1. Introduction
Assessing cell viability is a convenient and fundamental method to analyse the effects of various stressors on yeast cells in scientific research and in any brewing process, in which cell counting-associated technologies, such as concentration calculations and spotting tests, are widely adopted to provide an estimation of viable yeast cells [1]. These assays can be used to measure the results of yeast proliferation; to test the growth rate of yeast cells under different kinds of chemical, physical, or environmental factors; and as an internal control to achieve consistent fermentations in industry. At present, commonly used cell counting methods include plate counting [2], real-time quantitative PCR [3], haemocytometers [4], automatic cell counting instruments [5]and flow cytometry counting in biological operation.
The plate counting method is performed by spreading living cells on solid media to form colony forming units (CFUs), which can be counted with the naked eye. A corresponding method for automatic colony counting with ImageJ software has been developed [6]. The advantage of this method is that non-viable yeast cannot duplicate and form colonies on plates, but some shortcomings of this method are that the number of yeast cells only depends on different dilution concentrations and clumped cells will be registered as one count. Additionally, plate counting is also time consuming. Real-time quantitative PCR involves the application of related instruments and the drawing of accurate standard curves. Of course, this method relies on advanced expensive equipment and requires the pre-establishment of control genes, high-efficiency primers or the fluorescent dye PMA [7]. Most cell counters can only count specific types of cells. Professional yeast counters used in fermentation engineering are always expensive and may be assisted with stains that are unfavourable to the operator. In addition, flow cytometry changes the breakpoint of droplets so that the size of droplets formed by sheath fluid can wrap a cell [8], but it demands a more uniform cell size and can be better detected when cells exhibit certain fluorescence signals. However, the size of yeast cells varies widely, and a droplet may contain multiple yeast cells. As a consequence, it is impossible to count yeast cells accurately without the insertion of a fluorescent protein. Therefore, haemocytometer counting remains the most commonly used technique to determine the concentration of a yeast sample because of its ease of use and low cost [9; 10; 11; 12; 13].
Problems do exist with the haemocytometer method, such as it being time consuming and inefficient to implement for large-scale analyses. Using a photomicroscope, some researchers have developed effective software to automatically count cells on haemocytometer plates to avoid subjective manual counting and high-throughput statistics. However, some of these methods are limited to specific types of cells. For example, CellProfiler [14] can count mammalian and non-mammal cells via a high-throughput analysis. However, non-mammalian cells are limited to round cells, fission yeasts and breeding budding yeasts. CellC [15]can only count labelled cells. CellCounter [16]and OpenCFU [17] are designed for specialized cells and cannot count high concentrations of yeast. Moreover, in cases in which a non-homogeneous liquid is used to test yeast viability or there is background material in the fermentation process, making yeast cell counting even more challenging. Therefore, we intended to introduce the use of the free ilastik and ImageJ software for batch enumeration of yeast cells in complicated backgrounds.
As an open source image analysis program that can be run under the Macintosh, Windows and Linux operating systems, ImageJ has been reported to be suitable for mammalian cell counting [18]. Considering that mammalian cells with an irregular morphology generally grow and adhere to the wall and that their cell size is usually larger than that of yeast, mammalian cells are much easier to count than yeast cells. Hence, whether ImageJ can be used for yeast cell counting remains to be determined. Moreover, the culture media of mammalian cells has fewer impurities, and the background is easy to separate from cells in automatic counting. However, depending on the different purposes of yeast cultures, impurities in yeast media vary tremendously in type and quantity, which would interfere with the automatic counting of yeast cells by ImageJ. Currently, ilastik software can solve this problem very well. ilastik is easy to operate and can provide end users with machine learning-based image analysis [19]without substantial computational expertise; thus, we mainly use the workflow of ilastik to segment and classify images to maximize the separation of yeast cells from the background.
Taking the yeast Cryptococccus deneoformans as an example, a new rapid automated yeast cell counting method using ilastik and ImageJ was assessed in this study. As an opportunistic pathogenic fungus, the spotting test is frequently used to study the sensitivity of C. deneoformans cells to various environmental stresses and drugs, in which cell counting is the first step. Here, we describe an ImageJ macro, named “Yeast Counter”, and systematically test its performance for counting yeast cells. For samples in complicated backgrounds, the ilastik workflow was able to perform segmentation and classification with interactively supervised machine learning. Then, the number of yeast cells was counted using the “Yeast Counter” macro, which can set up customizable parameters based on cell size, perimeter, roundness and so on in the batch processing mode. According to the results of the spotting test, we observed that the customizable software algorithm for yeast counting reduced inter-operator errors significantly and generated accurate and objective results, while manual counting with a haemocytometer exhibited some errors between repeats and required more time than “Yeast Counter”.