High-throughput mouse phenotyping

5 downloads 4143 Views 1MB Size Report
Dec 23, 2010 - on the EMPReSS website (http://empress.har.mrc.ac.uk). ..... The query builder menu on the left-hand side enables the user to select and filter the results and the results ... systems are relatively hands-free and high throughput.
Methods 53 (2011) 394–404

Contents lists available at ScienceDirect

Methods journal homepage: www.elsevier.com/locate/ymeth

High-throughput mouse phenotyping Hilary Gates, Ann-Marie Mallon, Steve D.M. Brown ⇑, EUMODIC Consortium 1 MRC Mammalian Genetics Unit, MRC Harwell, Harwell Science and Innovation Campus, Harwell OX11 0RD, UK

a r t i c l e

i n f o

Article history: Available online 23 December 2010 Keywords: Mouse Phenotyping SOPs Databases

a b s t r a c t Comprehensive phenotyping will be required to reveal the pleiotropic functions of a gene and to uncover the wider role of genetic loci within diverse biological systems. The challenge will be to devise phenotyping approaches to characterise the thousands of mutants that are being generated as part of international efforts to acquire a mutant for every gene in the mouse genome. In order to acquire robust datasets of broad based phenotypes from mouse mutants it is necessary to design and implement pipelines that incorporate standardised phenotyping platforms that are validated across diverse mouse genetics centres or mouse clinics. We describe here the rationale and methodology behind one phenotyping pipeline, EMPReSSslim, that was designed as part of the work of the EUMORPHIA and EUMODIC consortia, and which exemplifies some of the challenges facing large-scale phenotyping. EMPReSSslim captures a broad range of data on diverse biological systems, from biochemical to physiological amongst others. Data capture and dissemination is pivotal to the operation of large-scale phenotyping pipelines, including the definition of parameters integral to each phenotyping test and the associated ontological descriptions. EMPReSSslim data is displayed within the EuroPhenome database, where a variety of tools are available to allow the user to search for interesting biological or clinical phenotypes. Ó 2011 Elsevier Inc. All rights reserved.

1. Introduction The mouse is a pivotal model organism for the study of gene function and the generation and analysis of disease models. It plays a key role in understanding mammalian physiology and disease processes underpinned by the extensive toolkit available for manipulating the genome. Many of the techniques and tools that are available for genome manipulation are described in this volume. The ultimate aim of altering the mouse genome in defined ways is to determine the function of genes and their role in the pathophysiology of disease. Progress in understanding the contribution of genes to disease processes will lead to improvements in diagnosis and the identification of targets for novel therapeutic approaches. A starting point for these investigations is the generation of a mouse knock-out. To facilitate this process an ES cell library comprising a knock-out for each gene in the mouse genome is being produced under the combined efforts of the International Knockout Mouse Consortium (IKMC). To build on this venture an International Mouse Phenotyping Consortium (IMPC) has been established to create mouse lines from each targeted ES cell, to determine the phenotype of the resulting mutant mice, and to archive them for further investigation by the wider scientific community. Comprehensive and systematic phenotyping of mouse knockouts requires centralised facilities with broad ranging expertise ⇑ Corresponding author. 1

E-mail address: [email protected] (S.D.M. Brown). See www.eumodic.org.

1046-2023/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.ymeth.2010.12.017

in physiological, biochemical and developmental systems [1]. Consequently the concept of the ‘mouse clinic’ has emerged encompassing a large, state of the art mouse functional genomics centre with the expertise to generate, phenotype and archive large numbers of mouse lines [2,3]. The phenotyping of a knock-out for every gene in the mouse genome will be a formidable undertaking, and there are considerable logistical, technological and scientific hurdles to be overcome. A pilot project for IMPC has been established under funding from the European Union called the European Mouse Disease Clinic or EUMODIC (www.eumodic.eu) in order to assess large-scale phenotyping and its operational and scientific requirements. EUMODIC brings together 4 mouse clinics: the Medical Research Council (MRC Harwell) and the Wellcome Trust Sanger Institute (WTSI) in the UK, Helmholtz Zentrum München German Mouse Clinic (HMGU) in Germany and the Institute Clinique de la Souris (ICS) in France. Knock-out mice are produced from the EUCOMM and KOMP ES cell resource, cohorts are bred and then passed through a series of primary screens which will provide information on a wide variety of structural, physiological and behavioural systems amongst others. Of course, the mice will require further and more detailed investigation before a full phenotype is known. However, such a broad-based primary screen is designed to reveal sufficient information about potential phenotypic attributes that will serve as the basis for a more focused secondary analysis. These analyses are performed in EUMODIC by so-called secondary phenotyping partners who select and import relevant mutant lines into their institutes for further detailed study. A very similar programme has been underway at

H. Gates et al. / Methods 53 (2011) 394–404

the German Mouse Clinic (Helmholtz Zentrum München) and the Wellcome Trust Sanger Institute (WTSI) in the UK as part of its Mouse Genetics Programme (MGP). The MGP programme employs a combination of tests that is similar to EUMODIC. However, EUMODIC employs two pipelines of tests, while the WTSI programme carries out the tests in a single pipeline. For the EUMODIC programme, the composition of the screen, logistics of breeding and phenotyping controls and data handling are described in detail below. While there are some differences between the composition of tests employed in the WTSI MGP pipeline and the EUMODIC pipelines, by and large the operation and phenotypic outcomes of individual tests that are utilised in common are the same. For this reason, we concentrate on the nature of the individual tests. However, it should be emphasised that test order can have a dramatic effect on test outcome, particularly for behavioural tests, and it should be noted that every phenotyping pipeline should be assessed in the context of test order and inter-test interval. The ultimate aim is to make the data and the mice available to the wider scientific community as a resource for further analysis and investigation. Data capture and analysis are critical for cataloguing and disseminating phenotype information to the wider biomedical sciences community. Accordingly, we place some emphasis on describing the underpinning informatics strategies that are pivotal to the operation of a mouse clinic and the acquisition and annotation of phenotype data. The challenge for mouse geneticists for the future is to populate a matrix comprising gene, environment and phenotype test, with an array of data points that will serve as a vehicle for the study of biological mechanisms in diverse systems.

