|Home | About | Journals | Submit | Contact Us | Français|
Intravital fluorescence microscopy has emerged as a powerful technique to visualize cellular processes in vivo. However, the size of the objective lenses has limited physical accessibility to various tissue sites in the internal organs of small animals. The use of small-diameter probes using graded-index (GRIN) lenses expands the capabilities of conventional intravital microscopes into minimally invasive internal organs imaging. In this protocol, we describe the detailed steps for the fabrication of front- and side-view GRIN probes and the integration and operation of the probes in a confocal microscope for visualizing fluorescent cells and microvasculature in various murine organs. We further present longitudinal imaging of immune cells in renal allografts and the tumor development in the colon. The fabrication and integration can be completed in 5–7 hours, and a typical in vivo imaging session takes 1–2 hours.
Intravital fluorescence microscopes have been used to study a variety of cellular-level processes in vivo, such as cell trafficking, intercellular interaction, and vascular changes. However, due to the large size of objective lenses required in these instruments, their applications have been limited mostly to superficial tissues, such as the skin, or to surgically exposed internal organs. To overcome this constraint, endomicroscopy based on miniature optical probes has been developed, providing minimally invasive access to tissues in intact organs (Flusberg et al. 2005). High quality miniature optical probes can be readily fabricated with commercially available GRIN lenses into dimensions (lengths and diameters) that are well-suited for minimally invasive high-resolution imaging in small animals (Kim et al. 2010; Fan et al. 2010). A front-view GRIN probe allows in vivo imaging of internal organs after laparotomy (Fan et al. 2010). A side-view probe is particularly suited for imaging the mucosa of tubular organs such as gastrointestinal tracts and respiratory airways (Kim et al. 2010) through natural orifices.
The advances in micro-optic and fiber-optic technologies led to the emergence of several robust design strategies for miniature optical probes and commercial instruments from, for example, Pentax (Kiesslich et al. 2007), Karl Stortz (Becker et al. 2007), Mauna Kea Technologies (Hsiung et al. 2010), and Olympus (Dela Cruz et al. 2005). The confocal endomicroscope made by Pentax and Optiscan, Ltd. offers spatial resolution similar to a GRIN probe, but the diameter of its scanning-fiber probe is relatively large (3.5–10 mm). Microprobes objectives from Olympus have narrower diameters (1.3–3.5 mm), but the objective has a fixed focal plane which restricts its use in live imaging (Dela Cruz et al. 2010). The Coloview endoscope from Karl Stortz is optimized for tracking the tumor progression in the colon on the macroscopic size scale (Waldner et al. 2011). A handheld flexible fiber-optic probe is commercialized by Mauna Kea Technologies, but its spatial resolution is typically lower than GRIN probes and is confined to the surface of the fiber bundle. In comparison, a GRIN probe provides microscopic resolution in all three dimensions, which is suitable for cellular studies. Thus it allows optical sectioning of thick tissues while possessing the mechanical rigidity adequate for the positional stability of tissue surface during imaging. Also, the view angle can be altered by attaching a prism mirror at the tip, e.g. by 90° for high-resolution side-view imaging of luminal organs. GRIN optical probes, with small diameters (0.35–1 mm), moderate lengths (20–100 mm), and high numerical aperture (0.4–0.6), present themselves as a useful tool for minimally invasive imaging of small animal models such as mice.
As in conventional confocal fluorescence microscopy, the optical penetration depth of a GRIN probe is limited to about 100 μm in most soft tissues. The penetration depth may be improved by a factor two with two-photon imaging and fluorophores emitting fluorescence in the near-infrared region. While the confocal endomicroscopy described here is well suited for imaging the outer layer of an organ, i.e. the cortex or mucosa, deeper or inner regions, such as the medulla, are not easily accessible by an optical probe without inflicting excess tissue damage. A potential solution to reach deeper tissue is to insert the probe physically into the target site (Barretto et al. 2009). With careful insertion, the collateral damage to surrounding tissue could be minimized. Another limitation stems from the mechanical rigidity of the probe, which makes it difficult to reach deep into the intestinal and respiratory tracts, such as the small intestine. This limitation may be mitigated with carefully planned surgical methods, such as a feeding-tube-like cannula (Kim et al. 2010). In addition, the field of view (FOV) of a probe is typically proportional to the diameter of the GRIN lens and is relatively small compared to a standard objective lens. Therefore, a GRIN probe is not suitable for macroscopic imaging. This limitation can be partially overcome by sequentially scanning the probe to acquire a large-area mosaic image.
