affects lightness perception. Nature 351, is lost in most tumours: gap-junction-mediated intercellular
228–230 (1991).
5. Adelson, E. H. Perceptual organization and the judgment of brightness. Science 262, 2042–2044 (1993).
peptide coupling for presentation by bystander MHC class I
6. Eagleman, D. M., Jacobson, J. E. & Sejnowski, T. J. Perceived luminance depends on temporal context. molecules and transfer to professional antigen presenting cells
Nature 428, 854–856 (2004). for cross-priming.
7. Bergstrom, S. S. Common and relative components of reflected light as information about the
MHC class I molecules present peptides to the immune system
illumination, colour, and three-dimensional form of objects. Scand. J. Psychol. 18, 180–186 (1977).
8. Gilchrist, A. L. The perception of surface blacks and whites. Sci. Am. 240, 112–123 (1979). for surveillance by CD8þ cytotoxic T cells (CTL). Because intra-
9. Barrow, H. G. & Tenenbaum, J. in Computer Vision Systems (eds Hanson, A. R. & Riseman, E. M.) 3–26 cellular antigens and antigenic peptides usually cannot traverse
(Academic, New York, 1978). membranes, only endogenous peptides can be presented by MHC
10. Anderson, B. L. A theory of illusory lightness and transparency in monocular and binocular images:
the role of contour junctions. Perception 26, 419–453 (1997).
class I molecules1. Antigenic peptides from infected cells are thus
11. Gilchrist, A. et al. An anchoring theory of lightness perception. Psychol. Rev. 106, 795–834 (1999). exclusively loaded on the cell’s own MHC class I molecules and not
12. Adelson, E. H. in The New Cognitive Neurosciences 2nd edn (ed. Gazzaniga, M.) 339–351 (MIT Press, on those of innocent bystander cells. This concept has been
Cambridge, Massachusetts, 1999). challenged by a process called cross-presentation2,3 . Cross-
13. Singh, M. & Anderson, B. L. Toward a perceptual theory of transparency. Psychol. Rev. 109, 492–519
(2002).
presentation implies the transfer of antigenic (usually intracellular)
14. Rutherford, M. D. & Brainard, D. H. Lightness constancy: a direct test of the illumination estimation antigens from diseased cells to professional antigen-presenting cells
hypothesis. Psychol. Sci. 13, 142–149 (2002). (APC) such as dendritic cells, activated monocytes or Langerhans
15. Anderson, B. L. The role of occlusion in the perception of depth, lightness, and opacity. Psychol. Rev.
cells2–4. The APCs subsequently present these antigenic peptides on
110, 762–784 (2003).
16. Anderson, B. L. The role of perceptual organization in White’s illusion. Perception 32, 269–284 their own MHC class I molecules, and migrate to draining
(2003). lymph nodes where activation and expansion of the specific
17. Anderson, B. L. Stereoscopic surface perception. Neuron 24, 919–928 (1999). CD8þ T-cell population occurs. Cross-presentation requires that
18. Boyaci, H., Maloney, L. T. & Hersh, S. The effect of perceived surface orientation on perceived surface
albedo in binocularly viewed scenes. J. Vis. 3, 541–553 (2003).
antigens somehow enter the MHC class I presentation pathway of
an APC.
Supplementary Information accompanies the paper on www.nature.com/nature. In this study, we investigated the possibility of direct gap-junc-
tion-mediated transfer of antigens between the cytoplasm of two
Acknowledgements We thank N. Witthoft for suggesting the chessboard variant of the illusion adjacent cells. Gap junctions are assemblies of intercellular channels
and C.U. Jo for inspiration and support. that form an integral part of multicellular organisms. A functional
channel is formed when a hemichannel, composed of six connexin
Competing interests statement The authors declare that they have no competing financial
interests.
molecules, assembles with a hemichannel from an adjacent cell5.