395

9. Associated documents 10. Supporting information 11. History review The EMPReSS database [7] contains all of the data described above for the EMPReSSslim SOPs. This includes detailed information about the parameters to be measured, including their name, the expected units, the datatype and if the parameter is required to be measured or not (Table 1). Importantly, EMPReSS also includes metadata parameters, particularly those that might affect the outcome of the test, for example the type of anaesthesia, the equipment or settings used and sample handling. It is recognised that complete uniformity even within centres will be hard to achieve so recording factors that might influence the results will allow the data to be displayed by metadata and only crosscompared where appropriate. In addition to the parameters and metadata, EMPReSS stores the phenotypic ontology annotations for the majority of parameters. These annotations describe the expected phenotype that would be identified by a specific parameter if the mutant data was statistically different from the control. Currently these annotations are terms from the Mammalian Phenotype Ontology (MP) [8], but in the future this will be extended to include the ontology of phenotypic qualities (PATO) (http://obofoundry.org/wiki/index.php/ PATO:Main_Page) annotations enabling potential cross species integration. This detailed information stored in EMPReSS is crucial to the capture of data within EUMODIC and its subsequent uploading to the EuroPhenome database (http://www.europhenome.org) [9] (see Section 2.4).

2. Methods 2.2. Pipelines and order of tests 2.1. Use of Standard Operating Procedures (SOPs) Given the scale of future ambitions to phenotype a mouse mutant for every gene in the mouse genome, a number of mouse clinics will have to be engaged in the programme. As each test is performed in more than one clinic, it is vital to have Standard Operating Procedures (SOPs) which cover the equipment, consumables, quality control and the protocol and deliver robust, validated phenotyping outcomes. The format for EUMODIC SOPs was developed under a previous project, also funded by the European Union, called Eumorphia. This brought together 18 research institutes from eight European Countries in order to develop and standardise a set of SOPs covering diverse systems – the European Mouse Phenotyping Resource of Standardised Screens (EMPReSS, [4]). All of the SOPs are available on the EMPReSS website (http://empress.har.mrc.ac.uk). Each of the SOPs established under Eumorphia have subsequently developed and evolved under EUMODIC. The SOPs conform to the MIMPP (Minimum Information about a Mouse Phenotyping Procedure) [5] standard (http://mibbi.org/index.php/Projects/MIMPP) which outlines the minimum information required to describe a phenotyping procedure in much the same way as MIAME describes minimal information required to describe a microarray experiment [6]. The current format for the EMPReSSlim SOP is divided into the following sections: 1. 2. 3. 4. 5. 6. 7. 8.

Purpose Procedure Parameters recorded Metadata recorded Equipment Supplies Quality control Notes

EUMODIC has employed a subset of the EMPReSS tests. The phenotyping tests used in EUMODIC operate in a set sequence called a pipeline. In order to obtain a broad-based assessment of phenotype, EUMODIC utilises a wide variety of phenotyping platforms. Given the density of the testing regime, it was considered impractical to operate a single pipeline. Rather, EUMODIC operates two pipelines each comprising different tests with a separate cohort of mice passing through each pipeline (Fig. 1). These two pipelines are collectively known as EMPReSSslim. It is however likely that future high-throughput phenotyping under IMPC will operate a single pipeline, as this will reduce the numbers of animals and the mouse breeding required (see Section 2.3 below). However, a single pipeline will need to be carefully designed. It will be important to retain the extent of phenotyping platforms and the richness of phenotypes detected, while at the same time carefully considering test order. A critical area of pipeline operation is to ensure that each test will not unduly influence the results of the next. In the Eumorphia project, considerable emphasis was placed on the appropriate order of tests performed. This is particularly true for behavioural testing where test history and handling is likely to have unpredictable effects on performance [10–12]. 2.3. Breeding cohorts Production of mice from ES cells and subsequent breeding of cohorts for phenotyping has proved to be a major rate-limiting factor for high-throughput phenotyping. Viability and fertility are often compromised in knock-out mutants comfounding the production of age matched cohorts. Where homozygotes are embryonic lethal, or homozygotes are sub-fertile, it is often advantageous to analyse heterozygotes. EUMODIC enters cohorts of seven age-matched males and seven females into each pipeline. However, the breeding required to provide complete full-sized cohorts is formidable and mice are often passed through each pipeline in mini-cohorts of less

396

H. Gates et al. / Methods 53 (2011) 394–404

Table 1 Example of information captured for a SOP. Parameter_name

Unit

Datatype

Required

MP ontology annotations

Body weight Systolic arterial pressure

g mmHg

FLOAT INTEGER

1 1

Pulse rate Description

bpm

INTEGER TEXT

1 0

Increased body weight (MP:0001260) or decreased body weight (MP: 0001262) Increased systolic blood pressure (MP:0006144) or decreased systolic blood pressure (MP:0006264) Increased heart rate (MP:0002626) or decreased heart rate (MP:0005333)

TEXT TEXT TEXT DATE/ TIME TEXT DATE/ TIME

1 1 1 0

Metadata_name Equipment name Equipment manufacturer Equipment model Time training period performed Comments Date/time when moved to testing room

0 0

In EMPReSS detailed information is captured about each SOP. This table shows an example of the information stored for the Non-Invasive Blood Pressure SOP. The first column (Parameter_name) lists all the parameters and metadata parameters that are measured, the second column (Unit) holds the unit that a parameter must be measured in and the third column (Datatype) controls the datatype of the data for the upload into EuroPhenome from a local LIMS system. The fourth column (Required) defines if a parameter is required (1) or optional (0) allowing different levels of data to be collected. Finally the last column (MP ontology annotation) describes the phenotypes as MP terms that the parameter is measuring.

Fig. 1. EMPReSSslim Pipelines. The pipeline diagrams depict the order and timings that the SOPs are performed in and the number of males and females that are tested. The SOPs are grouped into the six major disease systems: morphology and metabolism, cardiovascular, bone, neuro-behavioural and sensory, haematology and clinical chemistry and allergy and immune.

than seven. Statistical analyses have been performed on control data and it has been determined that, for most tests, seven animals provide the optimum number of detected mutations per mouse. The minimum effect sizes detected are considerably decreased if cohort size is reduced to four animals and an increase to 10 only gives on average 20% more mutations detected at a cost of 43% more mice. 2.4. Data Data generated from the EMPReSSlim pipelines is captured in the four EUMODIC mouse clinics Laboratory Information Manage-

ment System (LIMS) in accordance with the SOPs stored in EMPReSS. The database architectures and software implementations of the various LIMS differ markedly and each captures a large amount of additional data other than the phenotyping data. To capture the phenotype data it was therefore essential to develop a common data format that would exchange the data from the LIMS to EuroPhenome. An XML (Extensible Markup Language) data format is used for this purpose, whereby the phenotype data for each individual mouse, making up a mutant strain cohort, can be submitted for each procedure carried out. The phenotype data and cohort XML files are uploaded to the clinics FTP site which EuroPhenome