In this protocol, we first present a step-by-step guide for the user who needs to fabricate an optical probe. The fabrication is relatively easy to carry out. However, for those who are less skillful in optical polishing and assembly or do not have the necessary tools for fabrication, commercial services are available, where custom designed optical probes can be purchased (e.g. GRINTECH GmbH). In this case, this protocol serves as the general guideline for custom design of GRIN optical probes.
A graded-index lens (GRIN) is a cylindrical glass rod in which the refractive index varies radially. The refractive index increases typically with a parabolic function of radius from the center to the periphery. Due to the particular index profile, a GRIN lens, similarly to a lens waveguide, is able to focus, collimate and relay optical beams. The narrow diameter, cylindrical shape and relatively high numerical aperture (NA) make GRIN lenses a fitting building block for high-resolution miniature probes. An optical probe may comprise of three GRIN lenses (Jung & Schnitzer 2003): a coupling lens (CL), a relay lens (RL), and an imaging lens (IL). Commercial GRIN RL lenses have NA of 0.1–0.2, and CL and IL lenses have NA of 0.4–0.6. All of GRIN lenses are available in small diameters (0.35–1 mm) and specific lengths. In Fig. 1, we illustrate exemplary light paths within two types of optical probes in the front- and side-view configurations, respectively. The length of the CL corresponds to a quarter pitch (P). The RL should be sufficiently long to reach the target tissue, with a length at a multiple of half pitch (Fig. 1c). However, the length should not be overly long because it can result in mechanical fragility and lead to degradation of optical performance due to excessive chromatic and spatial aberrations. For mice imaging, the appropriate length of the RL is 20–30 mm for organs in the abdominal cavity or 60–100 mm for gastrointestinal tract imaging. For the front-view probe, the IL has a pitch of 0.25, and in the case of side-view probe the IL should have a pitch of 0.16–0.17 to compensate for the increased optical path length due to the prism mirror attached. The prism has a base length of 0.7 mm (or a diagonal length of 1 mm) to fit onto the 1-mm GRIN lens. The slanted surface of the prism is coated with metal (aluminum or silver for high reflectivity) to reflect light by 90° after the IL. The CL for a side-view probe can be purchased at custom length from a lens manufacturer or, alternatively, prepared by reducing the length of a 0.25-pitch lens (by 0.8–0.9 mm) to accommodate a 0.7 mm prism.
The steps of assembling the lenses and prism can be performed in a laboratory equipped with the appropriate tools. For polishing the IL, it is critical to ensure the flatness and parallelism of the lens surfaces to minimize optical distortion. This can be achieved with a manual lapping tool with the help of a custom-made jig (Fig. 2a and Supplementary Fig. 1). For sufficient precision and ease of assembly, we use two stereoscopes arranged in orthogonal directions (Fig. 2a, b). In Fig. 2c, we provide a pictorial outline of the fabrication processes for the front-view probe, with photographs taken at three representative steps. Figure. 2c (i) illustrates the step to glue an IL to a RL by using UV curable epoxy. A small volume of UV epoxy (about 1 μl) is dispensed onto the surface of the RL prior to mounting the lens. The user may need to manipulate the 0.25-pitch lens onto the relay lens using a pair of tweezers to achieve good alignment (within <30 μm) of the lenses with visual feedback using a stereoscope. Once the IL is rested onto the RL, UV light is illuminated to cure the epoxy (Fig. 2c ii). The same steps are repeated on the other side of the relay lens to glue a 0.25-pitch coupling lens. After the completion of the assembly of the lenses, additional protection over the optical probe is necessary. We used a metal sleeve to reinforce the optical probe, which helps to avoid damage to the optical probe during handling and imaging. After the optical probe is inserted into the metal sleeve with care, high thermal-curing epoxy is applied and cured in order to attach the metal sleeve to the optical probe (Fig. 2c iii). The attachment of GRIN lenses for same front-view and side-view probe are the same (Fig. 2d i–ii). For the side view probe, an additional step is required for the attachment of the prism onto the IL using UV curing epoxy (Fig. 2d iii–vi). The fabrication is completed with the insertion of a metal sleeve (Fig. 2d v–vi). Finally, in Fig. 2e we show the completed optical probes.
