Institution: NIH Library Sign In as Member / Individual

This Article Full Text Services Alert me to new issues of the journal Request Copyright Permission

Samples. Detailed information about the cerebral cortex samples used in this study is provided in Table 1. Human cortex samples were collected from two females and three males during autopsy [Homo sapiens (Hs)1, -2, and -3] or surgical procedures (Hs4 and -5) and were obtained from the Brain and Tissue Bank for Developmental Disorders at the University of Maryland (BTBUM; Baltimore) or the University of California at San Diego Medical Center (approved and monitored by the University of California at San Diego Institutional Review Board). Cortex samples of non-human primates were provided by the University of Louisiana at Lafayette New Iberia Research Center and the Salk Institute for Biological Studies. Procedures involving non-human primates were carried out in accordance with the guidelines established by the Institutional Animal Care and Use Committees of the New Iberia Research Center and the Salk Institute for Biological Studies. Common chimpanzee [Pan troglodyte (Pt)] samples were removed during postmortem dissections of three females and one male that died of natural causes (Pt1-4). Rhesus macaque [Macaca mulatta (Mm)] samples were dissected from four females and three males killed with a lethal dose of barbiturate (Mm1-7). Individuals were mostly adults, ranging in age between 29 and 71 years (average 43.4 years) for humans, 1 and 34 years (average 18.5 years) for chimpanzees, and 1 and 19 years (average 6.1 years) for rhesus macaques. Tissue was obtained from the cortex of several regions of the frontal, parietal, and temporal lobes of the left hemisphere (FCx, frontal cortex; FP, frontal pole; MFG, medial frontal gyrus; IPL, inferior parietal lobule; aIT, anterior inferotemporal cortex; STG, superior temporal gyrus; TP, temporal pole). Human heart samples were obtained from the BTBUM and correspond to one female and two males with ages between 20 and 25 years (average 22.3 years). Two pygmy chimpanzee (Pan paniscus) heart samples were obtained from the FrozenZoo of the Zoological Society of San Diego and correspond to neonate males. In all cases, tissue was immediately frozen after dissection and kept at –80°C until used for RNA extraction.

RNA Extraction. For each sample, total RNA was extracted from »100 mg of frozen tissue by homogenization in 1 ml of TRIzol (Invitrogen) with a 7-mm Polytron (Kinematica, Lucerne, Switzerland) at maximum speed for 2 min. Subsequent steps were done according to manufacturer’s instructions. RNA quality was evaluated in a 1% agarose gel by the 28S:18S rRNA ratio and was similar among all samples. RNA was stored at –80°C.

Oligonucleotide Arrays. Labeling and hybridization to the Affymetrix GeneChip Human Genome U95Av2 (HG_U95Av2) (Affymetrix, Santa Clara, CA) arrays were performed starting from 10 mg of total RNA as described (1, 2), with the exception that hybridization was done at 50°C. These arrays contain 12,625 probe sets for »10,000 different genes. Each of the tissue samples was processed independently and hybridized to a different array, except for the cortex samples of Pt1, -2, and -3, where two independent RNA extractions were performed from the same tissue and hybridized to separate arrays. Array results were analyzed using the MICROARRAY SUITE (MAS) software, Ver. 4.0 (Affymetrix). All arrays were normalized separately to a same average intensity of 200 based on the probe sets corresponding to the 60th to 90th percentile of hybridization signals. The ratio of the hybridization to the 3' and 5' ends of the mRNAs for the set of control genes on the array was close to one and comparable in all samples. A similar percentage of genes were detected in all samples. On average in the brain of humans, 54.6% (SD = 3.2%), and chimpanzees, 51.5% (SD = 3.0%) were detected, whereas 39.4% (SD = 2.7%) were detected in rhesus macaques. In heart, the proportion of probe sets detected by the arrays was 43.9% (SD = 2.1%) for humans and 46.3% (SD = 0.8%) for chimpanzees.

