1 Introduction

As a key second messenger, Ca2+ regulates a wide range of cellular processes, including gene expression, endocrine secretion, muscle contraction and synaptic transmission [1, 2]. In fact, fluctuations in intracellular Ca2+ concentration ([Ca2+]i) are pleiotropic signals for the orchestration of pathways regulating cell functions under both physiological and pathophysiological conditions [3,4,5].

The dynamics of [Ca2+]i changes is complex and variable, depending on the origin of the stimulus and triggering signal, the cell type and the specific cell state [6, 7]. Moreover, stimuli of different origins generate specific Ca2+ signals with peculiar amplitude and frequency, decoded by different downstream Ca2+-sensitive proteins and different signaling pathways [3, 8].

Primary aldosteronism (PA), the most common cause of endocrine hypertension, is commonly caused by aldosterone-producing adenoma (APA), which induces an autonomous aldosterone production due to a set of somatic or germline mutations in ion channel and pump genes that ultimately lead to an increase in [Ca2+]i of aldosterone-producing cells [9,10,11]. When due to APA, PA is surgically curable by adrenalectomy. Adrenocortical cells rely on Ca2+ for the regulation of early and late limiting steps of steroidogenesis since Ca2+ elevation induces the transport of cholesterol to the Cytochrome P450 Family 11 Subfamily A Member 1 (CYP11A1) in the inner mitochondrial membrane (IMM), and stimulates aldosterone synthase (CYP11B2) gene expression via activation of calmodulin and Ca2+/calmodulin-dependent protein kinases [12,13,14].

Studies had investigated Ca2+ signaling dynamics in aldosterone-producing cells in rodents ex vivo and in adrenocortical cell lines [7, 15,16,17,18,19,20,21,22]. Some have reported a high cell-cell variability of Ca2+ signals, while others have provided quantitative Ca2+ trace processing by implementing computational models [19,20,21,22]. However, the algorithms utilized for the analyses of the recorded Ca2+ levels were not consistent and reproducible, which highlighted the need for a standardized methodology for the assessment of adrenal cell [Ca2+]i dynamics in response to physiological stimuli as well as in pathological conditions.

We, therefore, sought for developing an observer-independent methodology for the analysis of [Ca2+]i dynamics in isolated human CD56+  adrenocortical cells by quantitatively monitoring intracellular Ca2+ levels using the Ca2+-sensitive-indicator Fura-2 in resting conditions and after challenging them with Angiotensin II, one of the most potent physiological aldosterone secretagogues [23]. To provide standardized protocol for the analysis, we optimized a computational methodology that quantitatively assess Ca2+ signal recordings of individual adrenocortical cells over time, from a large data set without the need for advanced coding skills or sophisticated computer programs.

2 Methods

2.1 Cell Preparation and Culture

APA and APA adjacent tissues were collected in the operating room, cleared from surrounding fat and connective tissue, cells were mechanically and chemically dispersed and immune magnetically separated with anti-CD56 (neural cell adhesion molecule) antibody pre-coated magnetic beads as reported [24].

CD56+ cells, isolated according to methods of Caroccia et al. [24], were checked for viability using trypan blue and seeded on 24 mm poly-lysinated glass coverslips 24 h prior to Ca2+ imaging experiment in DMEM/F-12 (Dulbecco's Modified Eagle Medium/Nutrient Mixture F-12) supplemented with 1% Penicillin/Streptomycin, 1 % Insulin-Transferrin-Selenium (ITS), and 5 % cosmic calf serum (CCS).

The collection of samples was approved by the local ethics committee. Recruited patients provided signed consents.

