SEARCH

SEARCH BY CITATION

Keywords:

  • Xenopus;
  • subtractive hybridization;
  • blastema;
  • cDNA;
  • regeneration

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

Suppression polymerase chain reaction–based subtractive hybridization was used to identify genes that are expressed during Xenopus laevis hindlimb regeneration. Subtractions were done by using RNAs extracted from the regeneration-competent stage (stage 53) and regeneration-incompetent stage (stage 59) of limb development. Forward and reverse subtractions were done between stage 53 7-day blastema and stage 53 contralateral limb (competent stage), stage 59 7-day pseudoblastema and stage 59 contralateral limb (incompetent stage), and stage 53 7-day blastema and stage 59 7-day pseudoblastema. Several thousand clones were analyzed from the various subtracted libraries, either by random selection and sequencing (1,920) or by screening subtracted cDNA clones (6,150), arrayed on nylon membranes, with tissue-specific probes. Several hundred clones were identified from the array screens whose expression levels were at least twofold higher in experimental tissue vs. control tissue (e.g., blastema vs. limb) and selected for sequencing. In addition, primers were designed to assay several of the randomly selected clones and used to assess the level of expression of these genes during regeneration and normal limb development. Approximately half of the selected clones were differentially expressed, as expected, including several that demonstrate blastema-specific enhancement of expression. Three distinct categories of expression were identified in our screens: (1) clones that are expressed in both regeneration-competent blastemas and -incompetent pseudoblastemas, (2) clones that are expressed at highest levels in regeneration-competent blastemas, and (3) clones that are expressed at highest levels in regeneration-incompetent pseudoblastemas. Characterizing the role of each of these three categories of genes will be important in furthering our understanding of the process of tissue regeneration. Developmental Dynamics 226:398–409, 2003. © 2003 Wiley-Liss, Inc.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

Among vertebrates, the capacity for tissue and organ regeneration is best developed among urodele amphibians, which can replace lost limbs either as larvae or adults. Anuran amphibians, such as the African clawed frog Xenopus laevis, regenerate hindlimbs well during early larval stages but lose this ability as they approach metamorphosis (Dent,1962). With hindlimbs that can initially regenerate completely, followed by stages in which regeneration is abortive, larval Xenopus thus offers the opportunity to study factors controlling the regenerative process in the same species. Moreover, Xenopus is a useful model for studying genetic mechanisms that promote and/or inhibit limb regeneration, because of the embryo's relative ease of molecular manipulation and because the largest number of cloned genes from all amphibians are those from Xenopus (Slack,2000).

Regeneration of appendages involves the sequential and overlapping steps of wound healing, cellular dedifferentiation, proliferation to form the emerging blastema, pattern formation, and finally, differentiation. Dedifferentiation is a key element required for blastema formation and, thus, for regeneration. The amputation surface is covered by a dermis-free wound epithelium, which in turn signals the underlying muscle, cartilage, nerve sheath, and connective tissues to produce undifferentiated, mononucleated cells. These cells proliferate and form the blastema, a growing mass of mesenchymal cells that gradually begin to differentiate proximally and form the tissues of the new limb (Stocum,1995; Tsonis,1996; Mescher,1996).

Larval Xenopus hindlimbs produce a regeneration-competent blastema that regenerates normally if amputated at stage 53 or earlier (Nieuwkoop and Faber,1967) but produce limbs with missing structures if amputated at later stages. Size and specific skeletal components of the regenerated limb are reduced in a predictable sequence as limb removal is delayed, until by stage 59, just before metamorphosis, amputation produces a regeneration-incompetent blastema, termed a “pseudoblastema” (Komala,1957) that gives rise to an elongated heteromorphic spike consisting essentially of skin-covered cartilage (Dent,1962). The regenerative ability of hindlimbs declines in a proximal-to- distal orientation (Dent,1962), depending on the state of ossification of the structure amputated (Wolfe et al.,2000). The cause of this diminished regenerative capacity during development is unrelated to the organism's hormonal status, but involves changes intrinsic to the mesenchymal cells of the limb (Sessions and Bryant,1988; Filoni and Paglialunga,1990). Loss of regenerative ability before metamorphosis could be due to the silencing of regeneration-specific genes or to production of factors inhibitory to regeneration. The distinct differences in regenerative ability at various larval stages may be used to discover, by differential gene screening, the molecules and molecular pathways with possible roles in stimulating or inhibiting regeneration.

Our goal is to identify developmentally important genes that are differentially expressed between regeneration-competent and regeneration-incompetent stages in Xenopus by using subtractive hybridization (Hara et al.,1991; Diatchenko et al.,1996; Chenchik et al.,1998). We have constructed subtracted cDNA libraries of genes expressed in hindlimbs at these two developmental stages and in their distal, regenerated tissues 1 week after amputation (7-day blastema). Bioinformatic analysis of the cloned sequences provides new insights into the kinds of genes responsible for regenerative capacity in appendages. Primary polymerase chain reaction (PCR) screening of selected clones (genes) isolated in our subtractive screens indicates that there may be three distinct categories of genes associated with the regenerative process: (1) genes that are expressed in both regeneration-competent blastemas and -incompetent pseudoblastemas; (2) genes whose expression is elevated in regeneration-competent blastemas vs. -incompetent pseudoblastemas; (3) genes whose expression is elevated in regeneration-incompetent pseudoblastemas relative to -competent blastemas.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

Subtractive Hybridization

To identify hindlimb regeneration associated genes that are differentially expressed between regeneration-competent and regeneration-incompetent stages in Xenopus, we constructed the following subtracted cDNA libraries by using SMART PCR Select cDNA Subtraction (see the Experimental Procedures section for details): stage 53 7-day blastema minus limb (53B-53L); stage 53 7-day blastema minus stage 59 7-day pseudoblastema (53B-59B); stage 59 7-day pseudoblastema minus limb (59B-59L); stage 53 limb minus blastema (53L-53B); stage 59 7-day pseudoblastema minus stage 53 7-day blastema (59B-53B); and stage 59 limb minus 7-day pseudoblastema (59L-59B).