H. Gates et al. / Methods 53 (2011) 394–404

scans daily for new data. These data files are validated, on export from the LIMS and on upload to EuroPhenome, against the EMPReSS database. The status of all uploads is monitored in the EuroPhenome tracking system which allows clinics to view the status and issues as soon as the upload has been processed. EuroPhenome is designed to compare each knock-out line to its background strain and so after data upload, the EuroPhenome annotation pipeline is run which compares mutant phenotype data with the appropriate controls data using the relevant statistical test and automatically assigns phenotype ontology terms to the data based on a predefined cut-off [13]. Currently the statistical comparison takes place between related datasets only when critical metadata parameters are identical. Pearson’s Chi-square test of independence is applied to categorical data and the Student’s t-test and the Mann–Whitney U test are applied to quantitative data. These tests were chosen because they are familiar to the community and computationally tractable for the size of data set. The use of these statistical tests is currently being reviewed within EUMODIC and outcomes from this will be implemented in the annotation pipeline. The EuroPhenome website (http://www.europhenome.org/) allows users access to both the annotated data and the raw data through a number of tools. The three major tools are the Phenome data viewer, the Phenomap tool and the Ontology tree. The Phenome data viewer allows users to build a specific query about a line of interest and results in a graphical view of the data chosen (Fig. 2). The Phenomap tool allows users to view a heatmap representation of all lines (or an user selected list) and the procedures for which there is data present. Red coloured boxes indicate a significant annotation is present at the chosen P-value and effect size, green indicates no significant annotation has been assigned and grey indicates the raw data is not present (Fig. 3). Additional functionality is available for users to change the P-value and effect size, which results in the Phenomap being automatically modified.

3. Tests performed The two pipelines of EMPReSSslim (Fig. 1) assess a number of disease systems and each comprises a series of tests carried out from 9 to 16 weeks. Cohorts of male and female mice are tested through each pipeline, data is captured by a local LIMS system and then exported to the EuroPhenome database (see above). EMPReSSslim consists of 20 phenotyping platforms capturing a large number of both phenotype data and metadata parameters. In each case, we describe the basis of the test and provide some detailed notes that elaborate on issues of test operation (including hints for troubleshooting) and, where appropriate, data capture. 3.1. Pipeline 1 3.1.1. Dysmorphology (9 weeks) The dysmorphology screen is a simple observational assessment to identify any morphological abnormalities in an adult mouse. A dysmorphology screen can provide important information on a variety of craniofacial, limb, and other visible malformations as well as visible phenotypes such as coat colour/texture, genital irregularities and dentition [14]. Observations in the EMPReSSslim dysmorphology screen include:  Whole body: length of body, thickness of skin and fat distribution.  Coat: dorsal and ventral variation in colour, presence, patterning and irregularities in the placement, texture and distribution of hair on the head, back and stomach.  Skin: texture, moisture, colour and patterning on the body, nose, ear, hand, foot and tail.

397

 Head: anomalies in the length, width, symmetry and curvature of the head; abnormalities in vibrissae; number colour and shape of teeth; shape, position and size of ears; opacity, size, positioning, colour and closure of eyes.  Genitalia: size, presence and structure. Colour may vary for female mice according to the stage of the oestrous cycle.  Forelimb: size, flexibility, shape, structure and number of digits present on the forepaws.  Hindlimb: size, flexibility, shape, structure and number of digits present on the hindlimbs.  Tail: presence, length, width, curvature and any subtle kinks. Operational Notes The dysmorphology test has been refined in EUMODIC and the SOP altered so that it does not overlap with behavioural and movement observations that are assessed in other tests such as modified SHIRPA (see below). Dysmorphology in EMPReSSslim is purely focussed on observations of the morphology of the mouse. Currently there are 181 parameters that can be recorded in the EMPReSSslim dysmorphology screen. To simplify this, an on-line tool has been developed to record the parameters in a hierarchical way, so that a detailed breakdown of the abnormality is only required if a high-level abnormality is recorded. Ontologies from the Mouse Anatomy (MA) (e.g. coat hair) and PATO ontologies that reflect qualities (e.g. sparse) are utilised so that the parameters measured are recorded in a logical way that reflects both the anatomy that is altered and the manner in which it is changed. 3.1.2. Non-invasive blood pressure (11 weeks) Measurement of the systolic blood pressure (mmHg) and heart rate (beats per minute) of conscious mice using a non-invasive tail cuff system. The test is performed on five consecutive days at the same time of day. The first day is a training day. Data for systolic pressure and pulse rate are averaged over the remaining 4 days. Each day a run is performed which consists of a series of 20 measurements called cycles. The first 5 or 10 cycles are employed to habituate the mouse to the equipment and test. The second 15 or 10 cycles are the measurement cycles that are recorded and averaged. Operational Notes Data are only recorded and included in the average when the following criteria are met: (1) A value has to be obtained for both systolic pressure and pulse rate. If data is recorded for only one parameter (or none) it is rejected. (2) Successful measurements are obtained for 7 cycles out of 10. (3) A minimum of 20 successful measurements are obtained over the 4 measurement days. Data can also be rejected following visual examination of the waveforms. This is however a skilled and time consuming method and not well suited to a high-throughput screen. A cross comparison amongst EUMODIC partners found that average readings were not significantly different if the few data points that do not have a good waveform are excluded. Thus, as a method for quality control, this protocol has not been adopted in EMPReSSslim. 3.1.3. Indirect calorimetry (12 weeks) The energy expenditure is evaluated through indirect calorimetry by measuring oxygen consumption with an open flow respirometric system. Indirect calorimetry provides detailed information on the energy metabolism of mutant mice and therefore specific supportive data that is required to describe a metabolic phenotype. There is

398

H. Gates et al. / Methods 53 (2011) 394–404

Fig. 2. EuroPhenome Phenome Data Viewer. This interface shows the results from the EuroPhenome data viewer when an user selects a gene (e.g. Akt2), a SOP (e.g. IPGTT) and one or more parameters. The query builder menu on the left-hand side enables the user to select and filter the results and the results pane on the right displays a graphical image for the specific parameters selected. The results pane also enables the user to drill down into the statistical results relevant to the particular graph.

some overlap with phenotyping results obtained by DEXA, body weight, intra-peritoneal glucose tolerance test (IPGTT), or clinical chemistry data (see below). However, indirect calorimetry data adds important information and plays an useful role in the selection of mutant lines as candidates for more detailed analysis of the metabolic phenotype.

