PunctaSpecks: A tool for automated detection, tracking, and analysis of multiple types of fluorescently labeled biomolecules

Highlights

• A comprehensive computational tool for processing and characterizing imaging data from fluorescence microscopy experiments.

• Characterizing imaging data with information about multiple biomolecules or biological processes.

• Calculating the number, areas, life-times, and amplitudes of fluorescence signals arising from multiple sources.

• Tracking diffusing fluorescence sources like moving mitochondria

• Determining the overlap probability of two processes or organelles imaged using indicator dyes of different colors.

• Applying the tool to different data sets recorded under physiological and pathological conditions including.

Abstract

Recent advances in imaging technology and fluorescent probes have made it possible to gain information about the dynamics of subcellular processes at unprecedented spatiotemporal scales. Unfortunately, a lack of automated tools to efficiently process the resulting imaging data encoding fine details of the biological processes remains a major bottleneck in utilizing the full potential of these powerful experimental techniques. Here we present a computational tool, called PunctaSpecks, that can characterize fluorescence signals arising from a wide range of biological molecules under normal and pathological conditions. Among other things, the program can calculate the number, areas, life-times, and amplitudes of fluorescence signals arising from multiple sources, track diffusing fluorescence sources like moving mitochondria, and determine the overlap probability of two processes or organelles imaged using indicator dyes of different colors. We have tested PunctaSpecks on synthetic time-lapse movies containing mobile fluorescence objects of various sizes, mimicking the activity of biomolecules. The robustness of the software is tested by varying the level of noise along with random but known pattern of appearing, disappearing, and movement of these objects. Next, we use PunctaSpecks to characterize protein-protein interaction involved in store-operated Ca²⁺ entry through the formation and activation of plasma membrane-bound ORAI1 channel and endoplasmic reticulum membrane-bound stromal interaction molecule (STIM), the evolution of inositol 1,4,5-trisphosphate (IP₃)-induced Ca²⁺ signals from sub-micrometer size local events into global waves in human cortical neurons, and the activity of Alzheimer’s disease-associated β amyloid pores in the plasma membrane. The tool can also be used to study other dynamical processes imaged through fluorescence molecules. The open source algorithm allows for extending the program to analyze more than two types of biomolecules visualized using markers of different colors.

Introduction

Recent advances in imaging techniques for visualizing biological processes at unprecedented spatiotemporal scales ranging from single molecule to whole-cell signals have revolutionized biological research [[1], [2], [3], [4], [5]]. It is now possible to study dynamical processes such as the flux through individual ion channels, diffusion of proteins in cell membrane, and movement of motor proteins and filaments along microtubules [[6], [7], [8], [9], [10], [11]]. These experiments often generate huge amount of imaging data on the dynamics of hundreds or thousands of such objects (terms objects, puncta, or events are used interchangeably throughout the manuscript) that needs to be analyzed quantitatively and efficiently to reveal their functional properties. Manual or semi-automatic analysis of these data sets is labor intensive, costly, inaccurate, and poorly reproducible, undermining the usefulness of technological advances. Thus, automated analysis of the data to extract quantitative information about the dynamics of the underlying biological processes is imperative.

Depending on the biological question being asked, fluorescence imaging data from a wide range of experimental scenarios such as single and multiple types of objects, variable particle density, and congested physiological environments (noise) need to be processed. These objects can be proteins tagged with fluorescent proteins or labeled by staining with specific dyes. More often these processes evolve in time and space. For example, Ca²⁺ release events evolve from blips or quarks due to opening of individual channels for a few milliseconds to micrometer-sized puffs and sparks caused by the concerted opening of a few closely located channels for hundreds of milliseconds to whole-cell waves lasting for minutes [12]. Imaging experiments record these events and their progression in a noisy cellular environment by recording changes in cell fluorescence over time [[13], [14], [15], [16], [17]]. A high throughput analysis of these events is key for understanding their function and specificity [12,18,19].

Some biological processes, on the other hand, remain confined to a very small space. They are short lived and can coexist in a large number. Yet, their dynamics have crucial implications for the cell function and survival. For example, cation-permeable pores formed by soluble beta amyloid (Aβ) aggregates in the cell membrane can destabilize the cell ionic homeostasis, and are believed to contribute to the pathology of Alzheimer's disease [[20], [21], [22], [23], [24]]. The activity and evolution of these pores is often recorded in multi-gigabyte image stacks using high throughput, massively parallel optical patch clamp technique [8,25]. Analyzing the spatiotemporal dynamics of localized events as they evolve into global signals or thousands of small spatially confined, short-lived events require accurate and efficient numerical methods.

In some situations, fully understanding a biological process requires characterizing the collective or concurrent behavior of more than one type of molecules or reactions visualized using markers of different colors. For instance, store operated Ca²⁺ entry (SOCE) refers to a process by which Ca²⁺ channels localized in the plasma membrane (PM) are activated when luminal Ca²⁺ stores in the endoplasmic reticulum (ER) are depleted. In many cells, SOCE is mediated by the highly Ca²⁺-selective plasma membrane channel, ORAI1, and the ER-resident Stromal Interaction Molecules (STIM1 and STIM2) that sense changes in the Ca²⁺ concentration of the ER lumen ([Ca²⁺]_ER). ORAI1 and STIM1/2 are labeled with two different fluorescent tags to record their activity simultaneously. Following cell stimulation that results in [Ca²⁺]_ER reduction, STIMs aggregate and translocate to the ER–PM junctions where they recruit and activate ORAI1. These junctions are sites where the ER membrane is closely apposed to the PM. Localization of ORAI1 with STIMs in these junctions compartmentalize the Ca²⁺ signaling machinery to discrete sites where the Ca²⁺ signals generated are channeled to activate distinct downstream cellular processes. This ensures the high signaling fidelity of responses triggered by ORAI1-mediated Ca²⁺ signals from the ER–PM junctions [[26], [27], [28], [29]]. The movement of these proteins in live cells following cell stimulation has been reported using various imaging techniques, such as confocal and total internal reflection fluorescence (TIRF) microscopy. The time-lapse image stacks acquired over several minutes show that they are highly dynamic and mobile following stimulation.

As mentioned above, the high-resolution imaging experiments record the activity of many objects or events usually immersed in a physiologically dense environment, introducing background noise during image acquisition. While the background noise can be removed, delineating individual objects or events and tracking their movement, amplitudes, spatial sizes, and lifetimes over time remains a major challenge. These types of data-sets demand efficient automated tools for accurately identifying and tracking the dynamics of each object over time to elucidate the underlying biological processes. A number of tools have been developed over the past either de novo or adapting preexisting tools tailored for the specific problem [24,[30], [31], [32], [33], [34], [35], [36], [37], [38], [39]]. In particular, several tools for analyzing localized Ca²⁺ signals in cardiac myocytes have been developed over the years [[40], [41], [42], [43], [44], [45], [46], [47]].

Due to the diverse spatiotemporally dynamical nature of the biological processes, there does not exist a “one-size-fits-all”, universally applicable method. Therefore, specialized tools tailored to address the problem at hand is often the way to go. We have developed an all-in-one MATLAB-based tool called PunctaSpecks, with a graphical user interface (GUI), that implements algorithms for identifying and characterizing fluorescently labeled objects in time-lapse image stacks. PunctaSpecks generates outputs that report on the distributions of spatial sizes, amplitudes in terms of mean or maximum intensity, mean active times, mean silent times, mobility patterns, and time-resolved trajectories of all objects or events in the recording, and overlap probabilities of two different types of objects such as ORAI1 and STIM in a given frame. We have implemented a variety of algorithms for preprocessing these movies to identify objects and extract their features of interest. A range of options are available to perform various analysis tasks on the identified objects and export data for later analysis. The open source MATLAB algorithm can be modified to analyze more than two types of biomolecules visualized using markers of different colors.