After construction of the subtracted cDNA libraries, clones for sequencing were selected either by random picking or by isolation from filter arrays after hybridization with specific probes. All clones sequenced were subjected to bioinformatic analysis. Clones exhibiting identity or overlap between reciprocal libraries (e.g., 53B-53L and 53L-53B) were assembled and in silico subtracted and extended by using matched sequences (identity of 50 bases or more) from the public databases. The processed sequences were compared with the public protein and nucleic acid databases by using BLAST to reveal known genes or genes with homology to known genes. The BLAST results indicated that, on average, 61% of the clones sequenced represented novel/unknown cDNAs and that 7% were hypothetical protein cDNAs.

We assessed the efficiency of our subtractions by using PCR-based assays as described by the subtraction kit manufacturer (BD Clontech, Inc.) using Xenopus control primers (e.g., ornithine decarboxylase [ODC] and EF1-α), and these results were well within the expected efficiencies for cDNA subtractions using the SMART PCR-Select subtraction kit (McGrew et al.,1999; Song et al.,1999).

Primary PCR Screen

As a first approximation of the effectiveness of the subtractions, 65 primer pairs representing randomly selected clones were used to assay the level of each RNA by using the PCR-amplified second-strand cDNA from the library construction as template (see the Experimental Procedures section). A scale for obtaining a qualitative assessment of the level of expression was developed for comparison (Fig. 1). Each primer pair was used in a 25-cycle PCR by using the second-strand cDNA as template. Of the 65 selected clones, 54% were differentially expressed when assayed by using their respective unsubtracted second-strand cDNA. The results of these assays showed that differential expression for the clones from the individual libraries was as follows: 53B-59B, 38%; 59B-53B, 67%; 53B-53L, 89%; 59B-59L, 33%; 59L-59B, 43%.

thumbnail image

Figure 1. Example of the ethidium bromide intensity and qualitative assignment used in the assay of polymerase chain reaction (PCR) -based expression analysis of 65 randomly picked subtracted sequences. A: Example of qualitative band intensity at 25 PCR cycles. B: Examples of differences in the band intensity of three clones amplified at 25 cycles from stage 53 7-day blastema and stage 59 7-day pseudoblastema second-strand cDNA. Clone 1 is expressed at +++ in both 50B and 53 B. Clones 2 and 3 are expressed at ++ in 53B and level + in 59B. The qualitative expression levels of clones 2 and 3 in stage 53 and 59 limbs are shown in Table 1.

Download figure to PowerPoint

Table 1. PCR-Based Expression Analysis of the 30 Randomly Selected Clones that Define the Three Specific Categories of Genes Expressed in 7-Day Hindlimb Blastemas/pseudoblastemas
GeneSt. 53 LimbSt. 53 BlastemaSt. 59 PseudoblastemaSt. 59 LimbAnnotationSpecies
  1. aAnnotation refers to either nucleotide or protein matches and reflects the categorizations described in the Experimental Procedures section. See Figure 1 for definition of intensity scale. Numbers refer to total PCR cycles needed to detect a signal equivalent to “+”. Species refers to the organism to which the nucleotide or protein sequence of subtracted cDNAs were most highly matched PCR, polymerase chain reaction.

Category I: Regeneration-competent and -incompetent blastemas
2435+++++Hypothetical proteinMouse
3930++++++++Hypothetical proteinMouse
4835+++++MLL-2 related homologXenopus laevis
4935+++++Hypothetical proteinHuman
52+++++Hypothetical proteinHuman
53++30Novel 
55+++++Thyroid hormone receptor beta AXenopus laevis
56++++++++++Ribosomal protein L1aXenopus laevis
5830+++++Novel 
62++30Novel 
64+++++++++Novel 
Category II: Regeneration-competent blastemas
2++++Novel 
3+++Novel 
630+++++LINE-1 reverse transcriplasePrimate
9++++Hypothetical proteinMedaka
1535++++pol polyproteinTakifugu rubripes
2335+++++++Novel 
2835++3030Cadherin-10Chicken
4235Novel 
47++++++col1A1Xenopus laevis
54++++Novel 
5735++++Novel 
5935++++−30Lambda recombinase-likeTetraodon nigrovindis
Category III: Regeneration-incompetent blastemas
21+35Novel 
2535MMP-28Xenopus laevis
3635+30Novel 
3735+++30ets-like factor ESE-2AHuman
40++30Novel 
5135Novel 
6135++++Novel 

The results of these primary PCR-based screens of selected genes indicated that 7.7% of the sequences are ubiquitously and abundantly expressed (i.e., detectable by using fewer than 20 PCR cycles) with equivalent intensity in all cDNA libraries screened. Of interest, all of the clones identified as abundantly expressed were isolated from the 53B-59B library. We found that 10.8% of the clones are expressed at low levels, requiring at least 35 PCR cycles to be detectable, and all of these clones were isolated from the blastema or pseudoblastema minus limb libraries. By carrying out the PCR-based analysis of expression of these 65 randomly picked clones, we were able to define three important categories (Table 1) of clones whose expression patterns are associated with blastemas and that may be important in our understanding of regeneration. Of the 65 clones analyzed, 30 were found to be defined by these three categories. The three categories are: (I) clones that are expressed in both regeneration-competent blastemas and -incompetent pseudoblastemas, (II) clones that are differentially expressed in regeneration-competent blastemas, and (III) clones that are differentially expressed in regeneration-incompetent pseudoblastemas (Table 1). These results identified differentially expressed markers that are potentially important for tissue regeneration. Category I clones expressed in both 53B and 59B could be genes necessary for blastema formation, irrespective of whether or not they are regeneration-competent or -incompetent. Category II clones, which are more abundantly expressed in 53B, could potentially be activators of regeneration. Category III clones more abundantly expressed in 59B could potentially be inhibitors of regeneration.

