Here we present a step-by-step protocol for a semiautomated approach to analyze murine long-term electrocardiography (ECG) data for basic ECG parameters and common arrhythmias. Data are obtained by implantable telemetry transmitters in living and awake mice and analyzed using Ponemah and its analysis modules.
Arrhythmias are common, affecting millions of patients worldwide. Current treatment strategies are associated with significant side effects and remain ineffective in many patients. To improve patient care, novel and innovative therapeutic concepts causally targeting arrhythmia mechanisms are needed. To study the complex pathophysiology of arrhythmias, suitable animal models are necessary, and mice have been proven to be ideal model species to evaluate the genetic impact on arrhythmias, to investigate fundamental molecular and cellular mechanisms, and to identify potential therapeutic targets.
Implantable telemetry devices are among the most powerful tools available to study electrophysiology in mice, allowing continuous ECG recording over a period of several months in freely moving, awake mice. However, due to the huge number of data points (>1 million QRS complexes per day), analysis of telemetry data remains challenging. This article describes a step-by-step approach to analyze ECGs and to detect arrhythmias in long-term telemetry recordings using the software, Ponemah, with its analysis modules, ECG Pro and Data Insights, developed by Data Sciences International (DSI). To analyze basic ECG parameters, such as heart rate, P wave duration, PR interval, QRS interval, or QT duration, an automated attribute analysis was performed using Ponemah to identify P, Q, and T waves within individually adjusted windows around detected R waves.
Results were then manually reviewed, allowing adjustment of individual annotations. The output from the attribute-based analysis and the pattern recognition analysis was then used by the Data Insights module to detect arrhythmias. This module allows an automatic screening for individually defined arrhythmias within the recording, followed by a manual review of suspected arrhythmia episodes. The article briefly discusses challenges in recording and detecting ECG signals, suggests strategies to improve data quality, and provides representative recordings of arrhythmias detected in mice using the approach described above.
Cardiac arrhythmias are common, affecting millions of patients worldwide1. Ageing populations show a growing incidence and thus a major public health burden resulting from cardiac arrhythmias and their morbidity and mortality2. Current treatment strategies are limited and often associated with significant side effects and remain ineffective in many patients3,4,5,6. Novel and innovative therapeutic strategies that causally target arrhythmia mechanisms are urgently needed. To study the complex pathophysiology of arrhythmias, suitable animal models are necessary; mice have been proven to be an ideal model species to evaluate the genetic impact on arrhythmias, to investigate fundamental molecular and cellular mechanisms, and to identify potential therapeutic targets7,8,9. Continuous ECG recording is a well-established concept in the clinical routine of arrhythmia detection10.
Implantable telemetry devices are among the most powerful tools available to study electrophysiology in mice as they allow continuous recording of the ECG (a common approach is to implant the leads in a lead-II position) over a period of several months in freely moving, awake mice11,12. However, due to the huge number of data points (up to more than 1 million QRS complexes per day) and limited knowledge of murine standard values, the analysis of telemetry data remains challenging. Commonly available telemetry transmitters for mice last up to 3 months, leading to the recording of up to 100 million QRS complexes. This means that pragmatic analysis protocols are much needed to reduce the time spent with each individual dataset and will allow researchers to handle and interpret this huge amount of data. To obtain a clean ECG signal upon recording, transmitter implantation needs to be optimal-the lead positions should be as far apart as possible to allow higher signal amplitudes.
The interested reader may be referred to a protocol by McCauley et al.12 for more information. Further, to minimize noise, cages and transmitters must be placed in a silent environment not prone to any disturbance, such as a ventilated cabinet with controlled environmental factors (temperature, light, and humidity). During the experimental period, lead positioning must be checked regularly to avoid loss of signal due to lead perforation or wound healing issues. Physiologically, there is a circadian alteration in ECG parameters in rodents as in humans, generating the need for a standardized approach to obtaining baseline ECG parameters from a continuous recording. Rather than calculating mean values of ECG parameters over a long period, analysis of a resting ECG similar to that in humans should be performed to obtain basic parameters such as resting heart rate, P wave duration, PR interval, QRS duration, or QT/QTc interval. In humans, a resting ECG is recorded over 10 s, at a normal heart rate of 50-100/min. This ECG includes 8 to 17 QRS complexes. An analysis of 20 consecutive QRS complexes is recommended in the mouse as "resting ECG equivalent". Because of the above-mentioned circadian alteration, a simple approach is to analyze two resting ECGs per day, one at daytime and one at night time. Depending on the light on/off cycle in the animal facility, suitable times are selected (e.g., 12 AM/PM), and basic parameters are obtained.
