【作者】 谭晓东；

【导师】 肖中举；

【作者基本信息】 南方医科大学 ， 神经生物学， 2008， 博士

【摘要】 哺乳动物的听觉有着令人难以置信的能力。高度的灵敏度,高度的精确性,亚毫秒级的快速反应能力及10⁶数量级的动态听觉范围,在听觉系统中达到了完美的统一。然而产生或者维系这些似乎难以兼顾的能力的听觉中枢信息处理机制,尤其是基本的编码机制却一直不清。这集中体现在大量以动作电位发放数（spike counts,SC）和/或以第一个动作电位延时（first spike lantency,FSL）来表征声音参数的研究中,究竟谁编码了声音信息,或者说他们分别编码了声音信息的哪些参数,至今未见统一的标准。此外,SC及FSL均受多个声音刺激参数的影响。这些参数之间是否存在相互关系,听觉神经元又是如何编码和携带这些不同的信息,这些方面的工作目前也少见相关的报道。本研究主要包括两方面内容:一是系统地比较SC和FSL表征声音频率和强度时的精确性和稳定性,以确定哪个参量更适合作为声音刺激的编码机制及参量间的相互关系;二是在确定能更好地表征声音刺激参数的参量后,进一步研究其表征的声音参数（频率和声强）之间的相互关系,以探讨FSL作为听觉中枢编码的可能机制。声音信息的编码机制在对听觉功能和机制的研究中占重要地位。如上所述,有两个参量通常用来作为听觉中枢神经元对声音频率和强度编码的可能机制,一个是SC,也叫发放率,另一个是FSL。听觉神经元的SC—强度曲线可分为三种类型:单调型（SC随着刺激幅度的升高而增加）,非单调型（发放率先随着刺激幅度的升高而增加,到一定水平后又随着幅度的进一步升高而减少）,饱和型（SC随刺激幅度的升高而增加,当增加至某一水平后不再随幅度的升高而变化而呈平台状）。这三种类型的强度-发放率曲线在听觉传导通路的各级中枢,如耳蜗核,上橄榄复合体,下丘,内侧膝状体,听皮层都可记录到。FSL一般都随着刺激幅度的升高而缩短。但也有例外的报道发现FSL可有不规则变化,如随着刺激幅度的升高而延长（paradoxical latency）,或者FSL并不随着刺激幅度的改变而改变而呈现出稳定延时（constant latency）。用SC对刺激声音的频率作图,可得到倒“V”字型的呈单峰状的功能曲线。引起神经元最大发放数的频率定义为该神经元的最佳频率（best frequency,BF）。不同刺激强度下的发放数-频率曲线和BF都是相对可变的,特别是当刺激强度较大的时候;而且BF通常也并不和特征频率（characteristic frequency,CF）重叠,这一现象在声刺激强度在阈强度以上10dB时尤为显著。FSL-频率曲线的形状与发放率-频率曲线恰恰相反,呈正“V”字型,其延时最短处所对应的频率往往就是该神经元的特征频率或在特征频率的附近,且随着刺激频率和CF之间的差距增加,FSL也越来越长,且频率相差越大,FSL延长的幅度也越大。但这两个参数中,哪个作为对声音的频率和强度编码的机制更合适目前并不清楚,一直以来,对它们之间和声音刺激的强度和关系的系统比较研究也鲜有报道。我们在小鼠下丘用在体细胞外记录的方法记录了听觉神经元对声音刺激的反应。将一约1.5 cm长的平头金属钉用齿科水泥竖直粘在小鼠头骨表面,将动物置于一泡沫塑料制成的模子中用弹性绷带立体固定好,模子悬吊于一个支架上。保持动物双耳不被遮挡以接受自由声场的声刺激。在下丘部位的头骨上开孔,暴露下丘部位的脑组织。通过Tucker-Davis Technologies系统3来产生短纯音作为声刺激,其时程（duration）为50ms,上升一下降时间（rise-fall time）为5ms。扬声器固定在动物头部正前方30 cm处,通过计算机用Brainware软件控制和改变声刺激的强度和频率。用玻璃微电极进行细胞外记录,记录结果用TDT 3采集和分析。在下丘听神经元根据其动作定位的发放被找到和分离出来后,通过改变短纯音声刺激的频率和强度以确定该神经元的CF和最小阈值（minimum threshold,MT）。接着用不同的频率-强度（frequency-amplitude,F-A）进行声刺激扫描（F-A scan）。声音刺激诱发的动作电位的波形,数目和峰电位的时间特性被电脑采集并整理。记录过程中,可同时用BrainWare软件监测SC水平,频率,刺激后放电直方图（post-stimulus time histograms,PSTH）,动作电位波形。在以上记录的基础上,再对数据进一步分析,对神经元的SC,FSL和频率及幅度的关系进行作图,并拟合分析SC及FSL的标准差（standard deviation,SD）和变异系数（coeffient ofvariation,CV）与频率和幅度的关系。我们共记录了163个下丘听觉神经元。数据分析处理的结果显示:1)从两者和声音刺激强度的对应关系来看,SC—强度关系曲线呈多型性,21个（占总数13%）呈单调型,49个（占总数29%）呈饱和型,95个神经元（占总数58%）呈非单调型。而FSL—强度关系曲线更为一致,表现为FSL随着声强的增加而缩短,且未发现有异相神经元（声强增加到一定幅度后,FSL反而随声强的进一步增加而延长）或恒定延时神经元（声强增加到一定幅度后,FSL不再随声强的进一步增加而缩短）;2)我们用one-way ANOVA法检测了每个神经元所有相邻两个幅度之间的反应（SC或FSL）是否有显著差异。在神经元SC-幅度曲线中仅有少数（2-3个）相邻数据点之间呈显著差异（见正文中图2.3A1,B1 and C1和表2.3A,B和C）,而FSL-幅度曲线中,大部分相邻的数据点都有显著差异（F值分别为134.9,212.2及336.7;P值均小于0.0001）。对SC-频率曲线和FSL-频率曲线的分析也得到了相似的结果。以上分析显示,以FSL表征的声音频率和强度其分辨率比以SC表征时要高;3)为进一步比较CV_SC和CV_FSL,我们对每个神经元都作出其比较图。所有神经元均进行了FSL的CV指数(The CVindex of FSL,CVI_FSL。指在同一神经元上,对幅度或频率做功能曲线,其位点的数量和所有位点的数量的比值)的分析。在163个神经元中,各有151个（93%）神经元在其特征频率下对不同声强反应,或在CV-声强曲线交叉点的声强下对不同频率反应,其CVI_FSL大于0.5,且低声强时CV_FSL值低于CV_SC;4)以MT修正后的幅度（MT+）表示,在MT+5,+10,+20,+60的幅度时,SC的BF与CF相等（在±500Hz的范围内）的神经元比例分别是85%（138/163）,50%（80/160）,32%（46/144）及12%（6/52）;而FSL的BF与CF相等的神经元比例分别是99%,89%,85%及21%。即神经元的调谐曲线中,任何声强下BF和CF相等的神经元所占比例均是FSL多于SC,且对于FSL和SC,这一比例都随刺激强度的增加而减小。综上所述,我们对SC和FSL表征声音频率和强度时的精确性和稳定性进行系统比较的结果表明,SC和FSL应是互相独立的,它们可能分别代表了声音刺激的不同参数。从小鼠下丘听觉神经元的反应和刺激声音的强度和频率关联看来,较之SC,FSL更加精确和稳定,能更好地表征刺激声音的频率和幅度。听觉系统外周（耳蜗）的工作原理相当于多个相互叠加的带通滤波器,由基底膜底部到顶部响应的频率由高到低顺序排列。这种在基底膜不同部位表现出来的频率选择性分布被称为耳蜗拓扑地形（cochleaotopographic）组构。与基底膜相对应,听觉通路的各级中枢也都存在着神经元CF或BF的拓扑结构（tonotopic structure）。神经系统信号处理过程中有几个基本原则,包括信息处理的等级假说,即初级中枢只能完成简单的信号加工,复杂的信息处理在高级中枢完成;及平行的信息处理假说,即感觉系统从外周到中枢有许多并列的信息传递通路,表现出高度有序的拓扑结构,具有能提取不同特征、处理不同信号的功能。平行级层投射（parallel-hierarchical）假说也被用来解释听觉系统对信号的处理过程,各级中枢的频率拓扑结构被认为是其结构基础,但这一假说尚缺乏直接的实验证据。为支持这一假说,需要证明听觉神经元有来自基底膜的滤波特性,即一个神经元对任何频率的声音刺激的反应均由与CF相关的频率决定或声音的共振频率决定,而且,该神经元的滤波特性应自基底膜传递而来。