As an initial test of the effectiveness of our primary PCR screen, we have isolated a full-length cDNA of one clone that shares a high degree of homology (98%) at the amino acid level to a previously reported Xenopus sal-3 (Xsal-3) homolog (Onuma et al.,1999). A probe prepared from our Xsal-3 clone for whole-mount in situ hybridization shows expression in limbs and 7-day blastemas at the regeneration-competent stage and in 7-day pseudoblastemas but not in limbs at the regeneration-incompetent stage, consistent with our primary PCR data (manuscript in preparation).

Filter Arrays

We arrayed clones from three of our six subtracted libraries. These were the 53B-53L, 53B-59B, and 59B-59L libraries. These three were chosen for array analysis because we predicted that they would harbor the greatest numbers of clones representing genes whose functions would be to either promote or inhibit the processes of regeneration. As indicated in the Experimental Procedures section, duplicate filter sets were generated and then one filter from each array set was screened with radioactive probes prepared from the experimental tissue. For example, with the 53B-53L array filters, 53B probe represents experimental and 53L probe represents control. Due to expected detection limits, not all arrayed clones were detected, and the level of detection was quite different with the three arrays (data not shown). We chose clones for sequencing only if they exhibited a level of expression that was at least twofold higher in experimental vs. control probes. We could make two opposing assumptions as to the relative abundance of different classes of genes found by random selection vs. those selected from array screening. The abundance of different classes of genes isolated from the array filter screening would be expected to be different from those picked at random. This prediction would be based on the fact that we had no prior knowledge of the level of expression of the randomly selected clones; thus, many clones picked might have represented abundantly expressed genes. Alternatively, because the libraries are subtracted, it is equally likely that random selection would yield a statistically significant number of clones that also would be found to be expressed at higher levels by array screening. Thus, the two methods of selection might be expected to result in similar levels of the different classes of genes.

We find that only a few classes of genes are significantly different between the two types of selection (see below; Table 3). For example, the percentage of genes in the translation category is twofold higher in the clones selected from the array screens vs. those randomly selected. This finding indicates that certain classes of genes will be enriched when using a preselection criteria such as array screening. However, several classes of genes are nearly equally represented between random selection and array selection. This latter observation indicates that our libraries are sufficiently subtracted such that random selection would be expected to yield results similar to those of screening arrays.

Table 3. Comparison of Gene Classifications for All of the Randomly Selected and Array Screen-selected Clones
 Random selectionArray screen 53B-53LRandom selectionArray screen 53B-59BRandom selectionArray screen 59B-59L
 53L-53B53B-53L59B-53B53B-59B59L-59B59B-59L
Total no. of Clones132294327146321143307279156
Percentage of total clones in each gene class         
 Unknown (no match)65.1571.4353.5239.0483.1867.1349.5167.7461.54
 Hypothetical protein11.364.765.209.593.434.9011.736.093.85
 Transcription4.554.085.202.052.804.902.282.152.56
 DNA - chromatin1.521.363.982.741.562.801.633.587.05
 Translation4.554.769.482.740.932.803.581.083.21
 Metabolism - enzymes1.522.724.5913.700.312.107.495.387.69
 Signaling - hormones2.272.383.982.053.122.804.562.151.28
 Structural proteins0.761.363.3623.971.251.409.777.898.97
 Retrotransposon related0.003.061.530.001.251.400.000.000.64
 Transport - binding proteins6.822.043.984.110.932.109.453.233.21
 Repetitive element1.522.045.200.001.257.690.000.720.00

Bioinformatic Analysis

As described in the Experimental Procedures section, all of the raw sequences were trimmed of vector and adaptor sequences, in silico subtracted and extended (if possible) by using matches to sequences in the public databases. By using the bioinformatic data obtained from the sequence information from both random selection and array screening selection, we placed the sequences of the clones into 10 functional gene classes as well as a novel class (no match to any database sequence). Representative sequences and their annotations that allowed for the 10 gene classifications are presented in Table 2. The percentage of each gene class from the randomly selected and array-selected clones from each library is presented in Table 3.