The next process is to integrate the probe into a microscope system. Here we describe the steps in detail to assemble mechanical mounts for front- and side-view probes, respectively, and to connect the mount to a custom-built laser-scanning confocal microscope. Additional approaches to adapt the probes into commercial intravital microscopes are also provided.
In Fig. 3a, we show the complete mount for the front-view probe. Most components of the mount are commercially available optomechanical parts. In Fig. 3b, we show the individual components, namely the GRIN optical probe (i), probe holder (ii), XY translation stage (iii), and Z-axis translation stage (iv). Additional components are necessary to attach the probe mount (v) to a microscope. Figure 3c illustrates this process for an exemplary custom-built confocal microscope (Kim et al. 2008; Fan et al. 2010). The objective lens (vi) of the microscope is attached to an independent z translation stage (vii). The z translation stage is attached to a cage cube (viii) with cage rods, and the probe mount (v) is attached to the cage cube (viii) with cage plates (ix) and post (x). The focal plane of the probe can be remotely adjusted with the z translation stage (vii) holding the microscope objective (vi), while the probe remains fixed in position during imaging. Figure 3d shows the complete assembly, where the optical probe (i) is fully integrated into the microscope.
In Fig. 4a, we present the assembled mount for the side-view probe. A side-view mount consists of several individual parts necessary for translation and rotation control of the probe. Figure 4b shows the individual components: a timing belt (xi), rotation pulley (xii), cage plate with bearing (xiii), and rotation shaft (xiv), in addition to the optical probe (i′), probe holder (ii′), XY translation stage (iii), and Z-axis translation stage (iv). In the assembled mount, the angular position of the side-view probe is controlled with the rotation shaft (xiv). Figure 4c depicts the components used to attach the assembled side-view probe mount (v′) to the custom-built laser-scanning confocal microscope with an objective lens placed in a horizontal plane (Kim et al. 2010). The complete assembly of the side-view probe in the confocal microscope is shown in Fig. 4d (Supplementary Video 1).
Once the assembly is completed, fine positioning of the probe with respect to the objective lens is necessary. These steps are described in detail in Procedures and Supplementary Video 2. In brief, this process involves centering the probe with respect to the objective lens using the XY translational stage (Figs. 3 iii, 4 iii). The z translational stage is used to bring the proximal surface of the optical probe to a position coincident with the focal plane of the microscope objective. For the initial use of any optical probe, we recommend the user to measure its optical performance. Yhe field of view (FOV), spatial resolution and chromatic aberration can all be characterized with commercially available fluorescent beads. In Box 1, we describe this procedure in detail.
The field of view (FOV) and the spatial resolution can be measured using fluorescent samples. The chromatic aberration can be quantified using a multicolor fluorescent sample. Detailed mathematical background for obtaining spatial resolution and FOV can be found in the literature (Cole et al 2011).
Detachment of the probe from the microscope is performed in the reverse order of the integration process. Typically, the probe mount (Fig. 3a, ,4a)4a) is first removed from the microscope, the probe holder is dismounted from the probe mount, and finally the probe itself is removed. The optical probe can be cleaned and stored until required.
Integration of the optical probes into a different microscope would require a modification of the protocol. In Box 2, we illustrate alternative approaches in adapting the probe mount to a typical commercial microscope.
The integration of a GRIN optical probe into an existing microscope system can be done with minimal modification. Here we describe two integration approaches onto a standard upright microsocpe system. Briefly, the front-view probe can be held in place using a micro-V clamp mounted on an XYZ stage. On the other hand, the side-view probe requires the redirection of the excitation beam to the side (rotate by 90°) using a simple 4f lens system.
As illustrated in Box 2 Fig. a, a micro-V clamp is mounted to a three axis motorized translation stage (Sutter Instruments, MP-285) using a standard ½ inch post (Thorlabs, TR series) and ½ inch post holder(Thorlabs, PH series). A front-view optical probe is held in place with a micro V-clamp (Thorlabs, VK 250). Alternatively, a manual stage (Thorlabs, PT3A) can also be used to hold the micro V-clamp.
