S. J. O’Brien
Editor’s Note
Around 30 years before the Human Genome Project mapped the tens of thousands of protein-coding genes of the human genome, debate over the eukaryotic gene complement was rife. The total number of genes was thought to be far less than the amount of DNA in the haploid genome, leading some to suggest that over 90% of the eukaryotic genome was nonfunctional or “junk”. Here geneticist Stephen J. O’Brien questions this assumption, arguing that the evidence for junk DNA is based on the response of the functioning genes to natural selection. Non-coding DNA is now thought to comprise most of the human genome, but the term “junk” is used with caution since functions have been ascribed to some so-called “junk” sequences. 中文
MANY recent studies have been concerned with the construction of biological model systems to describe adequately regulation of gene action during development of eukaryotes 1-5 . The number of genes in mammals and Drosophila has been suggested to be 1 to 2 orders of magnitude less than the amount of available DNA per haploid genome could provide 2-7 . Although Drosophila and mammalian nuclei contain enough unique DNA to specify for respectively 10 5 and 10 6 genes of 1,000 nucleotide pairs 8,9 , it has been argued that a much lower estimate of functional gene number is more reasonable 2-7 . Conversely, these conclusions indicate that more than 90% of the eukaryotic genome may be composed of nonfunctional or noninformational “junk” DNA. Here we demonstrate these estimations have not been fundamentally proven; rather they are based on simplifying assumptions of questionable validity, in some cases contradictory to experimental data. 中文
The perceptive model proposed by Crick 5 provides that the structural genes for proteins are situated generally in the interbands observed in the giant salivary gland chromosomes of Drosophila . The chromosome bands, which contain all but a few % of the DNA, are the sites of regulatory elements and presumably large amounts of noninformational DNA. The model thus predicts approximately 5,000 structural genes in Drosophila , the approximate number of salivary gland bands which can be observed. 中文
This model is strongly supported by the elegant work of Judd et al . 10 who examined 121 lethal and gross morphological point mutations that map in the zeste to white region of the tip of the X chromosomes in D . melanogaster . There are 16 salivary gland chromosome bands or chromomeres in this region corresponding to 16 complementation groups of the morphological or lethal point mutations. In addition a series of overlapping deficiencies supports the 1 band:1 complementation group relationship. Extrapolation over the entire genome gives approximately 5,000 complementation groups or genes to 5,000 chromosome bands. There are also estimates available on the total number of lethal loci in the Drosophila genome. By screening for large numbers of lethal chromosomes either in natural populations or following irradiation, it is possible to relate the frequency of allelism to the number of lethal loci by a simple Poisson distribution: and the number of lethals thus measured in Drosophila gives a result between 1,000-2,000 11,12 . 中文
The problem with extrapolation of the fine structure analysis and the lethal data to the functional gene number is our inability to answer the question: how many genes when mutated are capable of producing a lethal or gross morphological phenotype? The answer is not known specifically but the available data suggest that only a very small percentage of all gene products are critical enough to kill the organism if absent. In Drosophila over 30 genes have a known gene product 13 , of which there are 14 at which “null” alleles eliminate the protein, its activity, or the RNA product entirely, and of these (Table 1) only the bobbed locus has lethal alleles 14 . Most alleles, however, at that locus, which is the structural gene for ribosomal RNA, are viable even at very low levels of rRNA. The other genes, which code for enzymes whose function a priori seemed essential for normal metabolism, are in no case lethal when homozygous for completely “null” alleles. 中文
Table 1. Genes in Drosophila melanogaster with Known Gene Products and Recovered “Null” Alleles
* W. J. Young, personal communication.