Precise CO2 and O2 sensors measure the difference in CO2 and O2 concentrations in air volumes flowing through control or animal cages. The amount of oxygen consumed over a given period of time can thus be calculated, as the air flow through the cage is known. Data are expressed as ml O2 h 1 animal 1. The system also monitors CO2 production and consequently the respiratory exchange

H. Gates et al. / Methods 53 (2011) 394–404

399

Fig. 3. EuroPhenome phenomap. This interface shows the EuroPhenome Phenomap tool, which is a heatmap representation of statistically significant phenovariants produced from the annotation pipeline. The control panel on the top allows the user to filter the results such as changing the P-value of the annotations displayed. The colours on the phenomap are: grey boxes which depict a procedure for which we currently do not hold data; green boxes which depict when data is available but no significant annotations were found at the current P-value and effect size and red boxes which show a significant annotation at the current P-value and effect size. Further details on red boxes can be obtained by hovering over the box.

ratio (RER) and heat production can be calculated. An activity and food intake monitoring system can also be integrated to the apparatus for the measurement of activity. Indirect calorimetry gives the most accurate measurement of energy generated and calories burned by the body and is essential in characterising energy homeostasis. Importantly, the current fully automated commercial systems are relatively hands-free and high throughput. Operational Notes Calorimetry requires robust quality control of phenotyping data at each centre, as well as refined approaches for statistical analysis (e.g. linear models including body mass as covariate or normalisation based on lean body mass). Improved approaches to the generation of derived parameters and statistical analysis will be important in revealing new metabolic phenotypes. Importantly, there is evidence based on current practice that it is preferable for control data to only be taken from the same clinic and same time period as mutant data but not from general baseline.

3.1.4. Simplified Intra-Peritoneal Glucose Tolerance Test, IPGTT (13 weeks) The Intraperitoneal Glucose Tolerance Test (IPGTT) measures the clearance of an intraperitoneally injected glucose load from

the body. It is used to detect disturbances in glucose metabolism that can be linked to human conditions such as diabetes or metabolic syndrome. A simplified, high throughput variant of the IPGTT is used. Animals are fasted for approximately 16 h, a solution of glucose (2 g of glucose/kg) is administered by intra-peritoneal (IP) injection and blood glucose is measured at 15, 30, 60 and 120 min postinjection. Glucose concentration in whole blood is measured using a hand-held glucose monitor called a Glucotrend that is used to measure glucose concentration in human diabetics. Operational Notes Blood is collected from the tail tip by scoring the tip of the tail with a scalpel after applying a topical anesthetic cream. A drop of blood is placed on a test strip that is fed into the monitor to give a reading. Measurements obtained by this method (for whole blood) were consistently lower than plasma measurements using a clinical chemistry analyzer. However the handheld glucose monitor was chosen as it clearly provides a robust assessment of impaired glucose tolerance and is much quicker to use in a high-throughput screen. More recently, the manufacturers have introduced a modification whereby a conversion factor is automatically applied to the readout providing a glucose concentration that more closely corresponds to the

400

H. Gates et al. / Methods 53 (2011) 394–404

plasma concentration in accordance with the International Federation of Clinical Chemistry (IFCC). 3.1.5. Dual-energy X-ray absorptiometry (DEXA or DXA) (14 weeks) A method of quantifying bone mineral content and density and fat mass. DEXA uses an X-ray generator of high stability to produce photons over a broad spectrum of energy levels [15]. The photon output is filtered to produce the two distinct peaks necessary to distinguish bone from soft tissue, which further enables body composition (fat mass and lean mass) to be measured. The following are recorded for an anaesthetized mouse: body weight, body length, fat and bone mass, bone mass density and lean mass. This test therefore gives information on the metabolic status of the mouse (fat content) as well as recording bone density. Operational Notes To determine the machine variation between clinics, a pilot experiment was set up using two ‘‘phantom’’ mice of differing simulated body composition. These phantom mice were measured at each of the four primary clinics. Comparison of the data indicates that there are significant differences in results depending on the type of DEXA machine used. This indicates the need for caution in direct comparison of DEXA data between clinics. In addition, because of the degree of intra-clinic variability, the SOP now determines that calibration of the DEXA apparatus should be carried out daily using the QC and QA phantoms provided by the manufacturer [16]. 3.1.6. X-ray (14 weeks) Following DEXA scanning digital X-ray images are taken of the anaesthetised mouse to allow a detailed analysis of the mouse skeletal system and teeth. Digital X-ray Images are examined at the clinics and any skeletal or dental abnormalities noted along with a count of the number of ribs, vertebrae and digits. This test gives information about skeletal malformations [17]. Operational Notes Advantage can also be taken of the pipeline approach to screening, since results from X-ray can be combined with observations in the dysmorphology screen and movement in the SHIRPA screen (see 3.2.2), such as gait and struggling when held by the tail, to discover more subtle phenotypes. Information can also be compared with measurements of blood calcium in the clinical chemistry screen (see below).

3.1.8. Fasted clinical chemistry (15/16 weeks) A clinical chemistry analysis is carried out on blood after an overnight fast of up to 16 h. The parameters measured are: total cholesterol, HDL cholesterol, non-HDL cholesterol, triglycerides, glucose, free fatty acids and glycerol. This complements the clinical chemical analysis of a non-fasted bleed undertaken on the other cohort of mice phenotyped in pipeline 2 (see below). Operational Notes Research into standardization using C57BL/6J at ICS, has shown that metabolic parameters were significantly affected by fasting [17]. The fasted bleed incorporated in pipeline 1 to measure biochemical parameters in plasma may be expected to show lower variance compared to data from free-fed animals.