An additional side-view imaging path can be built to accommodate the associated relay optics. The purpose of the side-view microscope adaptor is to flip and relay the conjugate scanning plane, located close to commercial objective mounts by 90 degrees and onto the back aperture of a microscope objective (Box 2 Fig. b). A silver mirror is used to reflect light exiting the microscope mount. The conjugate scanning plane is relayed onto the back aperture of the microscope objective with a 4-f lens system. The first lens is placed at one focal length away from the scanning plane. The second lens is placed at two focal lengths away from first lens. The back aperture of the microscope objective is placed at one focal length away from second lens. A detailed layout of the mount assembly is shown in Box 2 Fig. c. The mount comprises of 30 mm cage system cubes (Thorlabs, C4W; n=2), a rotatable kinematic cage cube platform (Thorlabs, B4C), a 1″ cage cube optic mount (Thorlabs, B5C), optical component threading adapters (SM1S10 or SM1A25), a SM1-threaded 30 mm cage plate, 0.35″ thick (Thorlabs, CP02, n=4), cage assembly rods (Thorlabs, ER1, n=4; ER2, n=2), z-axis translation mount (Thorlabs, SM1Z), a 0.5″ stainless steel optical post (Thorlabs, TR2), and an optical probe holder (Fig. 3b). Box 2 Fig. d shows the complete assembly of the side-view probe in an upright microscope.
The imaging procedure consists of handling live animals and acquiring in vivo fluorescence images/videos in the organs of interest using an optical probe. Here, we briefly describe the general outline of the steps in inserting the front- and side-view probes into anesthetized mice for in vivo imaging. An anesthetized mouse is placed on the plate and its body is kept warm with a heating pad (Fig. 3d). The mouse is immobilized on the animal stage using a flexible adhesive tape. For intra-abdominal organ imaging, the front-view optical probe is inserted through a 3 to 5 mm-long incision in the skin and peritoneum (Fan et al. 2010). While monitoring the fluorescence images, the XYZ animal stage is controlled to transverse the mouse about the probe until the tissue of interest is found. The focal plane of the optical probe is adjusted by controlling the z translation stage (Fig. 3c vii). The objective focus control allows the focal plane in the tissue to be varied without moving the optical probe.
Prior to colon imaging with a side-view probe, the mouse is starved for 24 hours to empty the colon, and immediately before imaging the belly of the mouse is gently swept toward the rectum to squeeze out any stool remaining in the colon. The anesthetized mouse is placed on its back on the plate with its anus facing the optical probe (Fig. 4d) and moved slowly toward the probe by the XYZ translation stage. By orienting the view window of the optical probe toward the ventral side, the position of the tip of the probe within the colon can be observed. It is helpful to monitor the length of the optical probe inserted into the animal for position registration. A side view optical probe is rotated and translated along the mucosa using the customized probe mount (Supplementary Video 3). Once the optical probe is in place, the user may observe the images on a computer display and record images. The acquisition of the images can take part in either a single movie file or multiple still images depending on the need.
Prior to imaging, mouse should be housed in a cage with free access to food and water, temperature/humidity control, ventilation, and regular cleaning. After imaging, mouse should be kept in a warm environment and observed until it recovers from anesthesia.
Mix 1 ml of ketamine (100 mg/ml), 0.15 ml of xylazine (100 mg/ml), and 8 ml of sterile saline in aseptic condition. The mixture volume can be modified depending on the number and weight of mice. ▲ CRITICAL The ketamine/xylazine solution should be prepared and used on the same day.
Dissolve the FITC-dextran or TAMRA-dextran in sterile saline with 5% wt/vol and filter the solution through a 0.45 μm syringe filter. The solution can be aliquoted (100 μl) and stored in light-proof vials for future use.
Before starting the metal sleeve procedure, mix the thixotropic epoxy (pack A) and the high temperature resistance (pack B) on an aluminum tray. ▲ CRITICAL The mixture should be prepared and used on the same day because the curing starts even in room temperature (18–20 °C).
A stainless sleeve with an inner diameter of 1.02 mm and an outer diameter of 1.25 mm is used for metal sleeve. This is to protect the adjoining points of the optical probe during routine handling and imaging. The metal sleeve needs to be cut and trimmed to the desired length to adequately cover the adjoining points of the fabricated optical probe without exceeding its total length. Metal sleeve can be prepared with a grinding machine. Alternatively, commercial metal cutting services are also available from GRINTECH GmbH or local machine shops. ! CAUTION Care is necessary to avoid physical injury during the cutting and grinding of the metal tube.