Null alleles at the first eleven loci above were detected by the loss of histochemical stain development on an electrophoretic gel. The sensitivity of this assay detects at least 5% of normal enzyme levels. In several cases ( Acph-1, rosy, Adh, α Gpdh-1 ), analytical enzyme assays with a sensitivity near 0.1% of wild type enzyme levels also failed to detect trace activity in “null” homozygotes. In the two cases where cross reacting material (CRM) was measured ( Acph-1 and ry ) it was also negligible. 中文
Null alleles of at least two of the loci were induced in a crossing scheme that would have recovered lethal alleles ( α Gpdh-1 and Acph-1 ). A lethal “null” allele would also be detected as an exceptional heterozygote with normal alleles of different eletrophoretic mobilities in those cases of “null” alleles discovered in natural or laboratory populations ( Est-C, Est-6, Aph, Aldox, and Idh ). 中文
Five of the fourteen loci have alleles which produce visible recessive phenotypes; ry , cn and v affect eye colour, bb affects bristles, and α Gpdh-1 “null” mutations which, although they appear morphologically normal, lack ability to sustain flight. The fraction 5 of 14 should not, however, be taken as an estimation of the fraction of loci at which “null” alleles produce an observable phenotype. This number is probably an overestimate because 4 of the 5 loci in question (all except α Gpdh-1 ) were discovered initially as morphological mutations and their gene product was deduced and identified from their visible phenotype. 中文
The eye colour mutations affect enzymes involved in the biosynthesis of eye pigments, and the bobbed locus, which shows a syndrome of effects usually associated with protein synthesis, was identified as the gene for rRNA. The phenotype of the α Gpdh-1 “null” mutations might easily have been missed had not the importance of the enzyme in insect flight been known previously 15 . The 11 other loci were identified only as the genes for selected enzymes, and of these none exhibited lethality or any morphological phenotype when “null” alleles were found. 中文
In two cases double “null” mutants of alkaline and acid phosphatase (R. S. MacIntyre, personal communication) and of Z w and 6-Pgd (W. J. Young, personal communication) were constructed and proved viable, fertile, and morphologically normal. Also, in two of the five cases where there is an observable phenotype, bb and α Gpdh-1 , there occurs a modification of the phenotype in the afflicted stocks. In the case of bb the diminished rDNA cistrons become “magnified” to approach the wild type rRNA levels within a few generations 16 . Flies genetically deficient for α-glycerophosphate dehydrogenase lack the ability to sustain flight due to their disrupted α-glycerophosphate cycle 15 , but after 25 generations this phenotype becomes modified and flies recover the ability to fly normally (S. O’Brien, unpublished data). Biological adaptive capacity for physiological compensation for lesions in the structural genes of important functions must be very extensive to protect the fly so efficiently from genetically sensitive loci even in the presumably critical functions. 中文
One might argue that even the smallest cytologically observable mutations in most cases are recessive lethals 10,14,17 . Resolution of such cytology, however, demands that at least 1 of the 5,000 chromomeres of Drosophila polytene chromosomes must be absent to detect a deletion. The precision of the technique then is at the level of 10 6 nucleotides, the average amount per chromomere, enough DNA for 20 genes of average length. I suggest that there could be up to 20 functional genes in each region of which only one might be lethal in its mutant configuration. 中文