Table 2. Representative Clones (10 From Each Classification) Whose Sequence Annotations Define the Functional Gene Classes Presented in Table 3
ClassificationAnnotationAccession no.Library source
Hypothetical protein   
 C. elegans hypothetical protein Y50E8A.g751007453B-53L
 S. scorfa basic proline-rich protein1514579753B-53L
 Herpes virus 4 Nuclear antigen EBNA1962562453B-53L
 Mus musculus RIKEN cDNABC00481053B-53L
 M. musculus RIKEN cDNANM_02651753B-53L
 M. musculus ORF2359933653B-59B
 X. laevis unknown proteinM2418753B-59B
 X. laevis cDNA clone 13A3AJ00928853B-59B
 M. musculus IMAGE: 3968454BC00557653B-53L
 M. musculus similar to hypothetical protein DKFZp566A1524BC01134353B-59B
Transcription   
 X. laevis mRNA for XFD-13AJ24268053B-53L
 X. laevis retinal homeobox 1A (Rx1A) mRNAAF00104853B-59B
 X. laevis Xvent-1BAJ13109453B-59B
 X. laevis mRNA for XP8AB05658253B-59B
 X. laevis XFG 5-1 and XFG 5-2 genesX7106753B-59B
 X. laevis XFD-1'U6575053B-59B
 X. laevis UBF alpha geneX6569753B-59B
 X. laevis Xsal-3 short and long formAB03082753B-53L
 X. laevis Lmx1bAF41408653B-53L
 X. laevis Y/CCAAT-box binding factor A, NF-YAAF04120353L-53B
DNA - chromatin   
 X. laevis DNA polymerase-αAF20299259L-59B
 X. laevis SSB1 geneX8367353B-59B
 X. laevis SSB2 geneX8367453B-53L
 X. laevis RAG-2 geneL1932553B-53L
 X. laevis regulator of chromosome condensation, RCC1D0067953L-53B
 ZW10 homolog, centromere/kinetochore proteinNM_00472459B-53B
 M. musculus REV-1-like (Rev11)NM_01957059B-59L
 X. laevis nucleosome assembly protein 1AF27853859B-59L
 H. sapiens REV3-like, catalytic subunit DNA polymerase zetaXM_04036559L-59B
 X. laevis DNA polymerase alphaAF20299259L-59B
Translation   
 X. laevis ribosomal protein S6L1999659L-59B
 X. laevis gene for ribosomal protein S1Z3453053B-53L
 X. laevis 5S rRNA geneM1085053B-53L
 X. laevis tandem repeat with tRNA genesY0043053B-53L
 X. laevis ribosomal protein L14X0502553B-53L
 X. laevis ribosomal protein L1aX0655253B-53L
 M. musculus ribosomal protein L44984529559B-59L
 X. laevis ribosomal protein L34AY07917753B-53L
 X. laevis ribosomal protein S6AF02055153B-59B
 X. laevis ribosomal protein S11X7880553B-59B
 O. cuniculus translation repressor NAT1U7611359B-53B
Metabolism - enzymes   
 M. musculus GalNAc transferase667993159B-59L
 G. gallus ubiquitin-conjugating enzymeAF12021259B-59L
 X. laevis aldolase BAB02452453B-59B
 C. familiaris MDCK420486353B-53L
 X. laevis thimet oligopeptidaseAB03090453B-53L
 X. laevis casein kinase-2 alpha subunitAY03295453B-53L
 X. laevis hatching enzymeD8963259B-59L
 C. familiaris MMP-9AF16924453B-59B
 X. laevis thymosin beta 4 peptideD1069259B-59L
 C. carpio creatine kinase M3-CKAF05529059B-59L
Signaling - hormones   
 M. musculus tissue-type vomeronasal pheromone receptor991030653B-59B
 X. laevis gene for proopiomelanocortin-AX5937053B-53L
 X. laevis partial beta-TrCPAJ42893053B-53L
 X. laevis thyroid hormone receptor beta AU0467553B-53L
 X. laevis BMP-2AJ31515953B-53L
 X. laevis POMC-B geneX5936953B-53L
 X. laevis Xwnt-4U1318353B-53L
 X. laevis thyroid receptor interacting protein 7 (Trip7)AF17032753B-59B
 X. laevis TGF-beta5AF00933453B-59B
 X. laevis TGF-beta2X5181759B-53B
Structural proteins   
 X. laevis beta-tubulin mRNAL0623259B-59L
 X. laevis cytoskeletal actin type-5 geneM2476959B-59L
 X. borealis larval alpha 3 skeletal actinX6350253B-53L
 O. cuniculus titin409669953B-53L
 X. laevis lamina associated protein 2-betaAF04881753L-53B
 M. musculus nebulin732998853B-59B
 X. laevis integrin alpha5 subunitU1268353B-59B
 M. musculus keratin complex 1, acidic gene 12711066753B-59B
 X. laevis B2 vitellogeninX0335553B-53L
 R. plpiens type 2 myosin heavy chainAF24068959B-53B
Retrotransposon related   
 X. laevis transposon Tx1M2691553B-53L
 O. latipes LTR retrotransposable element1815752153B-53L
 T. rubripes pol protein642516853B-53L
 LINE-1 reverse transcriptase homolog12629653B-53L
 Rat retrotransposon L111226253B-53L
 Multiple sclerosis associated retrovirus polyprotein232328753B-53L
 Reverse transcriptase silkworm transposon Pao47944353B-59B
 T. rubripes putative reverse transcriptase1426521153B-59B
Transport - binding   
 Bovine mitochondrial phosphate carrier proteinX0534053B-53L
 X. laevis Vg1 mRNA-binding proteinAY02892053B-53L
 M. musculus opsin 1BC02602153B-53L
 X. laevis aquaporin-3AJ13184753B-53L
 M. musculus schwannomin interacting protein 1NM_01392853B-53L
 M. musculus GTP-binding protein 31548855153B-59B
 X. laevis polyA-binding protein ABP-EFM2707253L-53B
 X. laevis TBP6Y1046153L-53B
 X. laevis ATP gated ion channel PX27AJ34511453B-59B
 X. laevis epithelial sodium channel alpha subunitU2353553B-59B
Repetitive element   
 X. laevis repeat sequence XstirAB03992253B-59B
 X. laevis DNA for OAX-RNAX0183359B-59L
 X. laevis PTR-2 repeatK0126259B-59L
 X. laevis REM 3 sequenceX0068053B-59B
 X. laevis monomeric repeat unit of satellite DNA called RHM2X0003653B-53L
 X. laevis monomeric repeat unit of satellite DNA called RHM5X0003753B-59B
 X. laevis repetitive 1723 elementX0007853B-53L
 X. laevis clone K11 OAX repetitive sequenceAF22541253B-53L
 X. laevis 3.18 tandem repeatY0043053B-59B
 X. laevis repeat element from gastrula mRNAM3686753B-59B

In examining Table 3, it is immediately apparent that the largest percentages of clones from any library source are those that are novel with no match to sequences in any public database. The high percentage of novel sequences found in our screens likely represents that we are analyzing a highly specialized late developmental stage system. The vast majority of Xenopus sequences deposited into the public databases are from studies aimed at very early development. Very few molecular studies involve tissue regeneration. Many of the sequences from the subtracted cDNAs in our study may represent 3′-untranslated regions of mRNAs and/or nonconserved coding regions, sequences that may be underrepresented in the public databases. This latter fact could, therefore, contribute to the fact that we found a large fraction of unknown sequences in our screens.