2.2 Ca2+ Imaging

The day of experiment, CD56+ cells were washed twice with modified Krebs-Ringer Buffer (KRB) consisting of NaCl (135 mM), KCl (2.5 mM), MgSO4 x 7H2O (1 mM), MgCl2 x 6H2O (1 mM), K2HPO4 (0.4 mM), HEPES (20 mM), glucose (5.5 mM), and CaCL2 (1 mM) at pH 7.4. Afterwards, cells were incubated with 2 μM Fura-2-AM (Life Technologies, Milan, Italy) for 20 minutes at 37 °C in KRB added with 0.02% Pluronic acid and 250 μM sulphinpyrazone. After one wash with KRB, cells were transferred to a metal open-top holder and placed on the microscope stage for image acquisition. After acquisition of basal resting Fura-2 signal, cells were treated with Ang II at different doses (0.1 nM, 1 nM, 10 nM), then ATP (100 μM), and ionomycin (1 μM) were added to induce IP3-dependent ER Ca2+ release and to permeabilize plasma membrane in order to saturate the Ca2+ dye, respectively. Ca2+-dependent Fura-2 fluorescence levels was monitored using an inverted fluorescence microscope Nikon Eclipse Ti equipped with a Zyla-CMOS 4.2-P camera (Andor, Oxford Instruments, Abingdon, UK) and a 40x objective Nikon S Fluor Oil 1.30 DIC H/N2 WD 0.22). Excitation wavelengths were obtained with a 75 W Xenon Lamp (USHIO, UXLS50A) and a monochromator (Cairn Optoscan Monochromator, Cairn Research Ltd., Faversham, UK) controlled by NIS-ELEMENTS AR (Nikon) software provided alternating excitation wavelengths (340/380 nm). A neutral density filter, ND4 (Nikon, USA) and a FF-409-DiO3 Dichroic (Semrock) were used in the excitation pathway. Collection of emitted fluorescence was done using a 510/84 nm filter (Semrock). Images were acquired every 1 s with 100 ms exposure time, by a Zyla-CMOS 4.2-P (Andor, Oxford Instruments) controlled by the same software.

2.3 Data Processing and Workflow

A custom-made pipeline was developed to analyze the variations in individual cell Ca2+ signals assessed by Fura-2 (Fig. 1). The input was a confocal microscope image and the resulting output was distinct individual peaks with peak parameters associated to each. The pipeline enclosed 3 main steps: 1) data extraction and construction, 2) trace plotting and peak identification and 3) peak filtering.

Fig. 1
figure 1

Workflow of custom-made peak detection pipeline. Input data are the raw confocal microscope images of the Ca2+-sensitive probe fluorescence and the output results are dynamics parameters (number, frequency and shape) of individual cell Ca2+ peaks. The pipeline comprises 3 data processing steps. Step 1 (Data Extraction and Construction): assignment of ROIs, extraction of Fura-2 signal value for the two excitation wavelengths (Fiji-Image J), normalization of data to the background, and calculation of Fura-2 340/380 nm signal ratios (Microsoft Excel). Step 2 (Trace Plotting and Peak Finding): data smoothing followed by peak detection through the application of the OriginPro software algorithm. Step 3 (Peak Filtering): implementation of Microsoft Excel’s equations to identify spontaneous vs Ang II-evoked Ca2+ peaks. For Ang II-evoked peaks, first and subsequent oscillations are considered separately

Full size image

We used ImageJ-based Fiji software, Microsoft Excel 2019 MSO (Version 2304 Build 16.0.16327.20200), and OriginPro 2023 software (10.0.0.154 academic version, OriginLab Corporation, Northampton, MA, USA-).

Data extraction was executed manually on ImageJ-based Fiji image analysis software by assigning individual cell as a region of interest (ROI) on raw confocal images (Fig. 2). Fura-2 is a dual excitation ratiometric Ca2+ indicator commonly used for the measurements of intracellular [Ca2+]. The Ca2+-free form of Fura-2 has a peak excitation wavelength at ~ 380 nm, whereas the Ca2+-bound form has a peak excitation wavelength at ~ 340 nm. An elevation of Ca2+ concentration induces an increase in Fura-2 emission fluorescence when the indicator is excited at 340 nm, with a corresponding decrease in fluorescence at 380-nm excitation. The ratio between emission after excitation at 340 nm and 380 nm (340/380 ratio) represents a good estimation of the level of Ca2+ in the sample [25]. The fluctuations of Fura-2 fluorescence emission at the 2 different exciting wavelengths (340 nm and 380 nm) were recorded and extracted separately, each normalized to the relative background noise (a region of the coverslip devoid of cells), then expressed as 340/380 ratio.