To calculate the similarity between the expression levels of the samples within each species, we used the Pearson correlation coefficient of the signal intensities of all probe sets in the array. The average pairwise Pearson correlation (r) of the signal intensities in the cortex of the different humans was 0.936 (SD = 0.036); of chimpanzees, 0.946 (SD = 0.014); and of rhesus, 0.969 (SD = 0.011). These correlations are similar to those reported in human (r = 0.952, SD = 0.021) or chimpanzee cortex (r = 0.965, SD = 0.008) in the data set from a previous study (3). Although it did not reach statistical significance, the correlation among different cortical regions from the same individual (r = 0.974, SD = 0.018) was higher than the correlation between the same region from different individuals (r = 0.967, SD = 0.013). Furthermore, it was much higher than the correlation between species. Cluster analysis was carried out by average linkage hierarchical clustering using the CLUSTER and TREEVIEW programs (4). Before clustering, the hybridization signal from each probe set was median-centered and normalized across the samples, and the uncentered Pearson correlation was used as similarity metric.

Identification of Differentially Expressed Genes. Before the detailed analysis, the overall signal intensities for each array were normalized to the same value using a simple multiplicative factor. Global normalization procedures correct for small differences in sample amounts, RNA labeling efficiencies and array sensitivity, and tend to minimize signal differences between samples. However, it is important to determine that the results did not depend on a specific normalization approach. To test this, we examined the differences in hybridization signal between humans and chimpanzees using additional normalization and analysis methods (Table 3). First, in the BULLFROG analysis, arrays were also normalized to the same average intensity of 200 using the standard MAS 4.0 normalization method, which takes into account the hybridization signals of all probe sets excluding the 2% with the highest and the 2% with the lowest intensity. Second, in the analysis with DCHIP, that normalizes all arrays to the same overall image intensity based on the invariant set of probes, we additionally used the Perfect Match-only (PM-only) and Perfect Match/Mismatch (PM/MM) difference models (6) to calculate gene expression values. Finally, we analyzed the arrays using Teragenomics (TeraGenomics, Information Management Consultants, Inc., www.teragenomics.com). For the BULLFROG and DCHIP analyses, the criteria used to identify probe sets with signal differences between humans and chimpanzees were the same as described above. The criteria used to identify probe sets with signal differences in TeraGenomics were a consistent call of increase/marginal increase or decrease/marginal decrease, fold change >1.5 and absolute difference change >30 in at least 75% of the comparisons, a fold change >1.2 in at least 90% of the comparisons, and a present call in at least one of the arrays.

To identify genes with signal intensity differences between primate species, we used the BULLFROG, Ver. 4.5 (5) and DCHIP, Ver. 1.0 programs (6). In the BULLFROG analysis, first, all pairwise comparisons between the arrays of each species were generated using the MAS 4.0 software. Specifically, we generated 40 comparisons (this work) and 36 comparisons (analysis of data from ref. 3) between human and chimpanzee cortex, six comparisons between human and chimpanzee heart (this work), and 36 comparisons between human and chimpanzee liver (analysis of data from ref. 3). In the case of the human Hs3 from which three cortical regions (MFG, IPL, and aIT) were available, MFG was compared with Pt4 FP, IPL was compared with Pt2 IPL, and aIT was compared with Pt1 and Pt3 STG. The criteria used to identify probe sets with signal differences between species were a consistent call of increase/marginal increase or decrease/marginal decrease, fold change >1.8, and absolute difference change >50 in at least 75% of the comparisons, a fold change >1.3 in at least 90% of the comparisons, and a present call in at least one of the arrays. In the DCHIP analysis, expression values for each probe set were calculated using the average difference instead of the model-based expression indexes. The criteria used to identify probe sets with signal differences between species were a fold-change >1.8 by using the lower bound of the 90% confidence interval, absolute difference between means >100, t test P value <0.001, and a present call in >25% of the samples involved. For each tissue, the probe sets identified with the BULLFROG and DCHIP analysis were then combined to generate the final list (Table 2).