A second widely used argument which suggests a minimum of informational DNA in the eukaryote genome (less than 10% of the available DNA) states that the mutational genetic load would be inordinate if mammals used all their DNA to carry and transmit biological information. Ohno 4 states that with a mutation rate of 10 -5 in mammals containing enough DNA for 3×10 6 genes, if all this DNA were informative, each gamete would contain 30 new mutations, which would produce a genetic load sufficient to have exterminated mammals years ago. Evaluation of these mutational and substitutional load restrictions on functional gene number depends upon the unresolved question of the selective neutrality of gene substitutions, and will be treated from both perspectives. 中文
If one accepts that the majority of gene substitutions and polymorphisms are selectively neutral, then the restrictions imposed by a genetic load on functional gene number become negligible. Neutral gene substitutions certainly cannot contribute to any accumulating substitutional or mutational load which depends upon selective disadvantage for its action. We must therefore estimate whether the number of functional genes are minimal, or rather that most gene substitutions are inconsequential with respect to natural selection. Proponents of selective neutrality feel that most substitutions are neutral, which removes any restrictions on large numbers of functional gene loci. 中文
There have been serious objections raised concerning the role of selective neutrality 18,19 . One of the weakest tenets of this hypothesis is that it is based very heavily on the multiplicative aspect of fitness, which assumes that selection acts independently and in an additive fashion over all loci in a population. That this is not the case has been argued cogently by several authors 20-22 . The main point is that selection acts on the whole organism, not on the genotype at each polymorphic locus in each organism in a population 22 . If multiplicative fitness is an unrealistic assumption, then besides questioning selective neutrality as a major force, it also removes the restrictions imposed by the mutational and substitutional load on the number of functional genes. 中文
There are a number of ways, suggested by myself and others, that a population can escape the rigours of multiplicative fitness, or more specifically, immediate selective consequence. These include diploidy 23 , epistasis 20,21 , synonymous base substitutions 7 , frequency dependent selection 24 , linkage disequilibrium 25 , and alternative metabolic pathways (Table 1). All these factors, because they can effectively shield new mutations from the rigours of natural selection, even though the mutations may be deleterious in another genetic environment, counter the assumption of multiplicative fitness. If this assumption is removed, so also is the necessity of restrictive genome size in Drosophila and mammals. 中文
Long sequences (150-300 nucleotides) of polyadenylic acid are generally attached to messenger RNA in eukaryote cells 26-28 . Although post-transcriptional addition of poly A to messenger RNA has been postulated 29-31 , the presence of poly T of comparable length in the nuclear DNA suggests transcriptional addition also 32 . RNA-DNA hybridization kinetics show that up to 0.55% of mammalian nuclear DNA anneals with poly A, corresponding to 1.1% poly dA-dT sequences 32 . This suggests a minimum of 5×10 4 poly dA-dT sites. If each of these sequences is transcribed with an adjacent structural gene, the number of functional genes must be greater than 5×10 4 by the addition of post-transcriptionally added poly A messages, plus non-messenger RNA genes, plus all non-transcribed regulatory genes. This number may be considerable. 中文
