Characteristics of subjects
Two groups of patient samples, consisting of eleven SALS and nine control subjects, were used in this study. Fresh-frozen samples of human motor cortex (precentral-gyrus sections) were obtained from the NICHD Brain and Tissue Bank for Developmental Disorders under contracts NO1-HD-4-3368 and NO1-HD-4-3383 and selected for post-mortem intervals (PMI) prior to freezing not exceeding 24 hours. Additionally, two patient samples were obtained as total RNA from a commercial source (Ambion, Inc.). All sample material is certified to have been obtained following international ethical guidelines and with prior consent from a fully informed donor or a member of the donor's family. All disease samples were cases of spontaneous ALS, with a mean patient age of 68.2 ± 7.6 years. Control samples had a mean patient age of 68.7 ± 11.0 years and were selected for matching age and for causes of death unrelated to ALS or other neurological disorders. Detailed information related to source code, age, sex, race, and storage of patient samples is given in Table .
Individual slices of 10 μm were produced from tissue samples at -20°C by a Leica CM1510S cryostat (Leica Microsystems) and stored at -80°C until further processing. Two slices per sample were stained by hematoxylin/eosin staining (Bio Optica) and for Nissl substance (with a microfiltered solution of cresyl violet, Sigma), respectively, to assess integrity of cellular and tissue morphology. Individual slices were used for in-situ hybridization and immunofluorescence (see below). Ten adjacent slices per sample were pooled and used for RNA extraction with Trizol (Invitrogen) following the manufacturer's standard protocol, followed by confirmation of RNA integrity by agarose gel electrophoresis.
Microarray processing and data extraction
Complementary RNAs (cRNAs) labeled with Cy5-CTP (Perkin-Elmer) were synthesized from 1 μg of total RNA of each sample using the Low RNA Input Fluorescent Linear Amplification Kit (Agilent Technologies) following the manufacturer's protocol. A reference cRNA, labeled with Cy3-CTP (Perkin-Elmer), was synthesized from 1 μg of sample D2 RNA. Aliquots (750 ng) of Cy3 and Cy5 labeled cRNA targets were co-hybridized on Whole Human Genome Oligo Microarrays (Agilent Technologies). Microarray hybridization and washing were performed using reagents and instruments (hybridization chambers and rotating oven) as indicated by the manufacturer (Agilent Technologies).
Microarrays were scanned at 10-μm resolution using a GenePix Personal 4100A microarray scanner and the GenePix Pro 6.0 acquisition and data-extraction software (Molecular Devices, Corp.). Raw data were processed and analyzed with Acuity 4.0 (Molecular Devices, Corp.) and GeneSpring 7.2 (Agilent Technologies). To remove unreliable data, all genes from all samples were filtered for quality to include only probe data fulfilling all of the following criteria: the spot had < 3% of saturated pixels at 635 and 532 nm; the spot was not flagged "bad", "not found" or "absent"; the spot had relatively uniform intensity and uniform background (Rgn R2
635/532 > 0.6); the spot was detectable well above background (signal-to-noise ratios at 635 or 532 nm > 3). Filtering data by quality control criteria short-listed 19,431 genes as our complete data set for Lowess normalization and subsequent analyses, out of a total of 44,233 probes (including control and alignment probes) present on the microarrays. The microarray data have been deposited in the Gene Expression Omnibus (GEO) database [129
] under accession no. GSE4595.