Estimation of False Positives. To approximate the false positive rate for these analyses and to evaluate potential confounding effects related to the fact that the cortical samples of each species are not perfectly matched, the original BULLFROG and DCHIP criteria used to identify genes with expression differences were applied to a series of comparisons between the more diverse human samples. First, we compared the expression levels in the cortex of the three males (Hs2, -3, and -4) and two females (Hs1 and -5) used in this study. In this analysis, 14 genes showed differences in hybridization levels between both sexes. Second, we studied the variation between the tissue obtained from autopsy or surgical biopsy, and compared the Hs1, -2, and -3 samples to those of Hs4 and -5. We found 21 genes with signal differences related to the collection procedure of the tissue. Finally, to determine the extent of the expression differences between distinct regions of the cortex, we compared the hybridization patterns of the three frontal (Hs1 FP, Hs2 FP, and Hs3 MFG) and three temporal cortex samples (Hs3 aIT, Hs4 aIT, and Hs5 aIT). Only 12 genes were identified as different between cortical regions. These results are very similar to those of a more extensive study of gene expression in the rhesus brain, which indicates that, with the exception of the visual cortex, variation in gene expression patterns between different regions of the cortex is small compared with other brain regions, and suggests that basal expression levels in the cortex are rather homogeneous (M.C., T.M.P., and C.B., unpublished results). None of the genes showing expression differences related to sex, origin of the samples, and cortex region coincide with those identified in the human–chimpanzee comparison. Thus, the effect of the variation between the samples appears to be negligible compared with the differences between species.

Distribution of Human–Chimpanzee Hybridization Ratios. For the study of the distribution of the relative differences in hybridization signals between humans and chimpanzees, we used as a variable the base-2 logarithm of the average ratio of the signal intensity in humans divided by that of chimpanzees. All of the probe sets in the array (excluding controls and probe sets with <12 probe pairs) were treated as independent data points. The two normalization and analysis methods used in the identification of the genes differentially expressed between humans and chimpanzees were tested. In the analysis with the MAS program, first the fold changes obtained in the different comparison files between human and chimpanzee samples were averaged by using BULLFROG; i.e., we averaged the fold changes of the 40 (this work) or 36 (analysis of data from ref. 3) comparisons between human and chimpanzee cortex, the six comparisons between human and chimpanzee heart (this work), and the 36 comparisons between human and chimpanzee liver (analysis of data from ref. 3). The average ratio between humans and chimpanzees was then obtained by inverting the fold changes with negative values and eliminating the negative sign. Finally, the average ratios were transformed by calculating the base-2 logarithms. A second analysis was done using the DCHIP program. In this case, the fold change for each probe set was obtained directly from the comparison between the human and chimpanzee arrays. The fold changes were then transformed into base-2 logarithm ratios following the same procedure described above.

The Kruskal–Wallis test was used to compare the distributions of the human–chimpanzee ratios of the signal intensity (base-2 logarithm transformed) of 12,535 probe sets for the two cortex, heart, and liver data sets. As shown in Fig. 3, there were significant differences in the mean ranks of the distribution of signal ratios between the four tissues (H = 181.48, df = 3, P < 0.001). In addition, all possible comparisons between the distributions of these ratios for each pair of tissues were carried out with Mann–Whitney U tests. Significant differences (P < 0.001) were found when comparing the cortex and noncortex distributions of signal ratios, but not when comparing the two cortex distributions to each other or the heart and liver distributions. When the skewness of the distribution of signal ratios was compared between the tissues (Fig. 3), we found that the average skewness for the two cortex distributions was significantly greater than that of heart and liver (t = 4.31, df = 2, P < 0.05). These results support the existence of a bias toward higher signal intensities in humans than in chimpanzees in the cortex, which is not detected in heart or liver. Statistical analyses were performed with STATVIEW, Ver. 5.0 (SAS Institute, Cary, NC).

Sequence Difference Detection. Because the oligonucleotide arrays are designed based on human sequences, sequence differences between the non-human mRNAs measured and the array probes could result in low hybridization efficiencies and an underestimation of expression levels in non-human primates. To identify probes in the HG_U95Av2 arrays that might include sequence differences between humans and chimpanzees, we applied a new algorithm developed in the Barlow laboratory (J. A. Greenhall, M.A.Z., C.B., and D.J.L., unpublished results). The algorithm analyzes the detailed hybridization patterns of all of the oligonucleotide probes for each probe set, after normalizing for expression level differences, and finds probes that show consistent significant differences in the hybridization behavior between two sets of samples. A total of 28 arrays hybridized with human cortex and liver samples and 21 hybridized with chimpanzee cortex and liver samples were compared. The algorithm was written in Structured Query Language (SQL) by using QUERYMAN, a Teradata-specific compiler, to run on our in-house Teradata relational database and algorithms (Teragenomics, IMC, Inc.). The criterion to identify the probes with putative sequence differences between human and chimpanzees was a D-call cutoff value of 1.8 (which approximates a t test P-value of 0.00001).