In the cellular slime mould, Dictyostelium discoideum , 28% of the nonrepetitive nuclear genome is represented in the cellular RNA during the 26 h developmental cycle 33 . If only one of the complementary strands of DNA of any gene is transcribed, the estimate represents 56% of the single copy DNA. Because the nonrepetitive genome size of Dictyostelium contains approximately 3×10 7 nucleotide pairs 34 , there are at least 16,000 to 17,000 RNA transcripts of average gene size (1,000 nucleotides) present over the cell cycle. Similarly, 10% of the mouse single copy sequences are represented in the cellular RNA of brain tissue. This hybridization result implies that a minimum of 300,000 different sequences of 1,000 nucleotides each are present in the mouse brain alone 35 . Results of RNA-DNA annealing experiments with Drosophila larval RNA indicate that between 15-20% of the unique nuclear genome is represented in larval RNA (R. Logan, personal communication), which corresponds to 30,000-40,000 RNA gene transcripts of average length. As the mouse and Drosophila data include only certain tissues and developmental times respectively, they probably are underestimates of the total unique DNA transcribed by 10-30%, based upon the degree of differences in RNA sequences exhibited at various developmental stages in Dictyostelium . 中文
Interpretation of DNA-RNA hybridization experiments as an estimation of functional genes could be argued to be invalid because a large proportion of cellular RNA is the rapidly degraded “heterogeneous nuclear RNA” which never leaves the nucleus for translation 26,28,36 . RNA does not have to be translated to have a function, indeed RNA has a number of functions other than translation. Three points support gene function of such RNA when considered together: first, the actual presence of the gene; second, the transcription of information, and third, the transcription of different non-repetitive sequences at different developmental times and in different tissues 33,35 . 中文
The major arguments supporting the contention that much of eukaryotic DNA is neither transcribed nor functional are based essentially on the response of the functioning genes to natural selection. The tremendous amounts of physiological and/or genetic compensatory mechanisms which defer the presumed deleterious effects of mutations make such arguments subject to re-evaluation. Furthermore, the molecular data with the poly A sites and RNA transcript estimates suggest greater amounts of gene action than have been presumed. 中文
Although it is impossible to measure exactly the number of functional genes in eukaryotes, the acceptance of evidence for these minimum amounts seems a little premature. 中文
Supported by a postdoctoral award from the National Institute of General Medical Science. 中文
I thank Drs. R. J. MacIntyre, W. Sofer, R. C. Getham, J. Bell, and M. Mitchell for criticism and discussion. 中文
( 242 , 52-54; 1973)
S. J. O’Brien
Gerontology Research Center, National Institute of Child Health and Human Development, National Institutes of Health, Baltimore City Hospitals, Baltimore, Maryland 21224
Received August 28; revised December 11, 1972.
References:
奥布赖恩
编者按
在人类基因组计划定位了数万个人类基因组中的蛋白质编码基因之前,30年间关于真核生物的基因数目存在很多争议。基因的总数被认为远少于单倍体基因组中DNA的数量,这导致一些人提出90%以上的真核生物基因组是没有功能的或者是“垃圾”。在本文中,遗传学家斯蒂芬·奥布赖恩质疑了这个假说,指出垃圾DNA实为功能基因对自然选择的反应的证据。现在认为非编码DNA占据了人类基因组的大部分,但是“垃圾”一词要谨慎使用,因为已经发现了一些所谓的“垃圾”序列的功能。 英文
近来有许多研究都着眼于构建生物模型系统来充分描述真核生物发育过程中基因行为的调节 [1-5] 。就一个单倍体基因组所能够容纳的DNA量来说,已发现哺乳动物和果蝇中的基因数目要比其少1到2个数量级 [2-7] 。以1,000个核苷酸对构成一个基因来计算,尽管果蝇和哺乳动物细胞核含有足够的非重复DNA分别形成10 5 和10 6 个基因 [8,9] ,但是看来对功能基因的数量更低的估计是更合理的 [2-7] 。反过来说,这些结论则提示真核生物基因组超过90%的DNA可能由非功能性或者不编码信息的“垃圾”DNA组成。本文中我们的结果显示这些估计都没有得到有力地证明,相反它们都是基于一些简化的假设所获得,而这些假设本身的正确性值得怀疑,其中有些甚至与实验数据相矛盾。 英文
克里克 [5] 提出的模型指出,在果蝇巨大唾液腺染色体中,编码蛋白质的结构基因通常都位于观察到的染色体条带之间。调控元件和大量含有无编码信息的DNA序列位于染色体条带上,这些染色体条带包含了绝大多数DNA。据此该模型预测果蝇大约有5,000个结构基因,这也是能够观察到的唾液腺染色体带的大致数量。 英文
该模型得到了贾德等人的出色工作的有力支持 [10] 。他们研究了黑腹果蝇位于X染色体末端 zeste 到 white 区域内的121个致死性和显著影响形态的点突变。该区域内一共有16个唾液腺染色体带或染色粒,与16个形态学改变或致死性点突变的互补群互为对应。此外一系列重叠的缺陷支持一个条带一个互补群的关系。外推至整个基因组,大约5,000个这种互补群或者基因对应于5,000个染色体带。也有人估计了果蝇基因组中致死基因座的总数。通过筛查自然群体中或者经过辐射处理后的群体中的大量可致死染色体,就有可能通过简单的泊松分布将等位性的频率与致死基因座的数目联系起来,这样得到的果蝇中致死基因座的数目大约是1,000~2,000个 [11,12] 。 英文