Connect polishing disk (Thorlabs, cat. no. D50-SMA) with SMA connector which has hole diameter of 1.03 mm (Thorlabs, cat. no. 11040A) after removing plastic part of SMA connector (Supplementary Fig. 1).
Prior to imaging, sterilize optical probes with 100% ethanol solution.
A custom-made stainless 25.4 mm diameter disk with a 1.26 mm diameter hole in the center is used for holding the optical probe (Supplementary Fig. 2).
A custom-made stainless tube (4.28 mm OD, 1.2 mm ID) used for holding the optical probe is inserted into a ball bearing that is already glued to a custom-made stainless bearing collar. A rotation pulley is attached to the tube using a set screw. (see Supplementary Fig. 3).
▲ CRITICAL Due to its small features, the optical probe must be handled with care during insertion into the holder via the thru-hole. The outer diameter of the optical probe holder is matched to the internal thread of the XY translation stage (Fig. 3b iii).
An XY translation stage is mounted onto a single axis translation stage (Newport, Gothic-Arch X) using two assembly cage rods (Thorlabs, ER2) and a cage adaptor mount (Thorlabs, CP02B) shown in Fig. 3a.
A probe mount is designed to provide rotation as well as translation of a side-view optical probe as shown in Fig. 4a. The detailed procedure for assembling the attachment mount is illustrated in Supplementary Fig. 4. Briefly, a probe holder comprised of a rotation pulley that is attached to an XY translation cage mount. A rotation shaft is connected with a timing belt, a cage mounting adaptor (Thorlabs, CP02B) and cage assembly rod (Thorlabs, ER1) to the probe holder and XY translation cage mount. The assembled mount is attached to a single-axis translation stage (Newport, Gothic-Arch X) using two cage rods (Thorlabs, ER2) and a cage adaptor mount (Thorlabs, CP02B) (Fig. 4b, c).
The microscope objective and the probe mount are attached to a cage cube (Thorlabs, C4W) (Fig. 3c iii, 4c iii). The microscope objective has an independent focus control, z translation stage, (Thorlabs, SM1Z) that is used to manually shift the focal plane of the objective.
The imaging system is a custom-built video-rate scanning laser confocal microscope (Kim et al. 2008). The system uses continuous-wave lasers at 491, 532, and 635 nm (Cobalt Laser) for single-photon fluorescence excitation. Silver mirrors and dichroic splitters (Edmund Optics, Chroma Technology) were used to select and direct a laser beam to a raster beam scanner comprising a silver-coated polygon scanner (Lincoln Laser) and a galvanometer (Cambridge Technology). The microscope was designed to have a field of view of 250–300 μm when a 40X objective lens (Olympus, LUCPlanF1) is used. Three photomultiplier tubes (Hamamatsu, R3896) were used to collect fluorescence signals in three fluorescence channels, respectively. An 8-bit four-channel frame grabber (Matrox Solios eA/XA) digitizes the PMT signals at 20_MS/s for 512×512 pixels/frame. Customized software written in C++ is used to display and save the acquired images at a frame rate of 15–30 Hz in real time.
Troubleshooting advice can be found in Table 1.
We show fluorescence images taken using the front-view optical probe at different sites in various murine organs in vivo. In Fig. 5a–f, we used a transgenic mouse, in which green fluorescent protein (GFP) is expressed under the direction of the major histocompatibility complex (MHC) class II promoter. MHC-II+ GFP+ antigen presenting cells are observed in green, and the vasculature is visualized by intravenously injected dye-dextran conjugates. In Fig. 5g, h, a Tie2-GFP transgenic mouse was imaged, where the endothelial cells of blood vessels are visualized by GFP fluorescence. With the front-view probe, we also performed repeated imaging of immune cells in a mouse model of renal transplantation (Miyajima et al. 2011). A donor kidney was harvested from a MHC class II-GFP mouse in C57BL/6 background strain and then transplanted into a wild-type mouse in BALB/C background strain. The front-view optical probe and the imaging procedures described earlier were used to monitor the donor GFP-expressing antigen-presentation cells in the kidney allograft. The tissues in the kidney expressed relatively strong auto fluorescence, which visualizes kidney tubules in the cortex. At Day 1 after transplantation, a significant number of dendritic shaped GFP+ donor APCs are observed (Fig. 6a). At Day 3, most dendritic GFP cells have entered the capsules, which indicates the process of clearance (Fig. 6b). Time-lapse imaging can allow the migration of immune cells to be tracked in the kidney (Supplementary Fig. 5).