但目前为止对此未见研究。根据上述FSL和SC对声音频率和强度的表征的比较,我们采用FSL作为表征,分析了神经元对CF和对其他频率声刺激的反应之间的关系。实验方法包括手术,声刺激及数据记录等,与上述大致相同。在记录到单峰神经元（即反应—频率曲线呈单峰状）并确定其CF和MT之后,进行F-A扫描,频率变化范围为CF±5 kHz,步阶为1或0.5 kHz。得到的MT,CF,SC,FSL分别对频率及幅度作图,再进一步进行拟合、计算斜率并作出斜率对CF及MT的分布图等处理。实验中我们记录到57个单峰神经元,分析处理的结果显示,1)当声音的上升时间、相位及其他参数固定时,神经元对刺激反应的FSL可精确地表征刺激的频率和幅度,FSL-幅度曲线符合指数方程,可根据Pieron’s经验法则（Pieron’sempirical law）很好地进行拟合（拟合后的R²值范围为0.917到0.998之间）,并且非特征频率声刺激引起的FSL-幅度曲线和特征频率声刺激引起的FSL-幅度曲线相一致。2)下丘单峰神经元显示出滤波特性,表现在不同频率下的FSL—声强曲线基本相似。当FSL相对于经CF同一化（normalize）后的有效声强（即相当于CF的声强减去一个固定值Δx）作图时,所有频率下的FSL-幅度曲线都可以重合,此结果与单个滤波器的工作原理非常一致（详细公式推导见正文）。因此,下丘神经元的这种滤波特性应该来自于基底膜的同一点而不是多个点（即其他频率位点）。3)FSL对声强的传递特性可通过一个指数常数来描述,此常数在高阈值神经元中较大,而与CF无关（相对于CF作线性回归时,R²=0.0042）。我们的研究结果提示:当使用FSL作为表征时,在听觉高级中枢的神经元显示出对应于基底膜某一点的滤波特性,这种对应性通过基底膜的滤波特性实现,即任何频率下的刺激幅度均可转换成相对于共振频率或特征频率的有效幅度,即对应于基底膜的振幅。因此,我们的研究结果为听觉系统平行级层投射的信号处理方式的假说提供了实验证据。此外,上述结果还提示:当其他刺激参数如上升时间,相位等固定不变时,神经元反应的延时可同时表征声音刺激的幅度和频率。更多还原

【Abstract】 The auditory system of mammals has many unbelievable abilities.It integrates perfectly the high sensitivity,high accuracy,quick responsibility within milliseconds or even less,and 10⁶ level of dynamic range of detectable sound amplitude. Nevertheless,the mechanisms of signal processing of auditory system,which should have given rise to or at least maintained these seemingly unconcomitant abilities,are still unclear,especially the basic coding mechanism of auditory system.This uncertainty is demonstrated by the numerous studies focused on the representing of acoustic parameters by spike count and / or spike time,no consistent criteria of which one is the dominant coding mechanism,or which parameter is coded by which one of these two candidates.Furthermore,both SC and FSL were affected by various acoustic parameters such as frequency,amplitude,rise-fall time,duration,phase,etc. So,whether there are any relationships within these parameters,what’s the possible mechanisms underling the coding or representing of different parameters within one mechanism,these studies are also scarce.So the purpose of our study is that,first,to compare systematically the representation of acoustic amplitude and frequency by SC and FSL,to determine which one is more precise and stable to be the possible coding of acoustic stimulation parameter,and second,to further study the possible mechanism underling the temporal coding by investigating the relationship between amplitude and frequency represented by FSL. Spike counts（SC） or,spike rate and fast spike latency（FSL）,are both used to evaluate the responses of neurons to amplitudes and frequencies of acoustic stimuli. The spike count - frequency functions across a population of auditory neurons can be divided into three types:monotonic（the SC increases monotonically as the amplitude of the acoustic stimulus increases）,non-monotonic（the firing of non-monotonic neurons first increases and then decreases with further increases in amplitude）,and saturated（the SC-amplitude function reaches a plateau with increasing amplitude）. These three types are found at almost every central auditory level,e.g.,the cochlear nucleus,superior olivary complex,inferior colliculus,medial geniculate body,and auditory cortex.FSL generally shortens as amplitude increases.However,paradoxical latency shifts involving an increase in FSL with increasing amplitude,and constant latency cells that show a much smaller latency change with signal intensity or an independence of FSL with changing