Pathway-based microarray analysis
To analyze the gene expression changes in the context of known biological pathways we used the Gene Map Annotator and Pathway Profiler (GenMAPP) 2.0 software package [130
Input data for GenMAPP analysis were unique probe identifiers, corresponding UniGene cluster IDs, an average fold-change and an uncorrected P
value (Welch's t-test) diseased vs. control for all 19,431 short-listed genes of our complete data set. If a UniGene cluster ID was represented by multiple probes, the probe carrying the lowest (most significant) uncorrected P
value was used for pathway-based comparison with spinal-cord data [4
] (see below). GenMAPP dynamically links gene-expression data to the gene ontology (GO) hierarchy of biological processes, cellular components and molecular functions [131
]. For each of the GO categories in the hierarchy (nodes), and including all the genes in its child nodes, GenMAPP identifies genes that meet a user-defined criterion (≥ 1.2-fold-change in our analysis, as recommended by [132
] and [133
]). From the total number of genes analyzed, the number of genes in a category and the corresponding numbers of genes meeting the queried criterion a z score is calculated, which indicates the non-randomness (for high, positive z scores) of the proportion of genes meeting the criterion. Additionally, a corrected P
value is calculated through permutation analysis (2000 permutations) of the data, followed by Westfall-Young adjustment for multiple hypothesis testing. Permutation analysis sets limits to the resolution of extremely significant events, so that the P
value for all events with an uncorrected P
< 0.0005 equals zero. Only nodes with positive z scores and a corrected P
< 0.05 were considered in our analysis, with a higher z score indicating greater significance between nodes of identical P
Gene-based microarray analysis
Of our quality-filtered data set of 19,431 genes, those with an average change greater than twofold were screened by a two-sided one-way ANOVA using Welch's t-test (with a Kolmogorov-Smirnov-test P value of 0.38 or above indicating a normal distribution of tested values), followed by the Benjamini and Hochberg False Discovery Rate procedure as a multiple testing correction. Genes with a corrected P value < 0.05 were selected as differentially expressed candidate genes. All candidate probe sequences were tested against the NCBI nucleotide database (November 2005) by BLASTN to update their annotations and confirm the specificity of each probe.
Sample total RNA (2 μg) was reverse transcribed with the Protoscript reverse transcription kit (New England Biolabs) using the VN(dT)23 primer as recommended by the manufacturer. As a standard for relative RNA quantification (Standard cDNA), ten equivalent reactions using 2 μg pooled RNA from healthy control samples (222 ng each) were performed and the resulting cDNAs precipitated and resuspended in 50 μl water each. As a negative control for cDNA quantification, a further reaction using 2 μg of pooled RNA was performed, in which the reverse transcriptase was replaced by water. All cDNAs were tested by conventional PCR with primers for glyceraldehyde-3-phosphate dehydrogenase (New England Biolabs), using 1/50 of each cDNA (the equivalent of 40 ng total RNA).
Quantitative RT-PCR (qRT-PCR) amplifications were performed with a LightCycler (Roche Molecular Biochemicals) using the same starting amount and LightCycler® FastStart DNA MasterPLUS SYBR Green I reagents in a standard volume of 20 μl. Real-time detection of fluorimetric intensity of SYBR Green I, indicating the amount of PCR product formed, was measured at the end of each elongation phase. Fluorescence values measured in the log-linear phase of amplification were considered using the second-derivative-maximum method of the LightCycler Data Analysis software (Roche Molecular Biochemicals). Relative quantification was performed using serial dilutions of pooled Standard cDNA (the equivalent of 80, 40, 20, 10 and 5 ng total RNA, respectively, run in duplicate) to provide a standard curve for each run. For all experiments, the standard curve had an error of below 5% and extended over the relative quantities of all individual samples.
Candidate genes whose differential expression was tested by gene-specific qRT-PCR analysis were parvalbumin and the metallothioneins MT1B, MT1E, MT1G, MT1L, MT1M, MT1X and MT2A. Differences in the quantity of starting material were compensated by normalization with the housekeeping genes beta-2-microglobulin (B2M) and ribosomal protein L19 (RPL19). The primers used are detailed below. Normalized fold-changes between diseased and healthy samples were calculated and tested by a two-tailed Mann-Whitney U-test (not assuming equal variances).
cDNAs corresponding to four mRNAs were synthesized by PCR from human brain RNA using specific forward and reverse primers. PCR fragments were subcloned into pCR4Blunt-TOPO vector (Invitrogen) and the orientation of the insert was determined by sequencing. Following linearization of plasmids by restriction endonucleases, riboprobes containing Cy3-CTP (sense) or Cy5-CTP (antisense) (Perkin-Elmer) were synthesized by T3 and T7 polymerase, respectively, using the MAXIscript In Vitro Transcription Kit (Ambion, Inc.). The in-situ
hybridization procedure was performed as previously described [134
]. Fluorescent hybridization signals were obtained by scanning sections at 5-μm resolution using a GenePix Personal 4100A scanner (Molecular Devices, Corp.). No signal was detected in control brain sections hybridized with the sense riboprobes or pretreated with RNase before hybridization with the antisense probes. Evaluation of hybridization signals were obtained by using a computer-assisted image analysis system and the Photoshop 7.0 software (Adobe Systems, Inc.).