3.2. Pipeline 2 3.2.1. Neurological, sensory and behavioural function From 9 to 12 weeks an integrated series of tests exploring neurological, sensory and behavioural function are performed. Members of the Eumorphia consortium developed an integrated battery of tests exploring these systems that has been designed to limit influences from preceding tests in the battery with the outcome of subsequent tests [18]. In EMPReSSslim tests are performed in the order: open field, modified SHIRPA, grip strength, rotarod, acoustic startle and PPI, followed by hot plate. 3.2.2. Open field (9 weeks) Mice are placed in an open field arena (a square or rectangular box with perspex sides but no top) and their activity is monitored using a video camera or infrared sensors. The open field test is used to assess anxiety and exploratory drive by assessing conflicting behaviour in response to a brightly lit aversive arena and the natural tendency to explore a novel environment. Mice are placed in the periphery and left to explore the arena for 20 min. Activity is measured in terms of: total distance travelled and time spent in the periphery and a defined central zone (40% of the total area). Exploratory drive in the open field is largely dependent on levels of anxiety with an increased tendency to activity in the periphery in highly anxious mice accompanied by reduced activity within the central zone, which is considered to be anxiogenic [19]. Light intensity is a key anxiogenic component that will influence the ambulation in an U-shaped way [20] and factor-analysis has shown that varying illumination intensity affects indices of activity [21] with dim light being predominantly a measure of locomotor activity as opposed to anxiety-like behaviours. Operational Notes

3.1.7. Heart weight and histology (15/16 weeks) This protocol describes the dissection, observation, weighing and fixation of heart for sectioning and potential histological analysis. Changes in heart weight and wall thickness are used to indicate a cardiovascular phenotype. As the heart is removed by dissection the area around the heart is examined to see if there is any excessive fat and the heart is examined for any gross abnormalities. Photographs are taken to illustrate details of any abnormalities. Operational Notes Heart weight is normalised to tibia length to eliminate changes due to differences in mouse size. In EMPReSSslim, because of the pipeline approach, tibia length can be measured from the X-ray images (above) which eliminates the need for time consuming dissection of the tibia at necropsy.

The open-field test has been cross-validated between the EMPReSS clinics [18]. Four inbred strains were tested (C57BL/6J, C3HeB/FeJ, BALB/cByJ, 129S2/SvPas) and shown to give a similar order of performance for the open field parameters (C57 > C3H > BABL > 129); although absolute activity was related to the size of the arena. As discussed above, levels of light are a key component so the test has been altered in some clinics to standardise better the open field by altering aspects of the apparatus (changing the colour of the sides) and the conditions (e.g. light levels) in order to make the apparatus more comparable between clinics.

3.2.3. Modified SHIRPA (9 weeks) This test is based on the SHIRPA (Smithkline Beecham, MRC Harwell, Imperial College, the Royal London Hospital phenotype

H. Gates et al. / Methods 53 (2011) 394–404

assessment) test [22]. The test can be used to assess a variety of skeletal, motor and behavioural functions, such as: muscles, skeleton and joints, lower motorneurones and spinocerebellar, sensory, neuropsychiatric and autonomic function. The modified SHIRPA test [23,24] consists of a number of observations of behaviour in a cylindrical glass viewing jar and in an open Perspex arena, above which a series of reflex responses are scored. Each observation is given a standard score of 0, 1 or 2 depending on the observation; the scoring sheet is available on EMPReSS. In brief, the observations made are:  In the viewing jar: activity, tremor, defecation, urination, palpebral closure, lacrimation.  In the arena: transfer arousal (freezing or immediate exploration when moved to the arena), locomotor activity (number of squares crossed), gait, pelvic elevation, tail elevation, startle response (to a 90 dB click), touch escape (when stroked on back of neck).  When held by tail: positional passivity (degree of struggling), trunk curl, limb grasping.  Reflexes: pinna reflex and corneal reflex (when inner canthus or cornea touched with cotton probe), contact righting response when placed in a tube and rolled until upside down.  General: any evidence of biting or excessive vocalisation. Operational Notes For the high-throughput screening in EMPReSSslim, the modified SHIRPA was further adapted to remove observations that would be performed in other tests in the pipeline, e.g. coat appearance, which are also observed in the dysmorphology test at the beginning of pipeline 1. A set of videos have also been made to demonstrate the various behaviours and scores in order to help with training the operators in the clinics and standardisation of observations. 3.2.4. Grip strength (9 weeks) The grip strength test is used to measure the neuromuscular function and muscle strength of either the forelimbs or both forelimbs and hind limbs. Grip strength is assessed by the grasping force applied by the mouse on a grid that is connected to a sensor [25]. Three trials are carried out in succession measuring forelimb-strength only, followed by three successive trials measuring the combined forelimb/hindlimb grip strength. All grip strength values obtained are normalised against body weight. Operational Notes Comparison of the results obtained by different phenotypers at one clinic (MRC Harwell), have shown that results obtained can vary between individual experimenters. This is a product of the direction of pull and strength/speed of pull applied by different phenotypers. This phenotyper specific effect has been noticed by other centres. It was concluded that mice should be allowed to attach to the grid properly before being pulled away and the mice should be pulled horizontally from the device to obtain correct measurements. It is also important to note the ID of the technician performing the test and incorporate this into metadata. 3.2.5. Rotarod (10 weeks) The rotarod test is used to assess motor coordination, balance and neuromuscular function in rodents. The rotarod is a widely used test in neurological studies [26– 28,18]. A rod is rotated at a constant or accelerating speed (e.g. from 4 to 40 rpm). The mouse is placed on the rod (against the