amplitude have also been observed.The SC of a neuron,as a function of acoustic frequency,takes the form of an "upside down V" shape for single peaked tuning curves.The acoustic frequency that evokes the highest response is defined as the best frequency（BF）.Both the response-frequency curve and BF are relatively variable,especially at high amplitude, and BF does not generally correspond with the characteristic frequency（CF,the most sensitive frequency or frequency at minimum threshold）,particularly when the amplitudes of acoustic stimuli are more than 10 dB above the threshold.In contrast to SC,the FSL- frequency function takes the reverse form to the spike count-frequency function resembling a "V shape",in which the shortest FSL is always at,or near the CF,and FSL lengthens with the increasing difference between the frequency of acoustic stimulus and the CF.However,it is unclear which one of SC and FSL is more suitable as a parameter for evaluating the responses of neurons to acoustic amplitudes and frequencies,since systematic comparisons between SC and FSL tuned to different amplitudes and frequencies,are scarce.Besides,both SC and FSL can be affected by many other acoustic parameters.In our study,a 1.5-cm-long nail was glued onto the dorsal surface of each mouse’s skull with dental cement.The animal was held in a polyethylene-foam body-mold with an elastic band hung over a stereotaxic apparatus,which was fixed on an anti-vibration table in a soundproof room.The animal’s head was immobilized by fixing the nail to a small metal rod with screws.A 2x2 mm² area on one side of the IC was exposed by opening the skull and dura above the IC.Pure tone bursts were used as acoustic stimuli and were generated and delivered using a Tucker-Davis Technologies System 3.The tone burst signals lasting 50 ms each with a 5 ms rise-fall time.Pure tone bursts were played back using a computer with BrainWare software which controlled the frequency and amplitude of pure tone bursts either manually or automatically.Extracellular recordings were made with glass micropipettes,and data were acquired and processed online with TDT 3.The CF and minimum threshold（MT, defined as the amplitude of tone bursts at CF required to elicit a spike firing probability of 0.1） were first measured approximately by manually varying the frequency and amplitude of tone bursts after an IC neuron was isolated.A frequency and amplitude（F-A） scan was then performed in which frequencies were varied in the range,CF±5 kHz,in 1 or 0.5 kHz steps.The waveforms,numbers and timings of spikes evoked by acoustic stimuli were collected and stored as data sets.The data were monitored with respect to SC-level,frequency function,post-stimulus time histograms（PSTH）,spike shapes and feature space window using Brain Ware.Based on these data,spike counts or first spike latencies,and their respective standard deviations（SD） and coefficients of variation（CV） were plotted as functions of amplitude and frequency in offline processing.Totally 163 inferior colliculus neurons were recorded and the systematical comparison of the precision and stability（i.e.,the resolution and the coefficient variation,CV） of SC- and FSL-function as frequencies and amplitudes in the inferior colliculus of mice were made.The results showed that:1.the SC-amplitude functions were of diverse shape.Of 163 neurons,21（13%） were monotonic,47（29%） were saturated,and 95（58%） were non-monotonic.Nevertheless,the FSL-amplitude functions were