Protein was extracted from human motor-cortex samples using standard methods. Following SDS-PAGE separation on 10–12% gels and semi-dry protein blotting (transfer buffer 10 mM NaHCO3, 3 mM Na2CO3, 20% methanol, pH 9.9) to nitrocellulose Hybond membranes (Amersham Pharmacia Biotech), blots were blocked (0.1% Tween, 3% milk PBS), incubated with the appropriate primary antibody at 4°C over-night, and following three washes with PBS were incubated with secondary antibody. Primary antibodies (all Santa Cruz Biotechnology, Inc.) were rabbit antibodies ANXA2 (SC-9061), AQP1 (SC-20810), and CANX (SC-11397), and goat antibody NRGN (SC-18336), all at 1:100 dilution. As HRP-conjugated secondary antibodies (all Amersham Pharmacia Biotech) we applied a 1:4000 dilution of anti-rabbit antibody for detection of ANXA2, AQP1, and CANX, and of anti-goat antibody for NRGN. Immunoblots were visualized using the Enhanced Chemiluminescence System (Amersham Pharmacia Biotech). Band intensities were determined as background-corrected volume measurements using the ImageQuant TL software (Amersham Pharmacia Biotech), were equalized using CANX as a loading reference (which has not been linked to ALS and in our microarray analysis showed an mRNA ratio diseased vs. control of 0.97), and were subjected to statistical analysis using a two-tailed, heteroscedastic Student's t-test in Excel (Microsoft Corp.).
The following forward (FP) and reverse (RP) primers were used for qRT-PCR analysis: PVALB (GenBank Accession No: NM_002854, forward primer (FP): 5'-acgctgaggacatcaagaagg-3', 5'-caattttgccgtccccatc-3), reverse primer (RP): MT1B (GenBank Accession No.: NM_005947, FP1: 5'-actccaggcttgtcttggctcc-3', RP1: 5'-tgggagcagggctctcccaa-3', FP2: 5'-ttgcctaggaactccaggcttgt-3', RP2: 5'-gcagcggcacttctctgatgag-3', FP3: 5'-tgctgctcttgctgccccgt-3', RP3: 5'-aaagaatgtagcaaaccggtcaggg-3'), MT1E (GenBank Accession No.: NM_175617, FP: 5'-ccttcttccccaggctgctgt-3', RP: 5'-aatgcagcaaatggctcagtgttg-3'), MT1G (GenBank Accession No.: BC035287, FP: 5'-gcatctgcaaaggggcatcg-3', RP: 5'-aaaggaatgtagcaaaggggtcaaga-3'), MT1L (GenBank Accession No.: BC070351, FP: 5'-gggctcctgctcctgtgcca-3', RP: 5'-ggaatgtagcaaatgctcagggttg-3'), MT1M (GenBank Accession No.: NM_176870, FP: 5'-tggtgtctcctgcgcctgca-3', RP: 5'-aatgcagcaaatggctcagtatcgtatt-3'), MT1X (GenBank Accession No.: BC053882, FP: 5'-tgctgctcctgctgccctgt-3', RP: 5'-aaaagatgtagcaaacgggtcaggg-3'), MT2A (GenBank Accession No.: NM_005953, FP: 5'-cgactctagccgcctcttca-3', RP: 5'-gaaaaaggaatatagcaaacggtcac-3'), RPL19 (GenBank Accession No.: NM_000981, FP: 5'-ggctgctcagaagataccgtg-3', RP: 5'-ggcgcttgcgtgcttccttg-3') and B2M (GenBank Accession No.: NM_004048, FP: 5'-agcgtactccaaagattcaggtt-3', RP: 5'-tacatgtctcgatcccacttaactat-3'). Specificity of PCR products obtained was characterized by melting-curve analysis followed by gel electrophoresis and DNA sequencing.