401

rotation) and the latency for it to fall from the rod is used to indicate deficits in motor coordination, balance and neuromuscular function. An average latency is calculated for each of the three trials that are performed. There is no training period prior to the test [18]. The equipment has five compartments on the rod so that five mice can undertake the test at the same time so that adjacent mice do not influence each other. In EMPReSSslim the test is run with 3 or 5 mice on the rod at one time but the number of mice on the rod is recorded as metadata. Operational Notes The surface and diameter of the rod are key components that may influence performance in this test. Validation and standardisation studies undertaken in Eumorphia indicated that the use of a soft foam cover around the rod ensured more consistent and reliable results between clinics [18]. The foam should also be replaced prior to the testing of each new cohort. Latency to fall is measured automatically. The test is also stopped if the mouse holds onto the rod and completes a circle without falling (called passive rotation). A single passive rotation counts as a failure in performance equivalent to falling off the rod. 3.2.6. Acoustic startle and pre-pulse inhibition (11 weeks) This test is used to detect abnormalities in neural processing of auditory stimuli. An acoustic startle response (or exaggerated flinch) is produced in response to an unexpected auditory stimulus (or pulse). This response can be attenuated if the stimulus is preceded by a weaker stimulus. This is known as pre-pulse inhibition or PPI. PPI provides an operational measure of sensorimotor gating which reflects the ability of an animal to inhibit sensory information properly [24]. Clinical studies have shown that schizophrenic patients have a reduced PPI. The lack of sufficient sensory gating mechanism is thought to lead to an overflow of the sensory stimulation and disintegration of the cognitive functions. The startle reflex paradigm is often used to assess the effects of putative anti-psychotics and to explore genetic and neurobiological mechanisms underlying behaviours of relevance to psychosis [29,30]. Operational Notes Taken together with the preyer response in the SHIRPA, (a backwards flick of the pinna when presented with a sharp sound of 90 dB), the acoustic startle response can also give an indication of hearing impairments. 3.2.7. Hot plate (12 weeks) The hot plate test is used to assess acute pain sensitivity to a thermal stimulus. A mouse is placed on the hot plate with a surface temperature of 52 °C and the time to an initial reaction is recorded along with the type of reaction [31,32]. The mouse is removed from the hot plate immediately after showing the pain response or after 30 s, if it has shown no response, in order to avoid potential tissue damage. A shorter latency to response would indicate increased pain sensitivity (hyperalgesia) while a longer latency indicates a reduced pain threshold (hypoalgesia). Operational Notes Within EMPReSSslim, the number of parameters recorded has been standardised so that the following types of reaction are noted: paw shake, paw lick, jump or locomotion (which includes backing away), no reaction and ‘other’. Operators need to be trained carefully in the observation and interpretation of the response. Careful calibration of hot plate temperature is required. A step is incorporated within the SOP requiring the monitoring of the surface temperature of the plate before mice are tested.

402

H. Gates et al. / Methods 53 (2011) 394–404

3.2.8. Ophthalmoscope (13 weeks) The fundus of both eyes is observed using an indirect ophthalmoscope following application of a topical dilator. The test is used to determine whether a mouse presents with abnormalities of the fundus, such as retinal degeneration or abnormalities of the optic disc or retinal vasculature.

Table 2 Clinical chemistry parameters recorded in EMPReSSslim.

Operational Notes To allow statistical comparison of the structure of the fundus between mutants and controls, a number of parameters have been defined in EMPReSSslim, which are scored as normal or abnormal for each eye separately: retina, retinal pigmentation, retinal structure, optic disc, blood vessels, blood vessel structure, blood vessel pattern. A free text comment can also be added and photographs uploaded. 3.2.9. Slit lamp (13 weeks) Both eyes are inspected using a slit lamp to determine whether there are eye abnormalities that affect the iris, cornea and lens. As well as recording anterior segment abnormalities of the eye, the response to light is also recorded at this point. Operational Notes A number of parameters are recorded for the left and right eye: eye size, eye present or absent, cornea (normal/abnormal), corneal opacity, corneal vascularisation, lens (normal/ abnormal), lens opacity, iris/pupil (normal/abnormal), iris/ pupil position, iris/pupil shape, iris/pupil light response, iris pigment. A free text comment can also be added and photographs uploaded. 3.2.10. Terminal blood collection (13 weeks) Blood is collected by retro-orbital puncture on free-fed mice to measure a variety of clinical chemistry parameters as well as carry out haematology, FACS analysis and immunoglobulin concentration measurements. SOPs for blood collection and sample handling are detailed on EMPReSS. 3.2.10.1. Clinical chemistry. A routine clinical chemistry screen is performed on plasma using a Beckman Coulter AU400 autoanalyser. SOPs on EMPReSSslim describe the collection of blood in lithium–heparin-coated tubes; sample handling to centrifuge the blood and collect the plasma; reagents and specific calibration and quality control materials; and measurement in the analyser. The screen provides substantive information on a variety of clinical chemical blood parameters, including specific substrates and electrolytes (see Table 2). It can be used to detect defects in a variety of organ systems and metabolic pathways that can be linked to human diseases [33]. Operational Notes Haemolysis of blood samples can differentially affect the values of certain parameters. In EMPReSSslim lipaemia, haemolysis and icterus test severity scores (LiH) are used to automatically score the severity of haemolysis, lipaemia and icterus of the plasma samples. Results are not quantitative, but a severity score is given (see Table 3) and parameters are rejected according to this score. Research by members of the Eumorphia consortium showed that several biochemical parameters are influenced by anaesthesia and the method of blood collection [17]. The clinics within EUMODIC therefore all moved to blood collection by retro-orbital puncture and MRC Harwell developed equipment to enable the gaseous anaesthesia of the mice under UK animal experimentation legislation.

Measured at all centres

Optional measurements

Glucose Urea Creatinine Sodium Potassium Chloride Total protein Albumin Calcium Phosphorus Iron Lactate dehydrogenase Aspartate aminotransferase Alanine aminotransferase Alkaline phosphatase Alpha-amylase Total cholesterol Triglyceride

Free fatty acid Creatine kinase Uric acid Total bilirubin HDL-cholesterol LDL-cholesterol Ferretin Transferrin C-reactive protein

Table 2 shows the clinical chemistry parameters that are required and therefore measured at all centres and those that are optional.

3.2.10.2. Haematology. Haematology parameters (cell counts and haemoglobin concentration) are recorded in EDTA-treated blood. This test will give an indication of abnormalities in the production of blood and its components (blood cells and haemoglobin) and associated blood-forming organs. Parameters recorded are:    

Cell counts: white blood cells, red blood cells, platelets Haemoglobin concentration Haematocrit Calculated haematological indexes: mean cell volume, mean corpuscular haemoglobin and mean cell haemoglobin concentration

3.2.10.3. FACS analysis of peripheral blood cells. FACS (Flow Cytometry Activated Cell Sorting) of peripheral blood cells is used to detect and analyse the main lymphocyte subpopulations of B, T or NK cells. This test, along with the measurement of immunoglobulin concentration, provides a preliminary exploration of immunological function and allergy. Monoclonal antibodies labelled with fluorescent dyes are bound to distinct cell surface antigens. The cell populations are identified using FACS that detects differential expression of cell surface molecules. Reliable results are very much dependent on the appropriate preparation, acquisition and gating of leukocytes. FACS parameters recorded include:         

T cell CD4 + percentage T cell CD8 + percentage T cell CD3 + percentage CD25 + percentage of CD4-T cells Mature B cell CD19 + percentage Granulocyte Gr1 + percentage NK cell percentage Monocyte gate Number of leuocytes Operational Notes The EMPReSSslim method includes ammonium chloride erythrocyte lysis, which prevents interference by the large amounts of erythrocytes. A monoclonal antibody to the cell surface gly-