in close registration,in which FSL decreased with the increase of amplitude and no paradoxical（an increase in FSL with increasing amplitude） or constant（an independence of FSL on amplitude） neuron was observed;2.We examined the discriminability（resolution） of differences in amplitude based on FSL and SC,that is,whether the responses to tones of two neighboring amplitude were significantly different,by one-way ANOVA.For SC only 2 or 3 data points were significantly different（Fig.3 A1,B1 and C1,the asterisks for the open circles）.For FSL,however,most of the data points showed differences（Fig.3 A1,B1 and C1,the asterisks for the filled circles）.Same analysis was applied to the discriminability of frequency based on FSL and SC,and similar results were acquired.Based on these analyses,we concluded that the discriminability（resolution） of differences in amplitude and frequency based on FSL are higher than those based on SC;3.To compare CV_SC and CV_FSL,we plotted them against each other for single neurons.The CV index of FSL(CVI_FSL),defined as the ratio of the number of the points at which CV_FSL is smaller than CV_SC to all the observed points on the amplitude or frequency functions within an individual neuron,was measured for all studied neurons.Out of 163 recorded neurons,respectively 151（93%） neurons responding to varying amplitudes at the neurons’ CFs,or responding to varying frequencies at the amplitude of which the CV（SC）- and CV（FSL）-amplitude curves intersected,showed DI_FSL greater than 0.5.In other words,for more than haif of the observed points within one neuron, CV_FSL was lower than CV_SC.The CVs of FSL for low amplitude stimuli were smaller than those of SC;4.the fraction of neurons for which best frequency（BF） = characteristic frequency（CF）（within±500 Hz） obtained from FSL was higher than that from SC at any amplitude of sound.The BF_SC of 85%（138 of 163） of neurons matched their CF at 5 dB above MT,but the fraction decreased sharply to about 50% as the amplitude increased to over 10 dB above MT,and to 12%at 60 dB above MT. The BF_FSL matched well with CF within 20dB above MT（with 99%,89%and 85% neurons at 5 dB,10 dB and 20 dB above MT,respectively） and also gradually decreased as the amplitude increased,reaching a minimum of 21%at 60 dB above MT.Therefore,systematical comparison of the representation of acoustic amplitude and frequency by SC and FSL showed that,SC and FSL may vary,independent from each other and represent different parameters of an acoustic stimulus,but FSL with its precision and stability appears to be a better parameter than SC in evaluation of the response of a neuron to frequency and amplitude in mouse inferior colliculus.The peripheral auditory system（cochlea） behaves as if it contains a bank of overlapping bandpass filters,which is finely arrange on the basilar membrane,with higher frequencies near the base of the cochlea and lower frequencies on the apex. This frequency selectivity distribution is the so called cochleaotopographic （tonotopical） organization of basilar membrane.Paralleled to the tonotopical feature of basilar membrane,neurons at each level of auditory pathway also demonstrate tonotopically-organized arrangement according to their CFs or BFs.Several basic principles in the signal processing of central nervous system is suggested by previous works.One of these principles is the rank-order processing hypothesis,i.e.,lower levels of central nervous system can only perform simple signal