在这种利用精细结构分析和致死性数据外推到功能基因数目过程中存在一个我们无法回答的问题:有多少基因突变后能够产生致死性的或者显著影响形态的表型?这无法明确地回答,但是已有数据提示所有基因产物中只有很少一部分重要到其缺失能够导致生物的死亡。在果蝇中,超过30个基因的基因产物是已知的 [13] ,其中只有14个在具有“无效”等位基因时能够导致蛋白质及其活性或RNA产物的完全缺失,而这其中(表1)只有 bobbed 基因座具有致死性的等位基因 [14] 。然而,这个基因座是核糖体RNA(rRNA)的结构基因,其大多数等位基因即使在所产生的rRNA浓度非常低时也能有活力。其他编码酶的基因即使它们对于正常代谢是必需的,在纯合的完全“无效的”等位基因中也没有一个是致死性的。 英文
表1. 具有确定基因产物和可恢复性“无效”等位基因的黑腹果蝇基因
* 扬,个人交流。
上表中前11个基因座中的无效等位基因是通过凝胶电泳组织化学染色条带丢失检测出来的。这种方法的敏感性至少能够检测出正常酶水平5%的量。在一些例子中( Acph-1 、 rosy 、 Adh 、α Gdph-1 ),灵敏度接近0.1%野生型酶水平的酶分析法也不能在“无效”纯合子中检测到痕量的酶活性。在通过交叉反应物质(CRM)检测的两个例子中( Acph-1 和 ry ),测到的酶活性也是微乎其微。 英文
在杂交实验中,无效等位基因中的至少两个基因座被诱导而恢复了致死性(α Gdph-1 和 Acph-1 )。在自然或实验室群体中发现的“无效”等位基因中( Est-C 、 Est-6 、 Aph 、 Aldox 和 Idh ),也能检测到一种致死性“无效”等位基因,这种等位基因与有着不同电泳迁移率的正常等位基因形成一个异常杂合子。 英文
14个基因座中有5个具有能够产生可见的隐性表型的等位基因,其中 ry , cn 和 v 影响眼睛的颜色, bb 影响刚毛,而α Gdph-1 “无效”突变使果蝇尽管在形态学上表现正常,但丧失了持续飞行的能力。但是5/14这个比例并不能作为产生可见表型的“无效”等位基因在基因座中所占比例的估算值。因为所研究的5个基因座中的4个(除了α Gdph-1 )最初就是作为形态学的突变而被发现的,其基因产物已经从它们的可见表型中推断和鉴定出来,所以这个数值很可能是被高估的。 英文
眼睛颜色相关基因的突变会影响眼色素生物合成过程中的相关酶,而通常显示产生一系列蛋白质合成相关症状的 bobbed 基因座已被确定为rRNA基因。如果不是以前就清楚α Gdph-1 产生的酶在昆虫飞行中的重要性 [15] ,就很容易忽视α Gdph-1 的“无效”突变的表现型。另外11个基因座都仅仅是被鉴定为特定酶的基因,并没有发现其“无效”等位基因具有致死性或者导致任何形态表型的改变。 英文
有两个构建了双重“无效”突变体的例子,碱性和酸性磷酸酶突变体(麦金太尔,个人交流)以及 Zw 和 6-Pgd 突变体(扬,个人交流),它们均被证实能够生存、繁殖并且形态上正常。同时,在5个具有可见表型的例子中, bb 和α Gdph-1 这两个突变的受累动物出现了表型修饰。在 bb 中,减少的rDNA顺反子被“放大”,以至在数代内达到野生型rRNA的水平 [16] 。遗传上缺乏α–甘油磷酸脱氢酶(α Gdph-1 )的果蝇由于α–甘油磷酸循环被破坏而丧失了持续飞行的能力 [15] ,但是经过25代以后,这种表型发生了变化而且果蝇恢复了正常飞行的能力(奥布赖恩,未发表数据)。对具有重要功能的结构基因损伤进行生理补偿的生物学适应能力一定非常广泛,使得果蝇能相当有效地免受这些遗传敏感的基因座的影响,即使是那些功能被假定很关键的基因座。 英文
有人可能说大多数情况下即便是最小的细胞学可见突变都是隐性致死的 [10,14,17] 。但是要确定这种细胞学上的改变,所需要的分辨率是至少能检测到果蝇多线染色体中5,000个染色粒中的一个发生了缺失。这个技术精确度是在10 6 个核苷酸水平,差不多是一个染色粒的平均大小,即足以构成20个平均长度的基因的DNA。我认为每个区域可有多达20个功能基因,在其突变谱中可能只有一个是致死性突变。 英文
另一个被广泛采用的论点提出了真核生物基因组中信息DNA的最小量(小于可获得的DNA的10%),并且认为如果哺乳动物用所有的DNA来携带和传递生物学信息的话,那么突变了的遗传负荷就会过度。大野 [4] 认为对含有足够组成3×10 6 个基因的DNA的哺乳动物来说,如果突变率是10 –5 ,而且所有的DNA都是携带信息的,那么每个配子会含有30个新的突变,这样的遗传负荷会使得哺乳动物在很多年前就灭绝了。评估这些突变和替换的负荷所限制的功能基因数目取决于基因替换的中性选择这个尚未解决的问题,而且需要从两方面进行考虑。 英文
如果接受大部分基因替换和多态性都是中性选择的,那么遗传负荷对于功能基因数目的限制就变得微不足道了。中性的基因替换当然不会有助于替换负荷或突变负荷的累积,因为这些负荷的作用是由其选择劣势决定的。因此我们必须估计是否有个最少的功能基因数目,或者更确切地说,大部分基因替换不是自然选择的结果。中性选择的支持者们觉得大部分替换都是中性的,这就消除了突变对于大量功能基因座的任何限制。 英文
对于中性选择的作用曾有很多严肃的反对意见 [18,19] 。该假设最薄弱的一条是它很大程度上基于对适应的倍增性,即假设选择的作用是独立的并且以加和的方式在群体中所有的基因座发挥作用。数名作者已经中肯地指出事实并非如此 [20-22] 。要点在于选择是作用于整个生物体,而不是群体中每个个体的每个多态位点的基因型 [22] 。如果倍增性适应是一种不现实的假设,那么除了对于中性选择作为主要作用力的质疑,突变和替换负荷对于功能基因数目的限制也可不再考虑。 英文
包括我本人在内,很多人认为一个种群有多种方法摆脱所谓倍增性适应,或者更确切地说是直接选择的结果。这些包括二倍性 [23] ,上位显性 [20,21] ,同义碱基替换 [7] ,频率依赖的选择性 [24] ,连锁不平衡 [25] 和替代代谢通路等(表1)。由于这些因素可以有效地保护新的突变免于遭受严酷的自然选择,尽管突变在其他遗传环境中可能是有害的,这些因素也可以与倍增性适应的假设相抗衡。如果不考虑这个假设,那么果蝇和哺乳动物中也不必考虑对于基因组大小的限制。 英文
真核细胞中的信使RNA一般连有很长的(150~300个核苷酸)多聚腺苷酸序列 [26-28] 。尽管人们假设多聚腺苷酸添加到信使RNA末端发生在转录后 [29-31] ,细胞核DNA中存在类似长度的多聚胸苷酸提示转录时添加也是可能的 [32] 。RNA-DNA杂交动力学显示高达0.55%的哺乳动物核DNA退火后结合多聚腺苷酸,对应于1.1%的多聚脱氧腺苷酸–脱氧胸苷酸序列 [32] 。这提示至少有5×10 4 个多聚脱氧腺苷酸–脱氧胸苷酸位点。如果这些序列的每一个都与邻近的结构基因一起转录,再加上转录后加入的多聚腺苷酸序列、非信使RNA基因以及所有的非转录性调节基因,那么功能基因数肯定会超过5×10 4 。这个数目可能是相当可观的。 英文
在细胞型黏菌盘基网柄菌阿米巴虫中,28%的非重复性核基因组在26小时的发育周期中表达为细胞RNA [33] 。如果任何基因中只有互补DNA链中的一条被转录,那么此估计值代表56%的单拷贝DNA。因为黏菌的非重复性基因组含有将近3×10 7 个核苷酸对 [34] ,所以在细胞周期中存在至少16,000到17,000个平均基因大小(1,000个核苷酸)的RNA转录产物。与之类似,小鼠脑组织中10%的单拷贝序列表达为细胞RNA。这个杂交结果提示单独在小鼠脑组织内部可能存在至少300,000个平均拥有1,000个核苷酸的不同序列 [35] 。用果蝇幼虫RNA进行的RNA-DNA退火实验显示15%~20%的特异性核基因组都表达为幼虫RNA(洛根,个人交流),相当于30,000~40,000个平均长度的RNA基因转录产物。由于小鼠和果蝇的数据仅仅分别包括了特定的组织和发育阶段,所以根据黏菌不同发育阶段RNA序列差异的程度,很可能将转录的特异DNA总量低估了10%~30%。 英文
将DNA–RNA杂交实验作为对功能基因数量的估计的解释可能会被认为不可靠,因为细胞RNA的大部分都是很快降解的“核不均一RNA”,它们从不离开细胞核去进行蛋白质翻译 [26,28,36] 。但是RNA并不是一定要翻译成蛋白质才具有功能,事实上RNA除了翻译之外还有很多功能。综合考虑有三点可以支持RNA的基因功能:第一,基因的实际存在;第二,信息的转录;第三,不同发育阶段和不同组织中不同的非重复序列的转录 [33,35] 。 英文
主流的观点认为大部分的真核DNA既不用于转录,也非功能性,支持这一论点的主要依据是功能基因对于自然选择的反应。大量延缓了突变可能产生的有害作用的生理和(或)遗传补偿机制使得我们需要重新评价这个论点。此外,针对多聚腺苷酸位点和RNA转录物估算的分子数据显示比预计数量更多的基因活动的存在。 英文
尽管不可能精确测定真核生物功能基因的数量,但接受这些最小数量的证据似乎还有些为时过早。 英文
本研究由国立综合医学研究所的博士后奖金资助。 英文
感谢麦金太尔、索弗、盖塞姆、贝尔、米切尔博士的意见和讨论。 英文
(毛晨晖 翻译;曾长青 审稿)