In Fig. 7a-c, we show fluorescence images taken using the side-view optical probe in the normal mucosa of the descending colon, which show MHC-II+ GFP+ cells, epithelial cells lining the crypt structure, and the blood vasculature visualized by FITC-dextran in the blood stream. A three-dimensional dataset of the vasculature was acquired by scanning the probe over a wide area of the tissue at a speed of 100–200 μm/sec. Rotation and registration of individual images allow a large-area mosaic image to be constructed (Kim et al. 2010; Fig. 7d). The dataset can be rendered in various modes, such as a fly-through presentation (Fig. 7e). Using a similar procedure, vasculature and cellular images can be obtained from the upper gastrointestinal tracts and respiratory tracts (Kim et al. 2010).
The side-view optical probe is well suited to track the development of colorectal tumors in a spontaneous colorectal cancer mouse model. We used a genetically modified mouse model, in which the adenomatous polyposis coli (Apc) gene in the intestinal cells in the colonic mucosa can be knocked out by administration of adenoviral Cre. The inactivation of Apc is followed by the activation of reporter gene of GFP to enable the long-term tracking and quantifying the growth of Apc-mutated cells in the colon by side-view fluorescence endomicroscopy (Kim et al. 2010). Using the side-view optical probe and the imaging procedures described earlier, we observed multiple GFP expressing polyps at 4 weeks after the viral delivery of Cre (Fig. 8a). Blood vessels are simultaneously visualized by intravenously injected tetramethylrhodamine (TAMRA)-Dextran. At this stage, two micro-nodules with diameters of 50–150 μm are detected, where noticeable changes in vasculature are observed. Two weeks later, the same area can be identified at approximately the same distance from anus, and it is found that one of the nodules has grown to approximately 400 μm in diameter (Fig. 8b). Dilation of blood vessels surrounding the nodules is clearly observed. The injected dye can leak out of the permeable vessels in tumors, which makes it difficult to image the vasculature repeatedly at a short time interval.
Conjugate plane:One spatial point is relayed onto a second conjugate/reciprocal spatial point through a set of optical lenses. For optimal laser beam steering, a 4f conjugate lens system is used to relay the pivot point on the beam steering mirrors onto the center of back pupil plane of the microscope objective lens.
Field of view (FOV) refers the extent of observable area through the optical system.
Focal plane indicates the plane, lying on the focal length away from the center of an optical arrangement, in which an image is formed for light focused from infinity.
Numerical aperture (NA) is a dimensionless number that indicates the resolving power of a microscope objective. The number is obtained by multiplying the index of refraction of the medium in front of the microscope objective and sine of half-angle over which the system can accept or emit light.
Optical aberration is a distortion in an optical system that reduces the optical performance as compared to an ideal system. They can arise due to the misalignment or physical imperfection of optical components such as lenses and mirrors. Chromatic aberration refers to a shift of focal plane due to a change in wavelength.
Pitch is defined as the length of the GRIN lens within which a single ray will propagate a full sinusoidal cycle.
Spatial resolution is defined as the shortest distance between two spatially separated points on a specimen that can be resolved by an imaging system as separate entities.
Temporal resolution is defined as the shortest time difference between two separate points captured in succession.
The authors thank Robert Colvin and Cuffy Chase for providing the mice with renal transplantation, Raju Kucherlapati and Ken Hung for the mouse model of spontaneous colorectal tumor, Charles Lin, Daniel Cote, Lee Kaplan, Mauro Ferrari, and Gou Young Koh for discussions, and David Stevenson for proof-reading of the manuscript. This work was supported by grants from National Institutes of Health (R21AI081010, RC1DK086242, RC2DK088661, U54CA143837, R01AI081734), Department of Defense (FA9550-10-1-0537), and National Research Foundation of Korea (R31-2008-000-10071-0, 2009-352-C00042, 2011-0009503).
AUTHOR CONTRIBUTIONSJKK, WML, PK, MC, SHY designed and wrote the protocol. KJ and SK contributed figures.
COMPETING FINANTIAL INTERESTS
The authors declare no competing financial interests.