Fig. 2
figure 2

Representative confocal image of CD56+ cell after 24 hours in culture. a Raw image, b same as A after ROIs assignment. Black dots represent anti-CD56+ magnetic beads used for cell isolation. ImageJ-based Fiji software was utilized to assigned ROIs. Scale bar = 50 μM

Full size image

3 Results

3.1 Trace Plotting and Peak Identification

Ca2+ traces were plotted using OriginPro lab software. However, some samples present undesired background fluctuations that were reduced by applying a noise reduction algorithm based on the Savitzky-Golay (SG) method to all sample traces [26,27,28]. The SG method fits a polynomial regression to successive data points of a moving window (local least-squares polynomial approximation) to recover the trace shape. The degree of smoothness was optimized at a window size = 20 which resembles the number of data points used in each local regression. The mentioned data processing tool achieves data smoothing while preserving the distinct features and properties of various data patterns and frequencies. The mentioned settings have been determined for few traces selected as templates and then used in the following batch analysis for the required large data set.

As shown in Fig. 3, in both AAC and APA cells, the Ca2+ signals do not always return to the baseline level after Ang II stimulation. Thus, establishing of a straight-line baseline functional to the event detection is rather inefficient. Therefore, a critical step towards processing dynamic live cell recordings is to adopt a baseline calculation approach capable to drift and adapt to the peculiarity and specificity of the signal of each cell. Accordingly, to estimate the baseline, we implemented the “Asymmetric least square smoothing baseline setting” (ALS) in the peak detection algorithm of OriginPro software. A threshold at 0.05, smoothing factor at 5, with a number of iterations set at 10 were the parameters implemented to run the algorithm. These settings correct the baseline by asymmetrically weighting the trace deviations and the estimated baseline adjusts to the highly dynamic [Ca2+]i oscillations, allowing to adapt to the drift in signals (Fig. 4a). The settings of the peak finding algorithm were set as follows: Direction = positive; Method = Local Maximum; Local points=10. In addition to that, the algorithm promoted detection of peaks following the indicated peak filtering settings: Method = By Height; Threshold height= 5 %.

Fig. 3
figure 3

Baseline detection. a, b Representative Ca2+ signal traces of 2 different AAC showing the inadequacy of setting a straight-line (red line) as baseline when performing dynamic Ca2+ recordings and peak detection. In these samples, the set baseline vanishes the detection of oscillations (left panel) and fails to accurately detect AUC (right panel)

Full size image
Fig. 4
figure 4

Modeling human CD56+ AAC and APA cells intracellular Ca2+ dynamics by standardized peak detection and characterization. a. Baseline setting (red line) was adapted to the dynamic nature of the Ca2+ signals using the OriginPro 2023 OriginLab Corporation lab software as described in the peak detection algorithm settings. Ang II was used of [10 nM] to induce the Ca2+ peaks. ATP was used to induce non receptor mediated increase of [Ca2+]i. To confirm the cell responsiveness, ionomycin was used of [1μM]. Numbers on peaks represent the time in seconds at which each event has occurred. i + ii: AAC, iii + iv: APA. b Peak characterization was performed by calculating different features of each individual Ca2+ peak: amplitude (red line); full width at half maximum (FWHM, orange line); area under the curve (AUC, green area)

Full size image

We focused our analysis on selected peak features, defined as: (i) amplitude from baseline (height of the peak), (ii) area under the curve (AUC, which represents an estimation of the total amount of Ca2+ entering the cell) and (iii) full width at half maximum (FWHM, which describes the extension/duration of the peak (Fig 4b). The results of the optimized algorithm described are identified individual peaks (Fig 5a–d).