M. Barazangi et al .
Editor’s Note
At subduction zones, one tectonic plate plunges down beneath another into the Earth’s mantle. The descent of the lower plate can trigger deep earthquakes. Bryan Isacks of Cornell University, together with Peter Molnar, suggested that a “gap” in seismic activity below a subduction zone might indicate that part of the descending slab had broken off from the rest. Here Isacks and coworkers present evidence that this has happened in the subduction zone containing the New Hebrides islands in the southwest Pacific, and most probably in that of New Zealand. They also suggest that the New Zealand and Tonga-Fiji slab cannot penetrate below 700 km, owing to a “discontinuity” in the mantle that subsequently became a major focus of geophysical studies. 中文
Study of seismic wave propagation in the mantle beneath the New Hebrides island arc shows that the remarkable gap in seismic activity between deep and intermediate depth earthquakes at the northern part of the arc corresponds to a gap in the lithospheric slab descending beneath the arc: the deep earthquakes mark a detached piece of lithosphere. Although observations for New Zealand deep earthquakes are ambiguous, other evidence suggests the detachment of lithosphere beneath New Zealand. 中文
ONE of the outstanding features of the distribution of earthquakes in the upper mantle is the existence of gaps in seismic activity between depths of about 300 and 550 km. These gaps are prominent beneath South America, New Zealand, and the New Hebrides island arc. They are of great interest because of the implication that portions of lithosphere can break off from the descending plate and exist as isolated slabs within the mantle. 中文
Here we present evidence from study of seismic wave attenuation for the detachment of lithospheric slabs in the upper mantle beneath the New Hebrides arc. Detachment is also indicated by travel time data and reconstructions of the past movements of plates in the region. We concentrate on the study of seismic attenuation because the effects are large and easy to observe and interpret 1-3 . 中文
Isacks and Molnar 4 suggest that the gaps in seismic activity as a function of depth can be explained by two models. In the first the stress inside a continuous slab varies from down-dip extension at intermediate depths to down-dip compression at greater depths, and thus is near zero between these depths. In the second a piece of descending lithosphere breaks off, sinks into the upper mantle and leaves a gap between the piece and the plate to which it was attached. The character of seismic waves, especially shear waves, produced by deep earthquakes and recorded at stations along island arcs where large gaps in seismic activity exist provides evidence for determining which of these models is correct. Observation of attenuated, low frequency waves for the appropriate paths indicates a gap in the descending lithosphere. But if high frequency shear waves are observed, then the interpretation is ambiguous. The observations can be explained either by propagation along a continuous high Q lithospheric slab or by propagation through the upper portion of a discontinuous slab that provides a low attenuation path through the asthenosphere. The second explanation implies that for 1-3 Hz shear waves attenuation is small below depths of about 300 km. 中文
The New Hebrides island arc provides a unique opportunity to study the nature of the gap in seismic activity in the upper mantle. Seismograph stations are well distributed along the arc. The gap is well established between depths of about 300 and 600 km. The structure of the intermediate and deep seismic zones with the very steeply dipping intermediate zone and the approximately horizontal deep zone certainly suggests that the two zones may not be connected. Enough deep earthquakes have occurred since the stations were installed to provide a large sample of paths. Forty deep earthquakes in the 10 year period 1961-1970 were recorded at Noumea (NOU) and Port Vila (PVC) stations (Fig. 1). Lonorore (LNR) was established in 1968. NOU, PVC and LNR employ 1 Hz underdamped seismometers and 2 Hz overdamped galvanometers. The instruments have a relatively flat response for seismic wave frequencies between 0.5 Hz and 10 Hz and thus record very clearly the large variations of shear wave frequencies. Detailed description of the New Hebrides-New Caledonia seismic network is given by Dubois 5 . 中文
Fig. 1. Map showing the Tonga-Fiji-New Hebrides region of the southwest Pacific. ●, Historically active volcanoes; ▲, seismic stations; ■, locations of deep earthquakes used in Figs. 2 and 3. Water depths are in km.
