To quantify microvascular flow from high speed capillary flow image sequences, we developed STAFF (Spatial Temporal Analysis of Fieldwise Flow) software. Across the full image field and over time, STAFF evaluates flow velocities and generates a sequence of color-coded spatial maps for visualization and tabular output for quantitative analyses.
Changes in blood flow velocity and distribution are vital in maintaining tissue and organ perfusion in response to varying cellular needs. Further, appearance of defects in microcirculation can be a primary indicator in the development of multiple pathologies. Advances in optical imaging have made intravital microscopy (IVM) a practical approach, permitting imaging at the cellular and subcellular level in live animals at high-speed over time. Yet, despite the importance of maintaining adequate tissue perfusion, spatial and temporal variability in capillary flow is seldom documented. In the standard approach, a small number of capillary segments are chosen for imaging over a limited time. To comprehensively quantify capillary flow in an unbiased way we developed Spatial Temporal Analysis of Fieldwise Flow (STAFF), a macro for FIJI open-source image analysis software. Using high-speed image sequences of full fields of blood flow within capillaries, STAFF produces images that represent motion over time called kymographs for every time interval for every vascular segment. From the kymographs STAFF calculates velocities from the distance that red blood cells move over time, and outputs the velocity data as a sequence of color-coded spatial maps for visualization and tabular output for quantitative analyses. In normal mouse livers, STAFF analyses quantified profound differences in flow velocity between pericentral and periportal regions within lobules. Even more unexpected are the differences in flow velocity seen between sinusoids that are side by side and fluctuations seen within individual vascular segments over seconds. STAFF is a powerful new tool capable of providing novel insights by enabling measurement of the complex spatiotemporal dynamics of capillary flow.
The microvasculature plays a critical role in physiology, ensuring effective perfusion of tissues under changing conditions. Microvascular dysfunction is associated with myriad conditions including long-term cardiovascular morbidity and mortality, development of dementia, and disease of both liver and kidney and thus is a key factor of interest in a broad range of biomedical investigations1,2,3,4,5. While multiple techniques have been used to evaluate tissue perfusion, only intravital microscopy enables data collection at the temporal and spatial resolution necessary to characterize blood flow at the level of individual capillaries.
Microvascular flow can be visualized in fluorescence microscopy either by the movement of fluorescent microspheres or by the movement of red blood cells against the background of membrane-impermeant fluorescent markers (e.g., fluorescently-labeled dextran or albumin)6,7. Microvascular flow can be imaged in superficial cell layers using widefield microscopy, or at depth using either confocal or multiphoton microscopy. However, capillary flow rates are such that the passage of red blood cells cannot generally be captured at speeds less than 60 frames/s. Since most laser scanning confocal and multiphoton microscopes require 1–5 s to scan a full image field, this speed can generally be accomplished only by limiting the field of view, sometimes to a single scan line8. The process of limiting measurements to selected capillary segments (1) has the potential to introduce selection bias and (2) makes it impossible to capture spatial and temporal heterogeneity in the rates of capillary blood flow. In contrast, images of capillary networks can be collected at speeds exceeding 100 fps using widefield digital microscopes equipped with scientific complementary metal oxide semiconductor (sCMOS) cameras9,10. These inexpensive systems, common in typical biomedical laboratories make it possible to image microvascular flow across entire two-dimensional networks, essentially continuously. The problem then becomes one of finding an analysis approach that is capable of extracting meaningful quantitative data from the massive and complex image datasets generated by high-speed video microscopy.
To enable analysis of full-field flow data we have developed STAFF, novel image analysis software that can continuously measure microvascular flow throughout entire microscope fields of image series collected at high speed11. The approach is compatible with a variety of different experimental systems and imaging modalities and the STAFF image analysis software is implemented as a macro toolset for the FIJI implementation of ImageJ12. The underlying principle used here to visualize microvascular flow is that first, some contrast must be provided to be able to image the red blood cells within capillaries. In our studies, contrast is provided by a bulk fluorescent probe that is excluded by the red blood cells. The velocity of flow can then be quantified from the displacement of the red blood cells that appear as a negative stain within the fluorescently labeled plasma in images collected at high speed from a living animal8. We then use STAFF to make plots of distance along each capillary segment over multiple intervals of time called kymographs, then detect the slopes present in the kymographs13, and from those slopes calculate the rates of microvascular flow. The approach can be applied to images collected from any capillary bed that can be accessed for imaging. Here we describe the application of IVM and STAFF to studies of blood flow in the liver.
All animal experiments were approved and conducted according to the Institutional Animal Care and Use Committee guidelines of Indiana University, and adhered to the NRC guide for the care and use of animals.
1. Surgical Preparation for Intravital Microscopy
NOTE: This is not a survival surgery. Once section 1 "Surgical preparation for intravital microscopy” is begun, work cannot be paused until the completion of section 2 "Intravital microscopy”.