To test the accuracy of the method, we examined 36 genes expressed in the cortex from which reliable nucleotide sequences of the region interrogated by the arrays were available in both species (corresponding to 730 probes of 51 probe sets). All of the chimpanzee sequences were obtained in the course of this study, and the human sequences were obtained from the June 2002 draft version of the human genome of the University of California, Santa Cruz. The comparison of the 30.4 kb of human and chimpanzee sequence showed they were 99.45% identical. It is important to note that the algorithm relies on the hybridization of the probe set to its target sequence. In cases where the mRNA is not detected in one of the species, either because the gene is not expressed or because the majority of probes include sequence differences, the algorithm will not be able to detect these differences. For the 33 genes that could be analyzed in both species, 40 of the 82 individual probes that contained sequence differences between humans and chimpanzees in their central 15 nucleotides were detected, with a low false positive rate of only 2%. Therefore, the algorithm was able to identify a significant fraction of the probes that might not hybridize efficiently in chimpanzee due to sequence changes. For the entire array, 4,577 probes of 2,285 probe sets that might be affected by sequence variation were found. When the array data were reanalyzed ignoring the signals for these probes, 65 of the original 246 probe sets with different hybridization levels between humans and chimpanzees were predicted to have sequence differences that could account for the higher signal intensity in humans. These 65 probe sets are indicated by an asterisk in Table 2, along with the number of probes in each of them that were predicted to include sequence differences between humans and chimpanzees.

Real-Time RT-PCR. Real-time RT-PCR was performed in an ABI PRISM 7700 Sequence Detection System with the DNA-binding dye SYBR green (Applied Biosystems). cDNA was generated using the SuperScript First Strand Synthesis kit (Invitrogen) with 200 ng of DNase I-treated total RNA from the cortex samples of three humans (Hs1 FP, Hs2 FP, and Hs5 aIT), three chimpanzees (Pt1 STG1, Pt2 IPL1, and Pt 4 FP), and three rhesus macaques (Mm3 TP, Mm5 FCx, and Mm7 FCx). Before the real-time RT-PCR analysis, »1 kb corresponding to the region covered by the oligonucleotide probes of all of the genes assayed was sequenced in one chimpanzee (Pt1) and one rhesus (Mm5), and the sequences were compared with those of the June 2002 draft version of the human genome of the University of California at Santa Cruz. To ensure that interspecific sequence differences did not affect the amplification, primers for the amplification were designed in regions that do not contain sequence differences between the three species by using PRIMER EXPRESS, Ver. 1.5, software (Applied Biosystems). The genes of interest were amplified in the three cortex samples of the three species in triplicate. b-actin mRNA was also amplified in triplicate in each sample as internal standard to control for differences in the cDNA concentration. The results of the real-time RT-PCR were analyzed by using SEQUENCE DETECTOR, Ver. 1.7, and DISSOCIATION CURVE, Ver.1.0, software (Applied Biosystems). Relative quantification was performed with the standard curve method according to the manufacturer’s recommended procedures. The amplification levels of the gene of interest were normalized by dividing them by the b-actin amplification level for each sample, and the three samples for each species were combined in a single expression value. The criterion to identify gene-expression changes was a difference >1.3-fold between the average expression levels of each species. This value provided a high sensitivity for detection of gene expression differences and maximum consistency between the quantitative RT-PCR and oligonucleotide array results, while maintaining a low false positive rate (as exemplified by some of the genes that do not show expression differences between humans and chimpanzees in the RT-PCR; Table 2).