Several interesting observations come from examining percentages of the various gene classes and their differences among the various subtracted libraries. One potentially significant finding is that sequences that have homology to reverse transcriptases and retrotransposons were only isolated from the blastema and pseudoblastema libraries. Expression of retrotransposon-like sequences have been shown to be associated with enhanced, growth factor-stimulated gene activity and to exhibit developmentally specific patterns of expression (Greene et al.,1993; Trelogan and Martin,1995; Shim et al.,2000). Because the blastema is undergoing significant proliferation, the level of gene activation is likely to be significantly higher than that of the limb, which is in the patterning and differentiating stage of its development.

Clones containing repetitive element sequences were also found at highest levels in the blastema and pseudoblastema libraries. Although the exact function of expressed genes containing repetitive elements remains unclear, studies have shown that genes of the OAX family of small RNAs exhibit distinct patterns of expression both spatially and temporally during early embryonic development (Whitford et al.,2000). We found two independent cDNAs from the regeneration-competent blastema library that share homology to the OAX element. In addition, repetitive sequences in the untranslated regions (UTRs) of some genes may play a role in the regulation of transcriptional and translational efficiency, such as those found in the human zinc finger gene ZNF177 (Landry et al.,2001) and the Hlx homeobox gene required for hepatic growth and development (Bates et al.,2001).

Sequences that are classified as encoding transport and/or binding proteins are found at two- to threefold higher levels in the limb libraries relative to the blastema/pseudoblastema libraries. This finding may reflect that, as limb development proceeds from early differentiation, commitment to patterning, and the building of structures, more transport of proteins and other compounds may be required, thus signaling an increase in the production of proteins necessary for transport and/or binding.

cDNAs that encode transcription factors and proteins or RNAs involved in translation together represent the largest category (5–15%) of functional sequences identified in all of the libraries. This finding is not surprising because these tissues, whether blastema/pseudoblastema or limb, are in an activated state of either repair, proliferation, and/or undergoing patterning. Therefore, it would be expected that there would be a large demand for the machinery of protein synthesis.

Quantitative-PCR Analysis of Array Clones

To verify the accuracy of the filter array results, we compared the level of expression of five clones selected from the 53B-53L array by using quantitative PCR (Q-PCR). These five clones were shown, in two independent hybridizations, to be expressed at least twofold higher in the stage 53B probe vs. the stage 53 limb probe. Analysis of the expression of these five clones was carried out in triplicate with assay of ODC levels used to ensure equivalent input of first-strand cDNA. The results of the Q-PCR are shown in Table 4, and demonstrate that there is good correlation between levels of expression obtained from array screens and those obtained by Q-PCR. As expected, not all clones (e.g., clone E11) show complete agreement between the two assay methods (Altmann et al.,2001). The difference in expression level for clone E11 was calculated at >50-fold with the 53B probe. This array result is due to the lack of detection of this clone with the 53L probe, and not to a necessarily large signal with the 53B probe. One of the selected clones was F10. F10 was shown by Q-PCR to be expressed twofold higher in blastemas relative to limbs, consistent with the results we obtained by whole-mount in situ hybridization (data not shown).

Table 4. Q-PCR Analysis of Five Clones Selected From the 53B minus 53L Array Screen
CloneAnnotationArray ratioQ-PCR ratio
  1. aArray ratio refers to the fold difference in signal obtained with the 53B probe compared with the 53L probe. Q-PCR ratio refers to the difference in signal obtained from 53B RNA vs. 53L RNA. A value of >50 for the array ratio results when one probe (the 53L probe in this case) does not detect the same DNA on the duplicate filter array. Q-PCR, quantitative polymerase chain reaction.

A01Novel22
A07Novel3.22
D08Novel3.23.5
E11Novel>505.5
F10DEAD-box RNA helicase5.52

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

Limb regeneration in urodele amphibians consists of a series of hierarchical events initiated by an amputation or other severe injury. First, the cut surface undergoes epithelialization from neighboring epidermis to form the wound epithelium. Then, the stump tissue cells dedifferentiate morphologically and express inhibitors of tissue-specific gene expression (Shimizu-Nishikawa et al.,1999), giving rise to mesenchymal blastemal cells that proliferate, redifferentiate, and undergo morphogenesis to reform the missing limb (Geraudie and Ferretti,1998). Important for understanding tissue regeneration is the identification and characterization of molecules or classes of molecules involved in the process. One class of molecules, the extracellular matrix (ECM) proteins, has been shown to regulate cellular migration, proliferation, and differentiation (Adams and Watt,1993). Dedifferentiation includes breakdown of the ECM associated with fully differentiated cells of the limb stump. Growth and patterning of the dedifferentiated cells occurs in a hyaluronate-rich matrix with little collagen. Unlike urodeles, limbs of postmetamorphic frogs show little dedifferentiation after amputation and undergo a fibrotic scarring response with abundant collagen synthesis, rather than forming a regeneration blastema (Stocum,1995). Indeed, one of the randomly selected clones from 59L-59B is type 1 collagen α-1 (Table 1, gene 47).

We have used subtractive hybridization to identify genes differentially expressed 7 days after amputation in regeneration-competent and regeneration-incompetent anuran limbs. According to Dent (1962), the capacity for perfect hindlimb regeneration begins to decline between stages 51 and 53 and is completely lost by stage 60. To maximize the amounts of starting material, we chose hind limbs from stage 53 as the regeneration-competent stage and from stage 59 as regeneration-incompetent. Limbs from these two groups formed regeneration blastemas and pseudoblastemas, respectively (Komala,1957). These were removed 7 days after amputation, a time of growth when re-epithelialization and other early wound healing responses to amputation are complete, but tissue redifferentiation is not yet predominant in the limb (Geraudie and Ferretti,1998).