2. Intravital Microscopy
3. Define the Vascular Network Using TrakEM2 in FIJI
NOTE: The protocol can be paused after saving work at any point in section 3.
4. Prepare the Movie Sequence for STAFF Analysis
5. Install STAFF Macros into FIJI
NOTE: (Important) Multiple folders within the FIJI folder subdirectories have similar file names, some capitalized, some not. Be certain that the correct folders are selected when installing STAFF.
6. Quantifying Vascular Flow Using STAFF
7. Quantitative Analysis Using STAFF Output
STAFF analysis generates a complete census of microvascular velocities across entire microscope fields over periods of time extending from seconds to minutes. Representative results are presented in Figure 1, Figure 2, Figure 3, and Figure 4. Figure 1 shows an example of a time series of the microvascular network in the liver of a mouse, the generation of the skeletonized image that is used to define the axis of microvascular flow, and the STAFF-generated map of individual vascular segments identified for quantification. STAFF then uses the skeletonized image to break the microvascular network down into individual segments, then generates images of kymographs for each segment. These images are provided to the user, along with tools to identify the time intervals to be used for kymograph analysis (Figure 2). STAFF then uses the skeleton, and the user-supplied time intervals to break the kymograph of each segment into individual segment-time intervals. STAFF then identifies the predominant angle in the kymograph from each segment-time interval and provides velocity measurements as .csv data files (Figure 3) and in the form of stacks of color-coded velocity map images (Figure 4). In order to support exploration of the data analysis pipeline, STAFF also provides .csv files containing all kymograph angle measurements and goodness-of-fit values.
Figure 1: Generating the vascular skeleton. (A) Original image of a single frame from time series images collected from the liver of a living mouse. Scale bar = 100 µm. (B) Image shown in Panel A with central veins (CV) and portal triads (PT) indicated, to identify main directions of sinusoid flow. (C) Image from Panel A with overlay of TrakEM2 segmentation of sinusoids. (D) Binary image of TrakEM2 segmentation. (E) Skeletonization of TrakEM2 segmentation. (F) STAFF output image of individual vascular segments with labels. Please click here to view a larger version of this figure.
Figure 2: Kymograph analysis. (A) Magnified image of vascular segments shown in Figure 1F. (B) Typical kymograph from one segment collected over two inter-respiratory time intervals. Periods of respiration are noted with arrows. (C) Kymographs for segments 240, 252 and 254 from panel A. Please click here to view a larger version of this figure.
Figure 3: Velocity data analysis. (A, B) Tables of velocity measurements of four segments over 19 time intervals expressed either as velocity with direction (A) or as absolute velocity (B). (C) Histogram showing distribution of absolute velocity values across the entire field over the entire 20 s time period. (D) Graph of the velocities of the three segments whose kymographs are shown in Figure 2C over the entire 20 s period. Please click here to view a larger version of this figure.
Figure 4: Velocity map. (A) STAFF output image of the color coded velocity map for a single time interval for the field shown in Figure 1. (B) Composite of velocity maps for all time intervals presented as a 3D volume. Please click here to view a larger version of this figure.
There are multiple critical steps in this protocol. First, minimization of motion during intravital imaging of the liver is essential for generating movies that are usable for capillary flow analysis using STAFF. Due to the proximity of the diaphragm, short periods of respiration-induced motion occur, with the secured liver returning to its initial position after each breath. Securing the surgically exposed liver against the coverslip-bottomed dish using gauze, then imaging from below using an inverted microscope serves to immobilize the organ between respirations16,17,18,19. Second, an image acquisition speed of 100 fps is strongly recommended because the speed of flow that can be measured is a function of the speed of image acquisition and the minimum length of the segments in which flow is measured11. Third, producing a high-quality skeleton, the line drawing representation of the vascular network, is the next critical step in obtaining capillary flow velocities using STAFF. Skeleton line segments should lie near the midpoint of the vasculature between respirations over the time course being analyzed and vascular branching out of the image plane should be identified. The locations of these hidden branches along a vascular segment can be inferred by viewing the movie, by examining the kymograph or examining the velocity values over time for that segment. Viewing the movie, these vascular segments are seen either as having bidirectional flow, emanating away from or converging towards the location of the unseen branch, or as having an abrupt change in flow speed at that point along the vascular segment. The location of hidden branch points can be identified in kymographs as the location of a change in angle that is produced by the change in flow direction or velocity. In the spreadsheet, vascular segments with hidden branch locations may spuriously change direction (sign) or change between the values of the two contributing segments. If the time series includes too many branch points that are out of the image plane, STAFF analysis should be repeated after manually editing the skeleton, by placing a gap (using the paint tool) in the line drawing at the locations of the missed branch points.