The resulting gap junctions electrically couple cells by direct
Correspondence and requests for materials should be addressed to B.L.A. (bart.a@unsw.edu.au). exchange of ions and allow exchange of nutrients and second
messengers. Gap junctions are thought to be non-specific channels
that allow passive diffusion of molecules with a relative molecular
mass of up to 1,000 (M r 1K)6 and intracellular signalling controls
the gating7. Connexin 43 (Cx43) is broadly expressed, whereas the
other connexin family members are expressed in specific tissues
.............................................................. only. Cx43 is also expressed in various haematopoietic cells like
follicular dendritic cells, B cells, activated lymphocytes and mono-
Cross-presentation by intercellular cytes8. Importantly, many tumour cells are uncoupled from their
environment, for example after inactivation of their gap junctions
peptide transfer through gap by ras, src and neu oncogenes or by APC deficiency9,10. Viral
junctions proteins of the herpesvirus HSV-2 (ref. 11) and the human papil-
loma virus HPV-16 (ref. 12) are able to close gap junctions of
infected cells. In addition, gap junction intercellular communi-
Joost Neijssen1*, Carla Herberts1*, Jan Wouter Drijfhout2, Eric Reits1,
cation seems to be important for the bystander effect in cancer gene
Lennert Janssen1 & Jacques Neefjes1
therapy13.
1
Division of Tumor Biology, The Netherlands Cancer Institute, Plesmanlaan 121, To visualize peptide transfer between cells, we used A431 cells.
1066 CX Amsterdam, The Netherlands This human squamous carcinoma cell line does not express gap
2
Department of Immunohematology and Blood Transfusion Leiden, University junctions, as shown by biochemical and biophysical techniques14.
Medical Center, Albinusdreef 2, 2333RC Leiden, The Netherlands A431 cells were stably transfected with human Cx43 (Fig. 1a),
* These authors contributed equally to this work resulting in functional gap junctions. To study peptide transfer
............................................................................................................................................................................. between cells, stable fluorescently labelled (FL-) peptides were
Major histocompatibility complex (MHC) class I molecules synthesized. These peptides are not degraded in cells because they
present peptides that are derived from endogenous proteins1. are composed of D -amino acids with a protective group at the
These antigens can also be transferred to professional antigen- amino terminus 15 . A 9-mer FL-peptide was introduced in
presenting cells in a process called cross-presentation, which A431/Cx43 and control A431 cells by micro-injection, together
precedes initiation of a proper T-cell response2,3; but exactly how with dextran-TexasRed (TxR) (M r 70K) as an injection marker.
they do this is unclear. We tested whether peptides can be Cells were subsequently analysed by confocal laser scanning
transferred directly from the cytoplasm of one cell into the microscopy (CLSM) (Fig. 1b) and transfer was quantified in both
cytoplasm of its neighbour through gap junctions. Here we cell lines (Fig. 1c). Whereas dextran-TxR is maintained in the
show that peptides with a relative molecular mass of up to micro-injected cell, the 9-mer peptide diffused into surrounding
,1,800 diffuse intercellularly through gap junctions unless a cells only when the cells expressed Cx43. Closure of gap junctions by
three-dimensional structure is imposed. This intercellular pep- chemical inhibitors such as 2-APB (ref. 16) prevented this inter-
tide transfer causes cytotoxic T-cell recognition of adjacent, cellular peptide transfer between A431/Cx43 cells.
innocent bystander cells as well as activated monocytes. Gap- To test the efficiency of gap-junction-mediated peptide transfer,
junction-mediated peptide transfer is restricted to a few coupling small groups of A431/Cx43 cells were grown on coverslips. Peptides
cells owing to the high cytosolic peptidase activity. We present a of various lengths were micro-injected along with dextran-TxR and
mechanism of antigen acquisition for cross-presentation that the rate of transfer was determined using fluorescence recovery after
couples the antigen presentation system of two adjacent cells and photobleaching (FRAP) techniques. Peptides were permitted to
NATURE | VOL 434 | 3 MARCH 2005 | www.nature.com/nature 83
© 2005 Nature Publishing Group