Fig. 5
figure 5

Representative Ca2+ traces of AAC (a, b) and APA (c, d) cells with detection of Ca2+ peaks according to our pipeline. Algorithm developed shows flexibility and adjustability to different types of behaviors. Numbers on peaks represent the time in seconds at which each event has occurred

Full size image

3.2 Peak Filtering

The strategy here developed allowed flexible application of the pipeline to variable experimental conditions, such as the addition of multiple stimuli and the time of addition of the stimulus itself.

Through simplified equations designed on Microsoft Excel 2019 MSO, we could filter out residual noise and nonspecific signals thus performing a second step cleaning of the data, where peaks posing an area less than 0.5 combined with a height less than 0.016 were removed, using the following the equation:

$$= If ((Area \, column < 0.5) * (height \, column < 0.016, {}^{^{\prime\prime}}Delete^{^{\prime\prime}} )$$

The Ca2+ traces could be subdivided in various regions for subsequent detailed analysis according to the experimental procedure applied. Here, we defined a region where spontaneous intracellular Ca2+ activity occurs (i.e. before the addition of the stimulus) and a region where Ca2+ fluctuation occurs in response to the stimulus (Ang II-evoked peaks). The time of Ang II addition discriminates the two regions in the Ca2+ trace plot.

In case the addition of Ang II was provided at sec 130, the equation used is:

$$= If (time cell < 130, \;{}^{^{\prime\prime}}spon^{^{\prime\prime}} , If (AND (time cell \ge 130, time cell < 260), \;{}^{^{\prime\prime}}10 nM dose^{^{\prime\prime}} , \;{}^{^{\prime\prime}}none^{^{\prime\prime}} ))$$

However, Ca2+ traces could adopt either a unique single Ca2+ transient behavior upon stimulus addition or an oscillatory behavior as depicted before in Figs. 3 And 4 [29]. In case of a multi peak behavior, it is applicable to segregate detected individual Ca2+ peaks in a time dependent manner to allow the differentiation between the first response transient and the subsequent peaks following this equation (with A3 and A2 demonstrating the number code of the cell);

$$= If \left( {A3 = A2, OS \left( {oscillating} \right), {}^{^{\prime\prime}}First \, Peak^{^{\prime\prime}} } \right)$$

To be able to quantify the number of cells showing unique or a multi peak Ca2+ behavior, we employed the subtotal count function on Excel. As individual ROIs are indicated by different ROI-ID number, we sum all the detected peaks originating from 1 ROI using the distinct ID numbers and then we apply the following equation on the subtotal function results which resembles the number of ROIs showing oscillation activity with column I representing the subtotal count function results Supplementary file1.

$$= COUNTIF(I4:I172,{}^{^{\prime\prime}} > = 2^{^{\prime\prime}} )$$

Frequency of oscillations was also characterized when 2 or more peaks were detected (F= 1/T) (period=T represents time for one oscillation) (Fig. 6). To calculate the period (T), copy the Id index (ROI name) and index (Time) with the subtotal to a new sheet, then subtract each 2 consecutive row indexes \((= B4-B3\)). Remove nonspecific values resulting from data ID count by conditional formatting, highlighting cells, then pick the Text that contain (count). Filter out Column A (data Id ROIs) by color then Column C (T): by number, greater than 0. Continue to calculate the frequency in Hz(= 1/T).

Fig. 6
figure 6

Time defined segments of a single Ca2+ trace in CD56+ aldosterone-producing cell. All Ca2+ recordings followed the sequence shown: spontaneous phase (unstimulated conditions), Ang II-stimulated phase divided into first transient and subsequent peaks, ATP and ionomycin stimulated phases. Different segments are marked with different colors. Period of each phase differs for each slide of cells

Full size image

Ca2+ signals plots were divided into resting conditions, in which spontaneous activity could be detected (light blue), and Ang II-stimulated conditions (light pink).