Our results show that differential subtraction is a useful method for identifying genes differentially expressed in regenerating limb tissues. As little as 1 μg of total RNA was reverse transcribed to yield the first cDNA strand, and the second strand of cDNA was generated by a PCR amplification step, yielding significant amounts of cDNA for subtraction and cloning. The method does not require a large amount of starting tissue, making it ideal for studying molecular mechanisms in a regeneration paradigm.

We analyzed 65 randomly selected subtracted clones to determine their level of expression in normal and regenerating limbs. This primary PCR screen identified three categories of genes, the expression of which are elevated in blastemas and pseudoblastemas relative to limbs. These three categories are genes whose expression is elevated in both regeneration-competent blastemas and regeneration-incompetent pseudoblastemas (category I), genes whose expression is elevated in regeneration-competent blastemas (category II), and genes whose expression is elevated in regeneration-incompetent pseudoblastemas (category III). Although this study involved only one time point, 7 days after limb amputation, we assume that each gene expressed in response to limb amputation has its own expression profile in time and place. The genes (clones) identified in our primary screen will need to be followed up by examining their temporal expression profile with in situ hybridization and with functional studies to determine what genes are necessary and sufficient for limb regeneration. The three categories of blastema-associated genes could indicate that to construct a blastema, it is necessary to express certain genes, whether or not that blastema has the capacity to regenerate. Potentially, the genes expressed at the highest levels in regeneration-competent blastemas may include stimulators of regeneration or encode proteins that repress the activity of inhibitors of regeneration. Whereas, the genes expressed at highest levels in the regeneration-incompetent pseudoblastemas may represent genes involved in the inhibition of regeneration. The ability or inability of an appendage to undergo epimorphic regeneration may be regulated by an equilibrium between various stimulators and inhibitors of regeneration; and the best way to address this possibility is to assay multiple genes simultaneously.

To better identify the potential significances of the categories of genes expressed during limb regeneration, we placed all of the sequenced genes (from both random selection and array screening) into 10 bioinformatic gene classes (Table 2). Clearly, the largest class of sequences identified in our studies from any of the subtracted libraries is that having no homology to any known sequence. The high degree of novel sequences identified in our study may reflect that we are working with tissues that have not yet been subjected to intense study. Although amphibian limb regeneration has been studied for many decades at the morphologic levels, until now, the molecular events associated with this process have been completely unknown. Additionally, the subtracted libraries may be enriched in sequences representing 3′-UTRs and, therefore, reflect sequences that are underrepresented in the public databases. However, this latter possibility is unlikely because the subtraction kit that was used uses a novel mRNA cap-binding oligonucleotide to prime second-strand cDNA synthesis (Warbrick and Glover,1994). Therefore, the distribution of cDNA fragments used for input into the subtraction would reflect the entire coding region of most if not all mRNAs. In support of this conclusion is that only 30% of all the clones sequenced (whether randomly picked or array selected) contained a polyA tail.

In some cases, the classes of genes identified in reciprocal library data sets are often equivalently represented, and in others, there are classes of genes that are represented at significantly different levels (Table 3). For example, in the 53B minus 53L library and its reciprocal (53L minus 53B), we found significantly more transport-binding protein sequences in the limb library than in the blastema library. We speculate that this may reflect the fact that the limb is undergoing significant patterning and modeling, requiring many proteins that are involved in transport and binding; whereas the blastema is undergoing dedifferentiation and proliferation, which requires a significantly different set of genes than a patterning and differentiating limb.

The one clone (F10) we isolated from the array screen (see Table 4) belongs to the family of DEAD-box RNA helicases. This class of proteins is related to translation initiation factor eIF4A, which represents the archetypal RNA helicase. These proteins are involved in the process of altering RNA secondary structure; and on the basis of the observed distribution of this class of protein, they are strongly implicated in spermatogenesis, embryogenesis, and cell growth and division (Leroy et al.,1989; Longo et al.,1996). Therefore, it is not surprising that one of the genes we have identified, and whose expression is elevated in the regenerating blastemas, belongs to a family of proteins that function in cellular proliferation.

One surprising finding in our study was the number of transposable element-related and reverse transcriptase-related sequences that were identified. These sequences represent an interesting gene class of cDNAs isolated in our study and appear to have a strong correlation to blastema specificity. Because embryonic development and the process of tissue regeneration are related, both processes would likely involve similar genes. There is a family of retrotransposons in Xenopus with developmentally regulated expression. The LTR-retrotransposons Xretpos are expressed only in the ventroposterior regions of the zygote, and the expression can be induced by ultraviolet (UV) irradiation and BMP-4 overexpression (Shim et al.,2000). In addition, the LINE-1 retrotransposons have been shown to have development-specific expression during mouse embryogenesis (Trelogan and Martin,1995). Retrotransposon elements make up a major part of the Xenopus, as well as other genomes (Trelogan and Martin,1995; Sheen et al.,2000) and seem to be involved in Xenopus regeneration and embryonic development.

As expected, several other classes of genes are predominant in these subtracted libraries. Genes encoding transcription factors and components of translation, signaling pathways and hormones are predominantly identified, whether randomly selected or selected based on expression level as determined in the array screens. Cascades of transcription factor expression are involved in establishing the anteroposterior as well as dorsoventral axes in Drosophila embryo (Jackle and Sauer,1993). Transcription factors are ideal for controlling developmental processes that involve coordinated efforts of multiple factors in complex networks (Latchman,1998). One transcription factor we isolated from our screen, Xsal-3, belongs to a superfamily of C2H2-type zinc finger-containing proteins. Loss of the human homolog of sal-3 has been shown to cause Townes-Brocks syndrome, one phenotype of which is limb abnormality (Kohlhase et al.,1998).

Tissue regeneration is likely to involve factors that both promote and inhibit the process. We can use the cDNAs cloned from our studies to identify both classes of genes. It is expected that several of the genes we have identified in our screens of regeneration-competent blastemas may promote regeneration, whereas genes identified in the regeneration-incompetent pseudoblastemas may be involved in inhibiting regeneration. We now have the tools to begin to test this hypothesis by introducing cDNAs into transgenic animals, as well as by direct application of expressing plasmids into regenerating limbs.