H. Gates et al. / Methods 53 (2011) 394–404 Table 3 Haemolysis measured by the LiH test. Severity Score

Parameters not reported

+ ++ +++

Potassium, AST, iron & LDH Potassium, AST, iron, LDH, ALP & CK Potassium, AST, iron, LDH, ALP, CK, alpha amylase, sodium, chloride & total protein Potassium, AST, iron, LDH, ALP, CK, alpha amylase, sodium, Chloride, total protein, cholesterol & inorganic phosphate No results to be reported Sample severely haemolysed – no results to be reported

++++ +++++ ABN

Table 3 shows the severity score measured by the LiH test and parameters that are not reported according to the severity of the haemolysis measured.

coprotein CD45 is used to enable the creation of a CD45 + gate. This allows discrimination of leukocytes from debris, erythrocytes and thrombocytes. In addition, staining with propidium iodide (PI) gates out dead cells. Samples are acquired until a count of 30,000 living CD45 + cells is reached for each sample. Data are analysed using FlowJo software. 3.2.10.4. Immunoglobulin concentrations. Immunoglobulin concentration is measured in the serum of mice using a microsphere based multiplex assay (Luminex xMAP Technology). This test along with the FACS analysis provides a preliminary exploration of immunological function and allergy. Luminex beads coupled with antibodies specific for immunoglobulins are used. Immunoglobulin concentrations are recorded for:      

IgM IgG3 IgG1 IgG2b IgA IgE Operational Notes A commercial immunoglobulin bead assay was assessed but found to be unsuitable due to lack of reproducibility and limited ranges of detection. Coupled beads are now prepared by the GMC (HMGU) and distributed to the other mouse clinics.

4. Conclusion EMPReSSslim provides a comprehensive screen for determining abnormalities in a wide range of physiological, anatomical, behavioural and biochemical areas. It has been designed to allow screening of small cohorts of mice in a high-throughput manner. It is being tested by four mouse clinics in the EUMODIC project, under funding from the EU, in order to standardise the tests and produce meaningful results. The ultimate aim will be to develop and adapt the EMPReSSslim screen and use it to determine the phenotype of a KO for each coding gene in the mouse genome. Acknowledgments EMPReSS, EMPReSSslim and the EuroPhenome Database were supported by the European Commission under the Eumorphia project (Framework Programme 5: QLG2-CT-2002-00930) and the EUMODIC project (Framework Programme 6: LSHG-CT-2006037188). A full list of contributors to the EUMODIC project can be found at http://www.eumodic.org/contributors.html.