The following primers were used to amplify fragments for in-situ hybridization: ATP1A3 (GenBank Accession No.: NM_152296; FP: 5'-tcaagaaggaggtggctatg-3', RP: 5'-gagaagcagccagtgatgat-3');NRGN (GenBank Accession No.: NM_006176; FP: 5'-gactaggccagaactgagca-3', RP: 5'-agtggcacggagatgtagg-3'); PVALB (GenBank Accession No.: NM_002854; FP: 5'-agttgcaggatgtcgatgac-3', RP: 5'-ccagagtggagaattcgtca-3'); ANXA2 (GenBank Accession No.: NM_004039; FP: 5'-gatcatctgctccagaacca-3', RP: 5'-gagtcatacagccgatcagc-3').
Comparison with SALS spinal-cord and AD hippocampus expression data
For SALS spinal-cord data, series GSE833 [4
] of GEO was imported into Excel (Microsoft Corp.), and all UniGene ID numbers were updated (January 2006) using probe identities of the HuGeneFL array (Affymetrix, Inc.) in the NetAffx™ Analysis Center (Affymetrix, Inc.). Fold-changes were calculated for five SALS samples (GSM6827, GSM6828, GSM6834, GSM6835, GSM6836) vs. four control samples (GSM6826, GSM6829, GSM6830, GSM6831), for genes with a corresponding UniGene ID and showing a signal level of > 100 in at least five of these nine samples (3276 genes of a total of 7070 present on the HuGeneFL array). For 2266 of these genes the equivalent, quality-filtered (see above) non-redundant Agilent probes with the lowest uncorrected P
value (see GenMAPP analysis above) were short-listed to calculate a correlation coefficient r for the expression ratios in spinal cord and motor cortex using Excel. A correlation coefficient was determined for all 2266 shared genes and for 23 shared genes corresponding to candidate genes identified in motor cortex by gene-based analysis, respectively.
For AD hippocampus data, an analogous procedure was followed, using seven severe-AD samples (GSM21203, GSM21206, GSM21207, GSM21208, GSM21212, GSM21213, GSM21229) vs. nine controls (GSM21215, GSM21217, GSM21218, GSM21219, GSM21220, GSM21221, GSM21226, GSM21231, GSM21232) of the GEO series GSE1297, and including genes showing a signal level of > 100 in at least eight of these 16 samples, resulting in a shared dataset of 6375 reliably detectable genes. A correlation coefficient was determined for the expression ratios diseased vs. control for these shared genes and for 38 shared genes identified as differentially expressed in our SALS study, respectively.
Tissue slices were submerged for 10 minutes in acetone at -20°C, followed by washes in phosphate-buffered saline (PBS), and incubation with 0.5% Triton in blocking solution (2% fetal calf serum, 2% bovine serum albumin, 0.2% fish-skin gelatine) for 10 minutes. Tissue slices were then incubated with 10-mM copper (II) sulfate and 50-mM ammonium acetate (pH 5.0) for 30 min to quench autofluorescence, followed by washes with PBS. Primary antibodies were co-incubated over-night at 4°C at dilutions of 1:50 for AQP1 (AQP1 antibody H-55, Santa Cruz Biotechnology, Inc.) and 1:3000 for neurofilaments (200 kDa + 160 kDa (phospho) antibody SMI 31, Abcam) in blocking solution without Triton X-100, followed by washes in PBS and co-incubation of secondary antibodies at room temperature at dilutions of 1:1500 anti-mouse Alexa Fluor 568 (A-11034, Molecular Probes) and 1:500 anti-rabbit Alexa Fluor 488 (A-21124, Molecular Probes). Tissue sections were then washed in PBS, incubated with Hoechst33342 as a DNA counter stain, washed in PBS, briefly air-dried, and mounted for microscopy in 10% Mowiol, 1% 1,4-Diazabicyclo [2.2.2]octane, and 25% glycerol in 0.1-M Tris buffer (pH 8). Images were acquired using an Axiovert 200 M microscope with the AxioVision 4.5 image acquisition software, the Apotome module and the transparency projection option for Z stacks of the Inside4D visualization module (all Carl Zeiss Inc.).