EXPERIMENTAL PROCEDURES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

Animals and Surgery

Larval Xenopus were raised in the laboratory from fertilized eggs or obtained commercially (Xenopus I). Hindlimbs were staged according to Nieuwkoop and Faber (1967). All surgical procedures were performed after anesthesia in 0.02% benzocaine in 100% Holtfreter's solution. Hindlimbs at either stage 53 (regeneration-competent) or stage 59 (regeneration-incompetent) were amputated unilaterally at the mid-zeugopodia level, and 7 days later, 20 to 30 blastemas or pseudoblastemas were collected 1 mm proximal to the original level of amputation. Contralateral control limbs were removed at the same level, and all tissues were processed by standard methods for total RNA isolation using Trizol solution (Invitrogen/Life Technologies).

Subtractive Hybridization

The experimental strategy used consisted of generating cDNA from regeneration-competent and -incompetent limbs and 7-day blastemas and pseudoblastemas, followed by subtracting one cDNA library from another to obtain cDNA enriched with regeneration-promoting or regeneration-inhibiting genes. Clones from each subtracted library were then selected at random or by array screening for differential expression and sequenced, and the sequences analyzed by using bioinformatic tools. Finally, differential PCR analysis was used to determine the level of expression of several clones in the various subtracted libraries.

Three separate forward and reverse subtractive hybridizations were carried out: (1) between cDNAs from stage 53 7-day blastemas and stage 53 control limbs (53B-53L and 53L-53B), (2) between cDNAs from stage 59 7-day pseudoblastemas and stage 59 control limbs (59B-59L and 59L-59B), and (3) between cDNAs from stage 53 7-day blastemas and stage 59 7-day pseudoblastemas (53B-59B and 59B-53B). Synthesis of cDNA, from 1 μg of total RNA, and subtractions were carried out by using the SMART PCR Select cDNA Subtraction protocol essentially as described by the manufacturer (BD Biosciences Clontech, Inc.). The optimal PCR cycle for second-strand cDNA synthesis was 18 cycles as revealed by agarose gel electrophoresis of the products at 15, 18, 21, and 24 cycles. The cDNA products were phenol-chloroform extracted and the DNA rendered blunt-ended by using T4 DNA polymerase and then digested with RsaI. The RsaI-digested cDNA was adaptor-ligated and subjected to suppression PCR reactions carried out according to the supplied protocol.

The subtracted cDNA was phosphorylated and ligated into the plasmid, pUC18 in a 20-μl reaction. The ligation reaction was diluted to 50 μl, and 1 μl was used to transform 30 μl of Electromax DH10B cells (Invitrogen/Life Technologies) by using a Gene Pulser (Bio-Rad Laboratories) for electroporation.

Growth of clones for random sequencing and for preparation of DNA for arraying on nylon membranes was carried out in deep-well 96-well plates containing 1.2 ml of Magnificent Broth (MacConnell Research) and 120 mg/L ampicillin. The plates were incubated in a shaking 37°C incubator for 22 hr at 325 rpm. Plasmid DNA was isolated from the cells grown in the plates by using standard mini-prep methods. Plasmid DNAs were used directly for robotic arraying or diluted 10-fold to yield a concentration of approximately 0.1 μg/μl for sequencing.

Filter Arrays and Random Selection

Clones for arraying were selected from three of the subtracted libraries: 53B-59B, 53B-53L, and 59B-59L. For filter arraying subtracted clones, plasmid DNA was extracted from a total of 6,150 clones grown in deep-well 96-well plates. We arrayed 2,307 clones from the 53B-53L library, 2,307 clones from the 53B-59B library, and 1,536 clones from the 59B-59L library. DNA was spotted robotically by using a PBA Flexys Robot (Genomic Solutions). The DNA was spotted onto nylon membranes (Hybond N+, Amersham Biosciences) that were presoaked in 1× SSPE. Gridding of each DNA was done in duplicate onto a 96-well grid pattern with a 4 × 4 “subgrid,” such that each “well” of the grid contained eight different DNAs in duplicate. The 4 × 4 subgrid pattern was selected so that duplicate samples were nonadjacent to both samples of any other pair of samples within the subgrid. This method was done to simplify the screening process and to reduce false-positive identification during the data acquisition process. Each filter array was generated in duplicate for screening with radioactive probes. After gridding, the nylon membranes were irradiated in a StrataLinker (Stratagene) to permanently affix the DNA to the membranes.

Random selection of subtracted clones for sequencing was carried out with clones from all six libraries. We picked a total of 384 clones from each library (except the 59B-53B and 53L-53B libraries, from which we picked 192 clones each) for a total of 1,920 clones. The colonies were placed into deep-well 96-well plates and grown overnight, and the DNA was extracted and sequenced.

Screening Filter Arrays

Filter arrays were screened by using radioactive probes prepared by PCR amplification of first-strand cDNA. First-strand cDNA synthesis reactions (20 μl) were carried out by using the SMART II kit (BD Biosciences Clontech, Inc.). The first-strand reaction was then diluted to 70 μl with TE. Radiolabeled PCR-generated probes were prepared in a 50-μl reaction containing 3.2 μl of first-strand cDNA; 5 μl of 10× Advantage 2 polymerase buffer (BD Biosciences Clontech, Inc.); 4 μl each of 2.5 mM dATP, dGTP, and TTP; 4 μl of 0.25 mM dCTP; 1 μl of PCR primer (SMART kit); 5 μl of 32PdCTP (∼3,000 Ci/mmol); and 1 μl of Advantage polymerase. The PCR reaction was carried out for 18 cycles of the following: 95°C for 30 sec, 65°C for 30 sec, 68°C for 6 min. Radiolabeled cDNA was then purified from unincorporated nucleotides by using G-50 spin columns (Roche Diagnostics Corp.). Array filters were prehybridized (5× SSPE, 5× Denhardt's, 0.2% sodium dodecyl sulfate [SDS]) at 65°C for 2 hr, the heat-denatured probe was added, and hybridization was carried out at 65°C in the same buffer for 18 hr. Filters were washed in 0.2× SSC with 0.2% SDS at 65°C, then exposed to Phosphor Image screens (Bio-Rad Laboratories) for 3 days. Data from the exposed screens was collected and analyzed for expression intensity by using the ArrayVision V4 software from Imaging Research, Inc. Filter spots (and the corresponding DNAs) exhibiting at least a twofold higher intensity with experimental (blastema) vs. control (limb) probes were then selected for sequencing.