processing,while the more complicated works are left to the higher nuclei. Another one is the parallel signal processing hypothesis,suggesting that there are many parallel tonotopic structures on different levels of the signal transferring pathway,which allows the neurons at different levels to abstract and process different characters of the signal.Signal processing in the auditory system is also hypothesized parallel-hierarchical,and the tonotopic structures on different levels of auditory pathway are thought to be the structural basis.But,no substantial evidence of the hypothesis is directly observed.To evidence it,filtering characteristic of a neuron should be found and relayed from basilar membrane,i.e.,the response of a neuron to any frequency acoustic stimulus is due to frequency-amplitude trading relative to CF or the resonant frequency（RF） acoustic stimulus.And also,this neuronal filtering characteristic would be relayed from BM.However,both of them are unexplored.According to the above comparisons of the representation of sound frequency and amplitude by FSL andSC,FSL is better than SC when representing the frequency and amplitude,so it was used as the response parameter to calculate the relationship between the response for neuronal CF and other frequencies.The materials and methods such as the surgical operation,acoustic stimuli,data acquisition are mostly the same as above.After the single peak auditory neuron was isolated and its CF and MT were measured,F-A scan were then performed with the range of CF±5 kHz and step of 1 or 0.5 kHz.Based on these data,MT,CF,spike counts,spike rates,first spike latency and successive spike latency were calculated and plotted as functions of frequency and amplitude,and the furthermore processing such as slope of functions, slopes distribution over CF,MT and population neurons.57 single peak neurons were recorded and the results of analysis showed that,1. FSL can accurately represent the stimulus frequency and amplitude within a neuron with the exponential first decay function when other parameters such as rise time, phase and so on are fixed,and can be fitted well by the function of Pieron’s empirical law（the range of R² was between 0.917 and 0.998）.The FSL-amplitude functions for non-CF frequency are the same as that for CF frequency;2.Single peak inferior colliculus neurons show filter characteristic in that the FSL - amplitude functions at different frequencies are similar to each other.When FSL is plotted by the effective amplitude normalized by CF（i.e.,the effective amplitude at other frequency equals the amplitude at CF subtracted by a fixed value△x）,then all the FSL - amplitude curves at different frequencies will overlap with each other,which is resemble the feature of a single filter.Thus,the filter characteristic of the inferior colliculus neurons is relayed from a single point of basilar membrane;3.The FSL transferring character can be described by the exponential constant（constant "c or b"）,which was bigger for higher threshold neuron regardless of CF（when doing the linear regression against CF,R² = 0.0042）.Our results suggested that,high level neurons in auditory system showed a filter characteristic relayed from BM,and substantially supplied evidences that signal processing in the auditory system is parallel-hierarchical.Furthermore,when other parameters such as rise time and phase are fixed,the latency of neuron firing can represent amplitude and frequency simultaneously.更多还原