cDNA Arrays. Human arrays containing 8,273 human cDNAs (»7,500 different genes) obtained from Incyte Genomics (Palo Alto, CA) were spotted in duplicate at the Salk Institute microarray facility. For hybridization to the arrays, 1 mg of total RNA was labeled with Cy5 or -3 by an aminoallyl indirect labeling procedure, and 20 pmol of probe was used. In the expression analysis of human and chimpanzee cortex, we performed four different comparisons according to the following fluor-reversal scheme (first sample labeled with Cy5, second sample labeled with Cy3): Hs1 FP-Pt2 IPL1, Pt4 FP-Hs2 FP, Hs3 aIT-Pt1 STG1, and Pt3 STG1-Hs5 aIT. In the analysis of human and rhesus cortex, the comparisons were: Hs1 FP-Mm5 FCx, Mm5 FCx-Hs1 FP, Hs2 FP-Mm6 FCx, and Mm6 FCx-Hs2 FP. Hybridized slides were scanned by using an Array Scanner GenIII (Molecular Dynamics), and background-subtracted data of all spots in the Cy5 and -3 channels were scaled to a common value in each slide. Only spots with a signal greater than background in at least 25% of the hybridizations were considered. The criteria for detecting significant differences between humans and non-human primates was a probability value of <0.05 by a paired t test of the hybridization signals in each spot and an average relative change between the two species >1.3-fold. As in the quantitative RT-PCR, these criteria maximized the number of genes with expression differences in the same direction between oligonucleotide and cDNA arrays.

In Situ Hybridization. Expression analysis of carbonic anhydrase II (CA2) and TWIST by in situ hybridization was performed by using slides containing multiple sections derived from the anterior inferior parietal lobule and the cerebellum of one human (Hs3), one chimpanzee (Pt4), and one rhesus macaque (Mm2). The tissue was stored at –80°C before sectioning. Coronal or sagittal sections of 20 mm were thaw-mounted onto 25 × 75-mm slides (Brain Research Laboratories, Newton, MA), air-dried, and postfixed in 4% paraformaldehyde for 20 min at room temperature. Slides were then rinsed in 0.1 M phosphate buffer (PB) and water, air-dried for 30 min, and stored at –80°C before use. In situ hybridization was carried out by the protocol described previously (7) by using radiolabeled riboprobes. After hybridization, slides were applied to autoradiographic film (BioMax MR film, Kodak) for a period of 5 days. After development and fixation of the film, the slides were dehydrated, defatted, and dipped in Kodak NTB2 emulsion for 4-6 weeks. The slides were then developed, counterstained with Cresyl violet, dehydrated, and coverslipped with Permount. The films and emulsion-dipped slides were analyzed by visual inspection and light microscopy using bright- and dark-field illumination. Control sections incubated with sense RNA showed no specific hybridization. CA2 (1,055 bp) and TWIST (906 bp) probes were derived from the 3' region of the human and chimpanzee cDNAs, respectively. Sequence divergence between the three species for both regions is <3%.

1. Sandberg, R., Yasuda, R., Pankratz, D. G., Carter, T. A., Del Rio, J. A., Wodicka, L., Mayford, M., Lockhart, D. J. & Barlow, C. (2000) Proc. Natl. Acad. Sci. USA 97, 11038–11043.

2. Wodicka, L., Dong, H., Mittmann, M., Ho, M. H. & Lockhart, D. J. (1997) Nat. Biotechnol. 15, 1359–1367.

3. Enard, W., Khaitovich, P., Klose, J., Zollner, S., Heissig, F., Giavalisco, P., Nieselt-Struwe, K., Muchmore, E., Varki, A., Ravid, R., et al. (2002) Science 296, 340–343.

4. Eisen, M. B., Spellman, P. T., Brown, P. O. & Botstein, D. (1998) Proc. Natl. Acad. Sci. USA 95, 14863–14868.

5. Zapala, M. A., Lockhart, D. J., Pankratz, D. G., Garcia, A. J., Barlow, C. & Lockhart, D. J. (2002) Genome Biol. 3, SOFTWARE0001-0009.

6. Li, C. & Wong, W. H. (2001) Proc. Natl. Acad. Sci. USA 98, 31–36.

7. Geschwind, D. H., Ou, J., Easterday, M. C., Dougherty, J. D., Jackson, R. L., Chen, Z., Antoine, H., Terskikh, A., Weissman, I. L., Nelson, S. F., et al.. (2001) Neuron 29, 325–339.