In our experiment, we used ATP (yellow) and ionomycin (light gray) as controls to determine the responsiveness and viability of the cells, thus the corresponding peaks were eliminated from detection by the algorithm (Fig. 6).

4 Discussion

We herein describe a novel pipeline methodology to analyze ex vivo Ca2+ imaging data of individual adrenocortical cells in a simple and reliable way. Thus far, many computational tools for analyzing intracellular Ca2+ changes that have been used require programming skills and knowledge of software languages and scripts [19,20,21,22] that are not familiar to a significant part of the biologists conducting experimental work [30, 31]. Therefore, the designed workflow described here represents a convenient approach to produce reproducible data and encourages autonomy of researcher with low or even zero prior bioinformatic skills.

The workflow of the pipeline provides an optimized protocol to use the algorithm embedded in the OriginPro 2023 software to promote the best peak detection approach. Processing data using Savitzky-Golay method (Windows 20) promotes a clear peak display of different Ca2+ behavior while it preserves small scaled peaks. The pipeline provides the optimum settings for baseline adjustment and peak identification algorithm to allow the analysis of irregular and complex patterns of Ca2+ kinetics associated with highly dynamic baseline drifts. The user is only left with the task of modifying the softwares settings without need of any code scripting, as herein described.

To allow detected peaks segmentation based on the time of stimuli addition, Microsoft Excel-based equations are designed. They require careful manual handling to ensure accurate peaks characteristics segregation.

It is worth noting that the OriginPro 2023 software already provides an App to analyze Ca2+ signals of patient-derived induced pluripotent stem cell cardiomyocytes (iPSC-CMs). However, as it requires an embedded Python and NumPy, SciPy libraries, and furnishes only an average information of all peaks per each assigned cell and not an assessment of individual peaks. Therefore, this approach isinadequate to provide a detailed detected-peaks analysis.

The discussed analysis of intracellular Ca2+ fluctuations proved to be useful for investigating altered Ca2+ dynamics when applied to assess outputs of Ca2+ events in primary human cells isolated from aldosterone-producing adenoma (APA) and APA-adjacent (AAC) tissues of patients with PA, i.e. the most common curable form of arterial hypertension. It allowed to quantitatively monitor intracellular [Ca2+]i fluctuations of isolated human CD56+ adrenocortical cells using the Ca2+-sensitive indicator Fura-2 after Angiotensin II stimulation by a dynamic baseline drifting in order to process a large scale of heterogeneous data [4, 16, 32].

Thus, the methodology developed in this study uses reliable and easy to handle algorithms to track intracellular Ca2+, with no  requirement of scripts designing, thus abolishing the possibility of code scripting. This provides users with the ability to process datasets autonomously without prior programming knowledge. The simple approach offers the analysis of periodic and fading peaks. The settings of the peak detection algorithm are designed and tested to be highly efficient on various types of real time traces.

Importantly, the flexibility and robustness of the pipeline can be standardized across multiple studies as it yields non-subjective results and is applicable to large data sets and minimizes the possibility for human errors. The only con of our methodology is the inevitable need for manual handling, more specifically step 3, which includes the peak filtering Excel designed equations with careful operation, yet considered to be time efficient. Thus careful performance is required.

Of note, when we applied this methodology on our data set, we observed calcium behaviors upon Ang II stimulation that are reproducible in cells in close proximity.  Our primary cells in culture show a very low proliferation rate and thus what we see in our experimental set up likely reflect the in vivo behavior. From dispersed CD56+ adrenocortical cells,  after 24 hours in culture, we did observe the formation of multicellular structures resembling the cyto-architecture of intact human adrenocortical glomeruli structures, the so called ‘rosette”. The functional behavior of cell assemblies, which were extensively studied and proven to coordinate Ang II-elicited calcium spikes in murine models [21, 33, 34], was not investigated in human CD56+ dispersed cells before and we are pleased to provide the scientific community with the method to analyse the crucial feature of acrenocortical cell biology.