The inclined seismic zones of South America also have remarkable gaps in seismic activity between about 350 and 550 km. A detailed study of the records produced by almost all the South American deep earthquakes that occurred during the past 10 yr at stations along the western coast of South America is currently under way, and will be reported in a separate study. 中文
Fig. 1 shows the location of the stations in New Caledonia and New Hebrides used in this study. We have examined all the records produced at these stations by the New Hebrides deep earthquakes. The most striking observation is that predominantly low frequency (about 0.5 Hz) S waves are recorded at PVC and LNR from the deep earthquakes north of 15° S. 中文
Fig. 2 shows a cross-section through the New Hebrides arc that intersects NOU, LNR and passes close to PVC. New Hebrides deep earthquakes located at the western part of the deep zone (very close to the downward projection of the intermediate depth zone) produce attenuated, low frequency S waves at PVC. The ray paths pass just beneath the dipping seismic zone. Frequencies greater than 1 Hz are absent and the amplitude of the S phase is generally less than that of P phase. As Oliver and Isacks 1 and Barazangi and Isacks 3 show, this can be explained by a transmission through an attenuating low Q zone. In contrast, S waves recorded at NIU station on the Tonga island arc from Tongan deep earthquakes have predominant frequencies of 3-4 Hz and the amplitudes of S are generally larger than those of P. 中文
Fig. 2. Cross-section of the New Hebrides arc showing J-B ray paths to NOU, PVC and LNR stations and the corresponding records [E-W component at PVC, N-S component at NOU, and Z component at LNR] and the locations of seismic activity (vertical lines). Only low frequency S waves are recorded at the stations.
Deep earthquakes also produce low frequency shear waves at LNR, a station close to the active volcanic line of the New Hebrides arc. This is in marked contrast to the observation of high frequency S waves from Tonga deep earthquakes at stations along the active volcanic line of the Tonga arc 6 . 中文
We interpret the observation of low frequency S waves at PVC and LNR to be mainly the result of attenuation along the path. The effect of the source can be excluded since New Hebrides deep earthquakes produce seismograms at Fiji stations that are similar to those produced by the Tonga deep earthquakes (the New Hebrides and Tonga deep earthquakes are approximately equidistant from the Fiji stations). The effect of the station can be excluded since intermediate earthquakes located along the New Hebrides arc produce, without exception, high frequency (about 3-4 Hz) S waves at PVC and LNR. This is so even where the path lengths are comparable or greater than those from the deep earthquakes. Further, the Tonga deep earthquakes produce seismograms at PVC and LNR that are strikingly similar to those produced at the Fiji stations. Thus, attenuation along the path is the main cause for the observed low frequency S waves at PVC and LNR. 中文
The attenuation is chiefly below 300 km, judging by the abundant observations of high frequency shear waves from intermediate depth earthquakes to PVC and other New Hebrides stations. All observations taken together are best explained by the absence of lithospheric slab material between depths of about 300 and 600 km; the deep earthquakes of New Hebrides therefore represent a detached slab in the upper mantle. To our knowledge, this is the first direct evidence that the attenuating, asthenospheric layer is deeper than about 300 km in the upper mantle. 中文
New Hebrides deep earthquakes always produce low frequency S waves at NOU (Fig. 2). This is most probably due to attenuation in the upper mantle (the asthenospheric layer), because the ray paths to NOU completely miss the dipping New Hebrides seismic zone. 中文
New Hebrides deep earthquakes located at the eastern part of the deep zone produce attenuated, low frequency S waves at PVC and LNR. The ray paths, calculated for a laterally homogeneous mantle with a Jeffreys-Bullen (J-B) velocity structure, pass just above the inclined seismic zone (Fig. 3). S phases which similarly appear to pass above the Tonga inclined zone, however, have large amplitudes and high frequencies (we note that the time scale of NIU record is about twice that of PVC, as shown in Fig. 3). Even though the J-B ray path seems to miss the Tonga seismic zone, the high frequency S waves probably travel through the descending slab. Barazangi, Isacks and Oliver 7 describe other evidence that the slab descending beneath Tonga acts as a wave guide for high frequency shear waves and is therefore continuous. Thus by comparison the absence of these high frequency shear waves for the easternmost New Hebrides deep earthquakes can be taken as evidence for the detachment of lithosphere beneath the New Hebrides. 中文
Fig. 3. Two cross-sections of Tonga and New Hebrides arcs showing ray paths to NIU and PVC stations and the corresponding records [E-W component at PVC, and Z component of P, and N-S component of S at NIU] and the locations of seismic activity in the upper mantle (vertical lines). Note the great difference in the signature of S waves at NIU and PVC in spite of the similarity in ray paths (time scale of NIU record is about twice that of PVC).