Next, a heart rate plot over time is used to detect relevant tachy- and bradycardia, with consecutive manual exploration of these episodes to get a first impression. This heart rate plot then leads to the important parameters of maximum and minimum heart rate over the recorded period as well as heart rate variability over time. After that, the dataset is analyzed for arrhythmias. This article describes a step-by-step approach to obtain these baseline ECG data from long-term telemetry recordings of awake mice over a recording period of up to three months. Further, it describes how to detect arrhythmias using the software, Ponemah version 6.42, with its analysis modules, ECG Pro and Data Insights, developed by Data Sciences International (DSI). This version is compatible with both Windows 7 (SP1, 64 bit) and Windows 10 (64 bit).
1. Prearrangements
2. Analysis of basic ECG parameters
NOTE: In addition to validation/bad data marks, the software also automatically measures and calculates a large variety of derived parameters which are then reported in the Derived Parameter List.
3. Arrhythmia detection using pattern recognition (ECG PRO module)
NOTE: Ponemah's ECG PRO module uses selected QRS complexes as templates for further analysis. The ECG patterns of the templates are compared to all QRS complexes within the recording to calculate the percentage of similarity ("match") and to recognize arrhythmias (e.g., atrial or ventricular premature capture beats). The number of QRS complexes needed to be marked depends on the variability of the QRS-amplitude within the recording. In certain cases, selecting and marking one QRS complex gives a similarity of 80 percent with the respective recording, marking the majority of QRS cycles. However, this is an ideal case and during analysis, the number of QRS complexes that need to be marked as templates is usually higher.
4. Arrhythmia detection: a simplified manual approach using Data Insights
NOTE: For arrhythmia analysis, a correct annotation of P and R waves is necessary. However, even if clear P waves are visible within the ECG tracing, these P waves are sometimes not adequately identified even after adjusting the Attribute settings. As R waves are usually adequately recognized and annotated, a practical approach for further arrhythmia analysis using Data Insights is proposed below. For a general overview on arrhythmia detection using Data Insights and its predefined species-specific searches, the interested reader may be referred to Mehendale et al.13.
Recording long-term ECGs results in huge data sets. The options for further analyses are manifold and depend on the individual research project. This protocol provides a description of some very basic readouts that can be used by most researchers, especially for screening experiments, e.g., when characterizing a transgenic mouse line or when investigating the effects of a specific treatment in a disease model. A previous project involved the study of a novel drug candidate to determine whether it possessed cardiotoxic effects by analyzing ECG parameters over time. Telemetry transmitters were implanted 20 days before treatment, and ECG recordings were started 10 days before treatment to allow sufficient wound healing and acclimation of the mouse. Before treatment, the ECG was studied every three days; within the first week after treatment, the ECG was studied every day, after which the ECG was analyzed every seven days until the end of recording three weeks after treatment.
This approach allowed the detection of periods of reduced heart rate, increased atrioventricular (PR interval) and ventricular (QRS duration) conduction, as well as altered repolarization (QTc interval) in mice treated with the new drug as shown in Figure 5. This first step served as a "screening" that allowed the identification of time periods within the recording that potentially contained arrhythmias. A more detailed examination of the ECG revealed sinus pauses causing reduced heart rate two days after treatment and various degrees of atrioventricular (AV) blocks causing reduced heart rate six days after treatment. The latter finding was further supported by the prolonged PR intervals at this time point. To obtain these ECG parameters, 20 QRS complexes should be analyzed per time point and may therefore not be able to detect paroxysmal arrhythmia episodes at other time points.
To address this issue, it is advisable to specifically search for bradycardia and tachycardia episodes as well as for pauses using the ECG Pro module followed by manual review of detected episodes. This approach allows the detection of all relevant arrhythmias and the determination of the specific type of arrhythmia within the whole recording. For example, a tachycardia episode was detected in this study, which was identified as an atrial fibrillation.
As previously demonstrated, this approach further allows the determination of the time course of arrhythmia occurrence, e.g., the time to first AV block after macrophage depletion14. Representative traces, as shown in Figure 6, are obtained as described above (Figure 6A: normal sinus rhythm; Figure 6B: sinus pause; Figure 6C: AV-block I°, Figure 6D: AV-block II° type Mobitz 1; Figure 6E: AV block II° type Mobitz 2; Figure 6F: AV block III°; Figure 6G: atrial fibrillation).
Figure 1: Loading and reviewing data in Ponemah. (A) Load Review Dat dialog providing an overview of all the mice and signals recorded within the loaded experiment. (B) Graph Setup Dialog to set graphical windows providing both raw data (e.g., ECG signals) and derived parameters. Please click here to view a larger version of this figure.
Figure 2: Template setup in Ponemah. (A) Template Setup window to configure and select a new or browse already configured Template Library. (B) Graph setup page for Template settings. Please click here to view a larger version of this figure.
Figure 3: ECG tracings. (A) Screenshot of the windows containing the ECG trace; (B) heart rate plot; and (C) Template window. Please click here to view a larger version of this figure.
Figure 4: Analysis of attributes of an ECG tracing. (A) An ECG Analysis Attributes dialog. At the top of this dialog, several tabs (QRS, PT, Advanced, Noise, Marks, Notes, Precision) allow the adjustment of various settings. The settings are presented in the middle part of the dialog. At the bottom of the dialog, the ECG tracing is shown in the waveform window. At the top of the waveform window, the ECG tracing is shown; at the bottom, the derivative of the ECG tracing, including a visualization of the setting thresholds above, is shown. In the example presented here, a QRS Detection Threshold of 40% is defined, which is indicated by the pink background at the bottom. (B) Template Analysis Dialog: Select the desired Template Match Region to which all other ECG cycles will be compared. In this example, the T Wave is selected as the Match Region for analysis with a Minimum Match of 85%. This means that if the T Region does not match with at least 85% confidence, the cycle will not be marked as a match. Please click here to view a larger version of this figure.
Figure 5: Basic ECG Parameters over time in a drug intervention cohort. Blue panel: night time, yellow panel: daytime. From left to right: Heart Rate, PR interval, QRS duration, QTc interval. Please click here to view a larger version of this figure.
Figure 6: Representative ECG traces. (A) Normal sinus rhythm, (B) sinus pause, (C) AV-block I°, (D) AV-block II° type Mobitz 1, (E) AV block II° type Mobitz 2, (F) AV block III°, (G) atrial fibrillation. Scale bars = 100 ms. Abbreviation: AV = atrioventricular. Please click here to view a larger version of this figure.
Figure 7: Analysis flowchart. Abbreviation: HR = heart rate. Please click here to view a larger version of this figure.
The surface ECG is the primary diagnostic tool for patients suffering from heart rhythm disorders, providing insights into many electrophysiological phenomena. Nevertheless, sufficient analysis of cardiac surface ECG pathologies requires knowledge and definition of normal physiologic parameters. Many years of epidemiological research have led to broad consent on what is physiologic in humans and thus enabled physicians worldwide to clearly distinguish the pathologic. However, the analysis of surface ECG data is a major challenge in murine models; distinguishing between physiological and pathological ECG results can be difficult due to incomplete understanding and definition of basic ECG parameters15,16. In 1968, Goldbarg et al. were the first to describe ECG in healthy mice17. Besides showing heart rates and basic ECG patterns, such as PR interval and QRS duration, they described major differences between anesthetized and awake animals and differences between various anesthetics and different murine breeds, which was later confirmed by other groups16,17.