403

References [1] S.D.M. Brown, J.M. Hancock, H. Gates, PLoS Genetics 2 (8) (2006) 1131–1137. [2] V. Gailus-Durner, H. Fuchs, L. Becker, I. Bolle, M. Brielmeier, J. Calzada-Wack, R. Elvert, N. Ehrhardt, C. Dalke, T.J. Franz, E. Grundner-Culemann, S. Hammelbacher, S.M. Hölter, G. Hölzlwimmer, M. Horsch, A. Javaheri, S.V. Kalaydjiev, M. Klempt, E. Kling, S. Kunder, C. Lengger, T. Lisse, T. Mijalski, B. Naton, V. Pedersen, C. Prehn, G. Przemeck, I. Racz, C. Reinhard, P. Reitmeir, I. Schneider, A. Schrewe, R. Steinkamp, C. Zybill, J. Adamski, J. Beckers, H. Behrendt, J. Favor, J. Graw, G. Heldmaier, H. Höfler, B. Ivandic, H. Katus, P. Kirchhof, M. Klingenspor, T. Klopstock, A. Lengeling, W. Müller, F. Ohl, M. Ollert, L. Quintanilla-Martinez, J. Schmidt, H. Schulz, E. Wolf, W. Wurst, A. Zimmer, D.H. Busch, M. Hrabé de Angelis, Nature Methods 2 (2005) 403– 404. [3] H. Fuchs, V. Gailus-Durner, T. Adler, J.A. Aguilar-Pimentel, L. Becker, J. CalzadaWack, P. Da Silva-Buttkus, F. Neff, A. Götz, W. Hans, S.M. Hölter, M. Horsch, G. Kastenmüller, E. Kemter, C. Lengger, H. Maier, M. Matloka, G. Möller, B. Naton, C. Prehn, O. Puk, I. Rácz, B. Rathkolb, W. Römisch-Marg, J. Rozman, R. WangSattler, A. Schrewe, C. Stöger, M. Tost, J. Adamski, B. Aigner, J. Beckers, H. Behrendt, D.H. Busch, I. Esposito, J. Graw, T. Illig, B. Ivandic, M. Klingenspor, T. Klopstock, E. Kremmer, M. Mempel, S. Neschen, M. Ollert, H. Schulz, K. Suhre, E. Wolf, W. Wurst, A. Zimmer, M. Hrabé de Angelis, Mouse Phenotyping, Methods 53 (2011) 120–135. [4] S.D.M. Brown, P. Chambon, M. Hrabé de Angelis, et al., Nature Genetics 37 (2005) 1155. [5] J.M. Hancock, N.C. Adams, V. Aidinis, A. Blake, J.A. Blake, M. Bogue, S.D.M. Brown, E.J. Chesler, D. Davidson, C. Duran, J.T. Eppig, V. Gailus-Durner, H. Gates, G.V. Gkoutos, S. Greenaway, M. Hrabé de Angelis, G. Kollias, S. Leblanc, K. Lee, C. Lengger, H. Maier, A.-M. Mallon, H. Masuya, D.G. Melvin, W. Müller, H. Parkinson, G. Proctor, E. Reuveni, P. Schofield, A. Shukla, C. Smith, T. Toyoda, L. Vasseur, S. Wakana, A. Walling, J. White, J. Wood, M. Zouberakis, Mammalian Genome 18 (3) (2007) 157–163. [6] C.F. Taylor, D. Field, S.-A. Sansone, J. Aerts, R. Apweiler, M. Ashburner, C.A. Ball, P.-A. Binz, M. Bogue, T. Booth, A. Brazma, R.R. Brinkman, A.M. Clark, E.W. Deutsch, O. Fiehn, J. Fostel, P. Ghazal, F. Gibson, T. Gray, G. Grimes, J.M. Hancock, N.W. Hardy, H. Hermjakob, R.K. Julian Jr., M. Kane, C. Kettner, C. Kinsinger, E. Kolker, M. Kuiper, N. Le Novère, J. Leebens-Mack, S.E. Lewis, P. Lord, A.-M. Mallon, N. Marthandan, H. Masuya, R. McNally, A. Mehrle, N. Morrison, S. Orchard, J. Quackenbush, J.M. Reecy, D.G. Robertson, P. RoccaSerra, H. Rodriguez, H. Rosenfelder, J. Santoyo-Lopez, R.H. Scheuermann, D. Schober, B. Smith, J. Snape, C.J. Stoeckert Jr., K. Tipton, P. Sterk, A. Untergasser, J. Vandesompele, S. Wiemann, Nature Biotechnology 26 (2008) 837. [7] A.M. Mallon, A. Blake, J.M. Hancock, Nucleic Acids Research 36 (2008) D715– D7188. [8] C.L. Smith, J.T. Eppig, Wiley Interdisciplinary Review Systems of Biology and Medicine 1 (3) (2009) 390–399. [9] H. Morgan, T. Beck, A. Blake, H. Gates, N. Adams, G. Debouzy, S. Leblanc, C. Lengger, H. Maier, D. Melvin, H. Meziane, D. Richardson, S. Wells, J. White, J. WoodThe EUMODIC Consortium, M. Hrabé de Angelis, S.D.M. Brown, J.M. Hancock, A.-M. Mallon, Nucleic Acids Research 38 (2010) D577–D585. [10] S.D. Brown, W. Wurst, R. Kuhn, J. Hancock, Annual Review of Genetics 43 (2009) 305–333. [11] K.L. McIlwain, M.Y. Meriweather, L.A. Yuva-Paylor, R. Paylor, Physiology and Behaviour 73 (2001) 705–717. [12] R. Paylor, C.M. Spencer, L.A. Yuva-Paylor, S. Pieke-Dahl, Physiology and Behaviour 87 (2006) 95–102. [13] T. Beck, H. Morgan, A. Blake, S. Wells, J.M. Hancock, A.-M. Mallon, BMC Bioinformatics 10 (Suppl. 5) (2009) S2. 1–9.xx. [14] H. Fuchs, K. Schugart, E. Wolf, R. Balling, M. Hrabe de Angelis, Mammalian Genome 11 (7) (2000) 528–530. [15] H. Fuchs, T. Lisse, K. Abe, M. Hrabé de Angelis, in: M. Hrabé de Angelis, P. Chambon, S.D.M. Brown (Eds.), Standards of Mouse Model Phenotyping, Wiley-VCH, Weinheim, 2006, pp. 35–86. [16] C.C. Glüer, K.G. Faulkner, M.J. Estilo, Osteoporosis International 3 (1993) 227– 235. [17] M.–F. Champy, M. Selloum, L. Piard, V. Zeitler, C. Caradec, P. Chambon, J. Auwerx, Mammalian Genome 15 (2004) 768–783. [18] S. Mandillo, V. Tucci, S.M. Hölter, H. Meziane, M.A. Banchaabouchi, M. Kallnik, H.V. Lad, P.M. Nolan, A.M. Ouagazzal, E.L. Coghill, K. Galle, E. Golini, S. Jacquot, W. Krezel, A. Parker, F. Riet, I. Schneider, D. Marazziti, J. Auwerx, S.D.M. Brown, P. Chambon, N. Rosenthal, G. Tocchini-Valentini, W. Wurst, Physiology and Genomics 34 (3) (2008) 243–255. [19] J.N. Crawley, R. Paylor, Hormones and Behavior 31 (1997) 197–211. [20] J.A. Gray, The Psychology of Fear and Stress, second ed., Cambridge University Press, Cambridge, 1987. [21] R. Trullas, P. Skolnick, Psychopharmacology 111 (1993) 323–331. [22] D.C. Rogers, E.M.C. Fisher, S.D.M. Brown, J. Peters, A.J. Hunter, J.E. Martin, Mammalian Genome 8 (1997) 711–713. [23] P.M. Nolan, J. Peters, M. Strivens, D. Rogers, J. Hagan, N. Spurr, I.C. Gray, L. Vizor, D. Brooker, E. Whitehill, R. Washbourne, T. Hough, S. Greenaway, M. Hewitt, X. Liu, S. McCormack, K. Pickford, R. Selley, C. Wells, Z. Tymowska-Lalanne, P. Roby, P. Glenister, C. Thornton, C. Thaung, J.A. Stevenson, R. Arkell, P. Mburu, R. Hardisty, A. Kiernan, A. Erven, K.P. Steel, S. Voegeling, J.L. Guenet, C. Nickols, R. Sadri, M. Nasse, A. Isaacs, K. Davies, M. Browne, E.M. Fisher, J. Martin, S. Rastan, S.D.M. Brown, J.A. Hunter, Nature Genetics 25 (2000) 440–443.

404

H. Gates et al. / Methods 53 (2011) 394–404

[24] V. Tucci, G. Blanco, P. Nolan, in: M. Hrabé de Angelis, P. Chambon, S.D.M. Brown (Eds.), Standards of Mouse Model Phenotyping, Wiley-VCH, Weinheim, 2006, pp. 135–176. [25] O.A. Meyer, H.A. Tilson, Byrd, M.T. Riley, Neurobehavioral Toxicology 1 (1979) 233–236. [26] S.P. Brooks, T. Pask, L. Jones, S.B. Dunnet, Genes, Brain and Behavior 3 (2004) 206–215. [27] J.N. Crawley, Brain Research 835 (1999) 18–26. [28] M.P. McFadyen, G. Kusek, V.J. Bolivar, L. Flaherty, Genes, Brain and Behavior 2 (2003) 214–219.

[29] M.A. Geyer, Psychopharmacology 147 (1999) 11–13. [30] A.M. Ouagazzal, F. Jenck, J.L. Moreau, Psychopharmacology 156 (2001) 273– 283. [31] S.G. Wilson, J.S. Mogil, Behavioral Brain Research 125 (2001) 65–73. [32] I. Racz, A. Zimmer, in: M. Hrabé de Angelis, P. Chambon, S.D.M. Brown (Eds.), Standards of Mouse Model Phenotyping, Wiley-VCH, Weinheim, 2006, pp. 221–235. [33] M. Klempt, B. Rathkolb, A. Aigner, E. Wolf, in: M. Hrabé de Angelis, P. Chambon, S.D.M. Brown (Eds.), Standards of Mouse Model Phenotyping, Wiley-VCH, Weinheim, 2006, pp. 87–107.