We uncovered a common issue in image acquisition that typically goes unnoticed but had a strong effect on measuring flow velocities using STAFF. In image acquisition, instability or “flicker” in the intensity of the epi-illumination lamp light source or from area lighting in the microscope room can occur. From either source, light/dark banding over time with the period of the flicker occurs at zero angle in the kymographs. If the zero angle peak is the major peak, then even if the angle produced by motion of red blood corpuscles through the vessel where flow speed is being measured is obvious to the human eye, the directionality plugin fits a curve to the zero peak and reports a value very near zero degrees, resulting in velocity measurements that are unrealistically high. Even if the zero-angle peak is not the major peak, it will influence the directionality curve fit such that the velocity reported is shifted toward a higher value. To address this problem, STAFF provides a “Flicker correction” option that ignores peak angles occurring at 0°. This modification eliminated the effects of lamp flicker without affecting velocity quantifications.
The main limitation of obtaining flow velocities using widefield microscopy is that image acquisition is restricted to thin preparations such as mesenteric vasculature, or the zebrafish intersegmental vessels or the superficial layers of organs such as liver or kidney that can be exteriorized.
A significant advantage of using STAFF over existing methods of flow quantification is that it enables rapid and unbiased detection of spatial and temporal patterns of vascular flow. Collection and analysis of microvascular flow data using raster scanning systems are generally limited to measurements of single user selected capillaries at a time20,21,22,23. Methods exist to extract flow velocities across fields24,25,26, however none of these approaches support analysis across entire fields and all are labor intensive. Using STAFF, analysis of every capillary segment over time across the entire image field of a dataset with tens of thousands of images can be accomplished within a day. Manual analysis of even a single dataset of full-field vascular flow would take months to years. Thus, manual analysis is impractical even for characterization of normal flow and clearly does not permit comparison of multiple treatment groups.
The ease of STAFF quantification of spatiotemporal patterning of vascular flow provides the opportunity for future users to link vascular morphological observations to effects on capillary flow velocity patterning. By defining the relationship between the vascular morphology measures of vessel diameters and network topology to flow velocity patterning we may then be able to predict flow patterns from vascular morphology. Similarly, correlating vascular flow patterning and vascular morphology with events such as timing, localization and extent of immune cell infiltration, tissue damage from toxicant exposure or disease, or status of intracellular transport, would not only give us a better understanding of flow patterning and cellular function in health and disease, but would also provide a framework for identifying particularly harmful pathophysiological scenarios.
The authors have nothing to disclose.
Studies presented here were supported by funding from the National Institutes of Health (NIH U01 GM111243 and NIH NIDDK P30 DK079312). Intravital microscopy studies were conducted at the Indiana Center for Biological Microscopy. We thank Dr. Malgorzata Kamocka for technical assistance with microscopy.
#5 forceps | Fine Science Tools | 11251-20 | Dumont #5 Inox Forceps |
C57BL/6 mice | Jackson Labs | male 9-12 weeks old | |
Cannula | Instech | BTPE-10 | Polyethylene Tubing .011x.024in |
CMOS camera | Hamamatsu | C11440-42U30 | 4.0LT Scientific CMOS |
Coverslip-bottomed dish | Electron Microscopy Sciences | WillCo Dish glass bottom GWST5040 | |
Dissecting scissors | Fine Science Tools | ||
Fiji ImageJ Image analysis software | https://fiji.sc/ ; https://downloads.imagej.net/fiji/Life-Line/fiji-win64-20170530.zip | ||
Fluorescein dextran | Thermo Fisher, Invitrogen | D1822 | Dextran, Fluorescein, 70,000 MW, Anionic, Lysine Fixable |
Gauze sponge | Fisher | 22-415-504 | 2×2 inch Dukal sterile gauze sponges |
Heating pad | Reptitherm | RH-4 | between mouse and stage |
Heating pad | Sunbeam | 000732-500-000U | over mouse |
Inverted epifluorescence microscope | Nikon | Nikon TiE inverted microscope | |
Isis Rodent electric shaver | Braun Aesculap | GT420 | |
Isofluorane | Abbott GmbH | PZN4831850 | |
Luer stub adapter | Fisher | 14-826-19E | Catheter adapter |
Micro scissors | Castro Viejo | ||
Microscope objective | Nikon | Plan Fluor 20x, NA 0.75 water immersion | |
Needle | Fisher | 30 Ga.x1/2" | |
Needle holder | Olsen-Hegar | ||
Objective heater | BioScience Tools | MTC-HLS-025 | Temperature controller with objective heater |
Rectal thermometer | Braintree Scientific, INC | TH-5A | Mouse Body Temperature monitoring |
STAFF macros | https://github.com/icbm-iupui/STAFF | ||
Suture string | Harvard Bioscience | 723288 | silk black suture, 6-0, spool |