These early data emphasize why interpretation of murine ECG data is delicate and complicated. With growing interest in murine models for arrhythmia research in the past decades, more research has been focused on mouse electrophysiology and has generated evidence on the patterns of activation and repolarization in the mouse heart. The interested reader may be referred to a recent article by Boukens et al. for a detailed review of the murine ECG and its underlying currents15. Kaese et al. provided an overview on murine ECG standard values and major differences between human and murine ECG traces18. The first major difference is heart rate: healthy awake mice have a heart rate of 550-725 beats per minute, PR intervals of 30-56 ms, a QRS duration of 9-30 ms, and a repolarization phase that is very distinct from that observed in humans14. Further, the murine ECG regularly shows the occurrence of J-waves and a small and less distinctive T-wave, making analysis of the ST-segment and QT interval difficult18,19. Overall, murine models have become the most widely used model organism for cardiovascular research, including arrhythmias8.
Taking into consideration the above described interspecies differences that very likely also influence arrhythmogenesis, these models can provide valuable insights. The analysis of basic ECG parameters, such as heart rate and duration of different intervals, can be reliably done using software such as Ponemah, LabChart, or ECGAuto among many others with their respective analysis algorithms. Examples for data display are shown in Figure 5. Arrhythmia detection, however, is far more delicate, and there are no widely established approaches for murine long-term ECG analysis for arrhythmias. Different approaches have been used to overcome the technical and methodological difficulties associated with arrhythmia detection of long-term ECG recordings in mice. These approaches range from only using short recordings for the manual analysis for arrhythmias20 to simple considerations accepting inaccuracy as described by Thireau et al.21. These researchers performed heart rate variability analysis by simply excluding all sections of their recording with R-R intervals not contained in the range of the mean R-R interval ± 2 standard deviations to exclude all arrhythmias, ectopic beats, and artefacts without any manual review. This is the reason for this semimanual approach using Ponemah and its consecutive analysis modules, ECG Pro and Data Insights. This software solution can be used to analyze a vast range of physiologic signals, ranging from ECG in large mammals to blood pressure or temperature data in very small species.
The software comes with many resources on how to analyze different types of data. Nevertheless, although working quite well with ECG signals from larger animals, the low signal amplitude and therefore, high noise of signals derived from species, such as living and awake mice, can lead to a number of difficulties using a common approach to analysis. Noise will often mask P or T waves and thus disable the use of most of the predefined search rules within Data Insights. Care must be taken to define optimal values of the QRS detection threshold and to keep the attribute values used to identify QRS complexes and distinguish between clear cycles and noise events. A high percentage of the QRS detection threshold may result in undersensing (i.e., some R waves may be not detected), whereas a low percentage may result in oversensing (i.e., other peaks, such as T waves, may be misinterpreted as R waves). Further, specific questions in arrhythmia research in mice are understandably not the main topic of the materials provided by DSI, and finding specific information can be difficult. Within this protocol, a simple and pragmatic approach is used to define different arrhythmias extrapolating established human definitions.
For example, in human long-term ECG data, a pause longer than 3 s is considered significant22. This results in a human heart rate of 20/min., representing a third of the minimum physiologic heart rate of 60/min. As described by Kaese et al.18, the murine minimum physiologic heart rate equals 550/min., making 200/min. approximately a third of that rate. According to the human definition, pauses of more than 0.3 s can be assumed to be significant in mice. Further, it is a simple and pragmatic approach to describe differences in baseline parameters as relative changes to the respective control. This takes into consideration the differences between individual mouse lines and is an elegant way to identify the probable pathologic without relying on (often lacking) established normal values. This simple approach, summarized in Figure 7, is suitable for all groups studying cardiac arrhythmias in murine models using implantable telemetry devices. It leads to the evaluation of general ECG parameters as well as data on heart rate over time and the detection of a wide variety of arrhythmias. Therefore, this article attempts to provide a step-by-step approach for ECG and arrhythmia analysis and adds significantly to the guidance and manuals that have already been published.
The authors have nothing to disclose.
This work was supported by German Research Foundation (DFG; Clinician Scientist Program In Vascular Medicine (PRIME), MA 2186/14-1 to P. Tomsits and D. Schüttler), German Centre for Cardiovascular Research (DZHK; 81X2600255 to S. Clauss), the Corona Foundation (S199/10079/2019 to S. Clauss), the ERA-NET on Cardiovascular Diseases (ERA-CVD; 01KL1910 to S. Clauss), the Heinrich-and-Lotte-Mühlfenzl Stiftung (to S. Clauss) and the China Scholarship Council (CSC, to R. Xia). The funders had no role in manuscript preparation.