Primary PCR Screening of Expression

Based on sequence and bioinformatic information (e.g., novel genes, genes with homologies to developmentally important genes, or homologies to certain human genes), we designed primer pairs for 90 of the randomly picked subtracted sequences. Of these, 65 amplified the expected sized product by using the original plasmid as template. Therefore, we analyzed the expression of these 65 genes by using PCR. The template for the PCR reactions was the PCR-amplified second-strand cDNA generated during the initial steps of the subtraction protocol. The differential expression PCR screen used 20, 25, 30, and 35 cycles for each reaction. Qualitative levels of amplification (expression) were assayed by intensity of the ethidium bromide signal after 25 PCR cycles. Each reaction was assigned a scale of intensity from 0 to +++ (Fig. 1). When no signal was evident, an additional 5 and 10 PCR cycles were run. If 30 cycles were necessary to see a PCR product, that reaction was assigned the number 30, and 35 if the product was visible only after 35 cycles, etc. PCR was performed by using Platinum Taq DNA polymerase (Invitrogen/Life Technologies). PCR-amplified double-stranded DNAs produced from total RNA by using SMART II cDNA synthesis (BD Biosciences Clontech, Inc.) were used as templates. PCR parameters were 94°C for 30 sec, 62°C for 30 sec, and 68°C for 30 sec. The internal standards, elongation factor 1-α (EF1-α), ODC (King and Moore,1994), and ribosomal protein S8A, RPS8A (Mariottini et al.,1988) were amplifiable with 20 PCR cycles with all templates used.

Quantitative PCR

Quantitative PCR was performed by using the DNA Engine Opticon System model PTC-200, which is composed of the DNA Engine Cycler and the CFD-3200 Opticon Detector (Qiagen, Inc.). Fluorescence chemistry was performed by using the QuantiTect SYBR Green PCR master mix (Qiagen, Inc.). The reaction components and cycler conditions were optimized for the DNA Engine Opticon system, as illustrated by Qiagen's QuantiTech SYBR Green PCR Kit for quantitative real-time PCR and two-step RT-PCR. The template for the Q-PCR reactions was first-strand cDNA prepared from total RNA of stage 53 limbs and 7-day blastemas. The cDNA was prepared from 1 μg of total RNA, exactly as described for the first-strand reaction used in the synthesis of radiolabeled probes for screening the array filters. The input blastema and limb cDNA was equivalent as assayed by UV spectroscopy. Each Q-PCR reaction was carried out in triplicate. Primers to ODC were used as the internal control. Melting curve analysis of all products demonstrated a single peak, indicating that each set of primers produced a single product.

Bioinformatics and Gene Classification

DNA sequencing was performed by using PE-ABI Prism Dye Terminator Cycle Sequencing Ready Reaction fluorescent-based chemistry. Sequence data were collected on an ABI3700 and analyzed by using PE-ABI Sequence Analysis version 3.6 software (Applied Biosystems). Data were edited by using Sequencher version 4.0.5 (Gene Codes Corp.). Sequences were trimmed of vector and adaptor sequences in two stages. First, a local implementation of VecScreen (www.ncbe.nlm.nih.gov/VecScreen/VecScreen.html) was used. Although this strategy successfully removed vector and adaptor sequence from the ends, we found significant numbers of chimeric sequences joined by medial sequences, which in most cases were clearly of adaptor origin. We identified these medial sequences in two ways: (1) identity or near identity to adaptor sequences used by the BD Biosciences Clontech kit, or (2) presence of the sequence internal to one or more clearly chimeric sequences. We split chimeric sequences into separate sequence files before subsequent analysis.

To reconstitute as much of the underlying mRNA as possible, sequences from each library were then assembled by using CAP2 (Huang,1996) and extended against all available Xenopus sequences by assembling all sequences with at least 50-bp overlaps of 100% identity. Reconstituted sequences from reciprocally subtracted libraries were then purged of overlapping sequence to eliminate remaining common sequences.

Remaining sequences were then annotated as follows: (1) those with matches of at least 150 nucleotides of 100% identity to Xenopus mRNA sequences in GenBank were classified as “known Xenopus,” (2) those with at least 50 nucleotides of 98% identity to known Xenopus transcripts were classified as “known Xenopus variant,” (3) those with at least 100 amino acids of overlap at the 90% identity level with a known Xenopus protein sequence were classified as “Xenopus homologue,” and (4) those with an overlap of 80 or more residues at the 65% identity level with a known protein from any source were classified as “reasonable homologues.” Matches were found with BLAST (Altschul et al.,1997), but identity criteria were evaluated by software available from Bioperl (www.bioperl.org), supplemented with software written by the authors.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

We thank the other members of the Regenerative Biology group: Ellen Chernoff, George Malacinksi, Simon Rhodes, Janet Hock, Santosh Mishra, Patanjali Sankhavaram, and Paul Rosteck for helpful discussions. We also thank Elizabeth Osborne and Yu Zhu for their expert technical assistance and the Eli Lilly and Co. BioRTP DNA Technologies Sequencing Core for DNA sequencing support for this project.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES