AnaSP: A software suite for automatic image analysis of multicellular spheroids

Highlights

• We present a new software suite to analyze brightfield images of spheroids.

• AnaSP estimates several morphological parameters in a very limited time.

• We proved the high accuracy of the segmentation method proposed.

• Both AnaSP source code and a standalone executable version are freely available.

• The GUI developed makes AnaSP effective even without expertise in computer vision.

Abstract

Today, more and more biological laboratories use 3D cell cultures and tissues grown in vitro as a 3D model of in vivo tumours and metastases. In the last decades, it has been extensively established that multicellular spheroids represent an efficient model to validate effects of drugs and treatments for human care applications. However, a lack of methods for quantitative analysis limits the usage of spheroids as models for routine experiments. Several methods have been proposed in literature to perform high throughput experiments employing spheroids by automatically computing different morphological parameters, such as diameter, volume and sphericity. Nevertheless, these systems are typically grounded on expensive automated technologies, that make the suggested solutions affordable only for a limited subset of laboratories, frequently performing high content screening analysis. In this work we propose AnaSP, an open source software suitable for automatically estimating several morphological parameters of spheroids, by simply analyzing brightfield images acquired with a standard widefield microscope, also not endowed with a motorized stage. The experiments performed proved sensitivity and precision of the segmentation method proposed, and excellent reliability of AnaSP to compute several morphological parameters of spheroids imaged in different conditions. AnaSP is distributed as an open source software tool. Its modular architecture and graphical user interface make it attractive also for researchers who do not work in areas of computer vision and suitable for both high content screenings and occasional spheroid-based experiments.

Introduction

Nowadays, it has been extensively established that monolayer cultures are a deficient model to validate the effects of drugs and treatments for different human care applications [1]. In particular, numerous cell types have been shown to behave differently when cultured in 3D conditions [2]. In this scenario, animal models are extensively used to perform in vivo experiments [3]. Nevertheless, the use of animals is highly controversial due to ethical and scientific reasons. For instance, interspecies’ differences in drug metabolism and toxicity between human and animal cells can jeopardize the assessment of the efficacy of the treatments [4]. Consequently, more and more biological laboratories produce and use 3D cell cultures and tissues grown in vitro as models of in vivo tumours [5] and metastases [6]. Mimicking the in vivo tumour microenvironment more closely than traditional monolayer cultures [7], they span the gap between standard 2D cultures and whole-animal models [8]. In particular, multicellular spheroids, cell aggregates of “large” dimensions (i.e., up to 2 mm of diameter [9]), are routinely employed as 3D in vitro models for testing drug dosages [10], radiotherapy treatments [11], and in general to define new care protocols for different tumours types [12]. Furthermore, they are also exploited as a tool for creating complex tissues [13].

The spheroids were originally proposed as a useful 3D model to study cancer in vitro by Sutherland et al. in the second half of the Twentieth century [14], and over time several different systems have been proposed to build spheroids and promote a wider usage in pre-clinical research and pharmaceutical development [15]. In particular, Kelm and Fussenegger [16] listed numerous cell lines able to form multicellular aggregated, and Achilli et al. [17] presented an extensive overview of systems available to build spheroids (e.g., antigravity bioreactors [18] and hanging drop plates [19]). Once built, to proceed with high-content screening experiments [20], the spheroids are then typically manually transferred (by using a pipette) in multi-well plates (one spheroid for well), typically a 384 [21] or a 96-well plate [22]. The morpho-biological organization of the cells composing the spheroids largely depends on the spheroids’ size [12], [23], [24]. In particular, the gradient of oxygen and glucose from the outer cell layers determines a stratification of cells characterized by different proliferation ratios [25]. Moreover, beyond a critical size of about 500 μm, central necrosis develops in most of the spheroids built from permanent cell lines [26]. Consequently, the spheroids’ core results similar to the avascular hypoxic regions of tumours [27]. In practice, the spheroids are considered a useful in vitro tumour model for a very wide range of applications [24].

Already in 2008 Lin and Chang [28] reported that more than 500 scientific articles describing multicellular spheroids as a tumour model used in drug and radiotherapy had been published. And the number of studies based on spheroids is quickly soaring. In these studies, many different biological, and morphological, parameters are typically monitored. For instance, diameters [25], perimeter [29], area [30], volume [31], and sphericity [32]. However, a lack of quantitative analysis’ methods limits the usage of spheroids as a reliable model and routine assessment of emerging therapies [33].

In this work we proposed AnaSP, a software suitable for analyzing spheroids and automatic computation of different morphological parameters, then correlated to the effect of drug dosages and treatments. AnaSP can be used in high throughput experiments, even high content screening based on spheroids. Four main reasons make AnaSP particularly attractive: First of all, AnaSP has been designed for experiments performed not necessarily using an automated microscope (i.e., endowed with a motorized stage). In practice, the images of the spheroids, typically placed in a multi-well plate (a single spheroid for each well) can be acquired manually, one at a time, by simply using a standard widefield microscope. Starting from the acquired images, AnaSP automatically computes the binary mask of the spheroid (i.e., black and white mask with value 1 assigned to the pixels belonging to the spheroid) and several morphological parameters, such as minor and major axis, equivalent diameter, perimeter, area, volume and sphericity. The only assumption we made is that the spheroids are characterized by a local spherical symmetry, and the only condition required is that every image to be analyzed contains a single spheroid of interest. In case of more spheroids present in the image, AnaSP by default analyses the biggest one, although a working modality considering more spheroids of similar size can be selected. Second, AnaSP was designed by using a modular architecture, allowing the user to choose which morphological parameters to compute automatically for each spheroid analyzed, also with the possibility of adding new parameters to be computed. Third, AnaSP is endowed with a Graphical User Interface (GUI) designed according to requirements provided by biologists that have no programming skills, and following their suggestions. Accordingly, it results particularly user friendly also for users not familiar with computer vision. Only the design of new functions, related to new morphological parameters to be estimated, requires some programming skills, but the template we provided with the source code of AnaSP strongly reduces the difficulties that could be met. Finally, AnaSP is distributed as an open source software tool written in MATLAB (©, The MathWorks, Inc., Natick, MA, USA), and both the source code and a standalone executable version (i.e., not requiring MATLAB being installed) are freely available at: http://sourceforge.net/p/anasp, together with sample images that can be downloaded as well. This allows a wide distribution of the software, also helping to increase the number of laboratories involved in performing spheroid-based high throughput experiments.

The experiments performed by using real-world images of spheroids, of different cell lines, and acquired under different setup conditions, proved the high sensitivity and precision of the segmentation method proposed, and the effectiveness of AnaSP to accurately estimate several morphological parameters of the spheroids in a very limited amount of time. Concluding, we proved that AnaSP can be used to perform quantitative spheroid-based high throughput experiments, and also occasional experiments with a limited number of spheroids, without requiring any expensive instrumentation (e.g., automated microscopes, cytometer, plate readers) and computer vision skills.

The next sections are organized as follows: Section 2 presents a short overview of the approaches available to automatically compute morphological parameters in spheroid-based experiments, Section 3 provides a detailed description of the proposed approach, Section 4 describes the material used in the experiments performed to validate the segmentation method proposed, while the results are presented and discussed in Section 5. Section 6 reports important notes to extend the use of AnaSP also to researchers not familiar with Information Technology. Finally, Section 7 summarizes the main findings of the work.