We will present a detailed study of the travel times of P waves of the New Hebrides deep earthquakes later. The travel time residuals of P waves of the New Hebrides deep earthquakes at PVC and LNR stations along the arc are close to normal (about 1 to 2 s earlier). This is in contrast to P residuals of about 4 to 5 s earlier from the Tongan deep earthquakes recorded at stations along the Tonga arc 8 . Thus the travel time data support those obtained from seismic wave attenuation and indicate that the deep earthquakes at the northeast of the arc represent a detached lithospheric slab. 中文
During the past seven years three deep earthquakes occurred south of 15° S, south of the horizontal oblong-shaped zone, and are located along a line parallel to the southern part of the New Hebrides arc (Fig. 5). Records produced by these earthquakes at NOU in New Caledonia and at stations located in the northern part of the New Hebrides arc (to the north of about 16° S latitude, LNR and LUG) show attenuated, low frequency S waves. But records produced at PVC from the most recent event located at about 18° S and 173° E show large amplitude, high frequency S waves. This is the only one of the three shocks recorded at PVC with good quality records. This observation is quite clear, however, and may imply the continuity of the descending slab in the southern part of the New Hebrides arc. More data are required before a meaningful interpretation can be made for the southern deep earthquakes zone. 中文
In New Zealand earthquakes reach a depth of about 300 km in the North Island and about 200 km in the northernmost part of the South Island. In addition three earthquakes occurred at depths of about 600 km in 1953 and 1960 beneath the North Island 9 . Fig. 4 shows a cross-section of the New Zealand arc and examples of seismograms from the local New Zealand network. Deep earthquakes produce high frequency S waves at Wellington (WEL). Two quite different explanations can be made for this. One is that the slab beneath New Zealand is continuous and reaches depths of at least 600 km, and thereby provides a path for high frequency S waves. The second is that although the slab may have a gap beneath about 300 km, the portion above 300 km is sufficient to provide a “window” through the zone of high attenuation. This second explanation implies that the principal zone of attenuation is located above 300 km, and thus implies a significant difference between the New Zealand and New Hebrides regions with respect to attenuation below 300 km. 中文
Fig. 4. Cross-section of the New Zealand arc showing ray paths to WEL, TNZ and ONE stations, horizontal records to WEL, ONE and GPZ stations, and the dipping seismic zone beneath the North Island (vertical lines). Insert map shows the locations of stations (●) and the location of the deep earthquakes of New Zealand (■).
The second alternative, that a detached lithosphere is present beneath the North Island of New Zealand, is supported by the variation of down-dip length of the inclined seismic zone as a function of latitude along the New Zealand—Kermadec—Tonga plate boundary. If New Zealand deep earthquakes are excluded, the down-dip length increases regularly northward as predicted by locations of the pole of relative motion between the Australian and Pacific plates 10,11 . The New Zealand deep shocks are thus distinctly anomalous in this respect and seem to mark a detached piece of plate. 中文
New Zealand deep earthquakes produce low frequency, attenuated S waves at Tarata (TNZ) and Onerahi (ONE) which are located to the west of the line of active volcanoes in the North Island. Mooney 12 mapped a zone of anomalously high attenuation in the uppermost mantle also located west of the active volcanoes. Thus, the low frequency S waves at TNZ and ONE are probably the result of attenuation in the uppermost mantle to the west of the dipping seismic zone. 中文
An interesting observation is that New Zealand deep shocks produce high frequency S waves at stations in the South Island. Fig. 4 shows an example recorded at Gebbies Pass (GPZ), a station located along the aseismic 13 east coast of the South Island. This observation suggests that beneath at least the northern part of the South Island no major zone of attenuation is present. The lithosphere marked by intermediate depth earthquakes beneath the North Island may thus extend beneath part of the South Island. 中文
An unusual feature of the earthquake distribution in the Tonga-Fiji region is the occurrence of deep earthquakes west of the inclined seismic zone of the Tonga arc (Fig. 5). In the past 10 yr about 8 well located deep earthquakes occurred beneath the Fiji Islands. It is not clear whether these earthquakes represent a continuation of the descending Tonga slab, a slab(s) detached from the present descending slab, or a slab detached during an earlier episode of underthrusting. Evidence obtained from focal mechanisms by Isacks et al . 14 suggests that these earthquakes represent a contorted continuation of the northern edge of the descending Tonga slab. In any case these earthquakes and the New Hebrides deep earthquakes show a considerable horizontal extent away from the corresponding descending slabs. This suggests that the earthquakes mark lithospheric slabs that are unable to penetrate the 600-700 km discontinuity of the upper mantle, and hence the discontinuity may represent the lower limit of the asthenosphere. These slabs may pile up above the discontinuity until their assimilation in the mesospheric lower mantle. 中文
Fig. 5. Map showing contours of earthquake depths for Tonga and New Hebrides arcs, the deep seismic zone of New Hebrides, and well located deep earthquakes beneath the Fiji islands and the Fiji plateau. ▲, Events with depth range between 525 and 575 km; ●, events with 576 to 625 km depth; ■, events with 626 to 675 km depth. The open circle represents an event with a depth of 470 km.
The New Hebrides and New Zealand deep earthquakes mark detached pieces of lithosphere in the upper mantle; considerable seismic wave attenuation probably exists below about 300 km of depth in the upper mantle beneath the northern part of the New Hebrides arc, which implies that the asthenosphere may extend deeper than 300 km in the mantle in this region; the spatial distribution of deep earthquakes between the New Hebrides and Tonga arcs suggests that slabs of lithosphere are unable to descend beneath about 700 km in the mantle. 中文
We thank R. D. Adams for original seismograms, Walter Mitronovas for discussions, and Jim Gill and Dan Karig for preprints of their research. M. B. thanks Maurice Ewing for a research grant. 中文
This work was supported by National Science Foundation grants. 中文
( 242 , 98-101; 1973)
Muawia Barazangi * , Bryan L. Isacks * , Jack Oliver * , Jacques Dubois † and Georges Pascal †
* Department of Geological Sciences, Cornell University, Ithaca, New York 14850
† Office de la Recherche Scientifique et Technique, Outre-Mer, Noumea, New Caledonia, and Institut de Physique du Globe, Université de Paris, 6 Paris
Received October 24, 1972.
References: