The structure and mechanical properties of collecting lymphatic vessels: an investigation using multimodal nonlinear microscopy

Abstract

This study employed nonlinear microscopy on fresh, unstained and unfixed collecting lymphatic vessels to determine the wall structure and its relationships to the mechanical properties of the tissue. Fresh bovine mesenteric collecting lymphatic vessels were mounted in a vessel bath and imaged under different luminal pressures (0–30 cmH₂O pressure head), and longitudinal tensions. The entire wall thickness was imaged, using two-photon fluorescence to visualize elastin, second harmonic generation to image the collagen, and coherent anti-Stokes Raman scattering to image the cell membrane. The adventitial fat cells were coupled to the wall within the elastin-rich network of fibres. The medial smooth muscle cells were too densely packed to resolve the boundaries of individual cells in en face images, but in tissue sections their appearance was consistent with electron microscopic data. Two distinct populations of collagen fibre were revealed. Large fibre (15–25 μm diameter) bundles were present in the inner media and small fibres (2–5 μm diameter) were distributed throughout the wall. The responses to longitudinal tension and luminal pressure indicated that the larger fibres resist the longitudinal strain and the smaller oppose pressure forces. Individual elastin fibres were of uniform thickness (1–3 μm) and interwove amongst themselves and between the collagen fibres. The network was probably too sparse directly to support mechanical loads and we speculate that its main function is to maintain the organization of collagen bundles during recovery from contraction.

Introduction

The lymphatic system is an important, but frequently overlooked, component of the circulation. In man it drains up to 4 L of plasma filtrate from the tissues per day and returns it to the vascular system. Key to this process are the collecting lymphatics, which are filled by the lymphatic microvessels and which, by means of a series of contractile elements separated by one-way valves, are able to pump the lymph, often against a hydrostatic pressure gradient. Failure of this process, as a result of disease or surgery, has dramatic and severe consequences in the generation of massive tissue oedema (Levick & McHale, 2003). Lymphatic pumping has attracted attention from clinicians, physiologists and, more recently, biomechanicists (Ohhashi et al. 1980; McHale & Meharg, 1992; Quick et al. 2007; Venugopal et al. 2007; MacDonald et al. 2008). However, much still remains to be done, and one area in which data are surprisingly lacking is in understanding of the mechanical properties of the vessel wall itself. This information is urgently required, both for exploration of regional variations in the anatomy of the pumping activity of the lymphatic network and for the development of accurate theoretical models.

There appear to have been very few measurements of the mechanical properties of lymphatic vessels. The first measurements were made by Ohhashi et al. (1980) in bovine mesenteric vessels under different conditions of smooth muscle activation, and were followed by a more quantitative estimation of the elastic modulus by Deng et al. (1999) in the canine thoracic duct. The latter study drew attention to the nonlinearity of the response above 5 cm water pressure head. A recent study from our laboratory found considerable variability in elastic modulus between bovine mesenteric vessels and also estimated the viscoelastic components of the response and longitudinal tension in the vessels (MacDonald et al. 2008). In contrast to these scant data, the mechanical properties of arteries and veins have been very extensively studied (Bergel, 1961; Dobrin & Rovick, 1969; Dobrin & Canfield, 1977) and significant progress has been made in relating differences in mechanical properties to variations in hemodynamic conditions prevailing at different points in the vascular tree. There is also a complementary body of literature relating mechanics to the detailed architecture of the vascular wall (Dobrin, 1978; Zoumi et al. 2004; Megens et al. 2007; Wang et al. 2008). A primary aim of the present study is to establish a basis for similar analyses of the lymphatic system by determining the structure and organization of cells and extracellular matrix in collecting lymphatics and the structural responses to mechanical loads.

The collecting lymphatics contain varying proportions of smooth muscle and an extracellular matrix containing fine bundles of collagen and elastin fibres. There is also a fat-rich adventitia incorporating a microcirculation. At least in the case of the thoracic duct, considerable variation of wall structure is reported (Gnepp, 1984). However, the available information is inadequate to establish the structural bases of the passive mechanical properties of the vessels and of how the muscle cells exert their mechanical effects. This situation may, in part, arise from methodological constraints. It is a laborious task to obtain the required three-dimensional information on cell and matrix organization by conventional histology and in the present study we have employed nonlinear microscopy to provide three-dimensional images of intact vessels for this purpose. In addition to the quality of images obtained, a great advantage of this technique is that it can be applied to intact, fresh, unstained tissue, avoiding the artefacts of histological processing and allowing the responses to mechanical strains to be imaged.

Non-linear microscopy is an excellent technique for three-dimensional, non-invasive imaging of biological structures; the nonlinear signal generation confines optical excitation to a focus where the photon flux is highest, thus providing intrinsic optical sectioning. The use of infrared excitation affords an increased depth penetration over conventional optical microscopy, removing the need for sample sectioning (Hsu et al. 2002; So, 2003). Moreover, contrast is derived from intrinsic sample properties, thus avoiding the need for histological or immunohistochemical stains. Recent applications include skin (Palero et al. 2007), arterial plaque (Megens et al. 2007), fish gills (Moger et al. 2008) and articular cartilage (Mansfield et al. 2008).

In the present study we employ a multimodal non-linear imaging system which generates contrast from the three principle tissue components believed to be responsible for the mechanical properties of the lymph vessel wall; collagen, elastin and smooth muscles. Elastin-containing networks are visualized by two-photon excitation of their autofluorescence (TPF) (So et al. 1998) and collagen fibres are identified by their strong second harmonic generation (SHG) signal (Campagnola et al. 2002). The cellular structure was identified using coherent anti-Stokes Raman scattering (CARS) microscopy. CARS microscopy derives its contrast from intrinsic molecular vibrations in a sample (for reviews see Cheng & Xie, 2004; Potma & Xie, 2005; Rodriguez et al. 2006). A pump beam of frequency ω_p and a Stokes beam, ω_s, interact with the sample via a four-wave mixing process. When the beat frequency (ω_p − ω_s) is tuned to match a Raman active vibrational mode, molecules are coherently driven with the excitation fields, resulting in the generation of a strong anti-Stokes signal. In this study CARS was employed to visualize the plasma membrane by exciting the Raman modes of CH₂ groups in membrane phospholipids (Cheng et al. 2002).

The simultaneous implementation of the three non-linear contrast mechanisms allows detailed characterization of the interrelationships of the two fibre networks of the extracellular matrix and of cells and matrix. Because they require no labelling or tissue processing, no distortion of the cellular envelope occurs, alleviating a problem of conventional microscopy.

Methods

Tissue preparation

Collecting lymphatics were excised from bovine mesentery at an abattoir (Stillman’s, Taunton, UK) immediately after death, before the fatty tissue solidified. These vessels are easily distinguishable by eye from other veins and arteries. The vessels were placed in normal Krebs solution (with 95% O₂, 5% CO₂ gas mix, BOC) and kept at 4 °C until warmed for use at 37 °C.

The seven vessels successfully studied were cannulated at both ends and immersed in a purpose-built bath filled with Krebs solution and connected at both ends to pressure reservoir heads containing Krebs solution to adjust transmural pressure. The bath, made of Teflon, with a coverslip window in the base was mounted directly on the stage of the inverted microscope (Fig. 1). The mounting cannulae were carried by micrometer screws (Orlin Tech Ltd, UK) which allowed the vessel to be placed close to the window and the longitudinal tension in the vessel to be adjusted.

Fig. 1.

Apparatus. Top: Schematic diagram of the nonlinear microscope. Below: Lymphatic vessel perfusion rig.

Four vessels were fixed in 4% formal saline for 24 h at 4 °C and sectioned into annular rings using a cryostat microtome, using saline as the embedding medium to avoid background multiphoton signals from medium transferred to the slide. Two sections were stained for smooth muscle actin (ABCAM, UK) (McCloskey et al. 2002).

Microscopy

Two-photon autofluorescence and second harmonic generation were excited using a mode-locked femtosecond Ti:sapphire oscillator (Mira 900D; Coherent, USA) which generated 100-fs pulses at 76 MHz with a central wavelength of 800 nm. The average power at the sample was attenuated to between 5 and 30 mW. The dual wavelength excitation required for CARS microscopy was generated using an optical parametric oscillator (OPO) (Levante Emerald, APE, Germany) pumped with a frequency doubled Nd:Vandium picosecond oscillator (High-Q Laser Production GmbH, Germany). The pump laser generated a 6-ps, 76 MHz pulse train at 532 nm with adjustable output power up to 10 W. The OPO produced collinear signal and idler beams with perfect temporal overlap of the pulse trains and provided continuous tuning over a range of wavelengths. The signal beam was tuned to 924 nm and the idler formed the Stokes beam at 1255 nm. This produced ω_p − ω_s of 2848 cm⁻¹, corresponding to the CH₂ vibrational mode of membrane phospholipids. Negative lipid contrast was used to improve contrast of lipid components, using pump and Stokes wavelengths of 919 nm and 1263 nm, respectively, which corresponds to the ‘dip’ in the CARS signal around 2965 cm⁻¹, which arises from destructive interference between the resonant lipid signal and the non-resonant electronic background (Cheng et al. 2003). The maximum combined output power of the signal and idler was approximately 2 W; however, the average power at the sample was restricted to no more than 100 mW.

Imaging was performed using a modified commercial inverted microscope and confocal laser scanner (IX71 and FV300; Olympus UK, UK). A schematic of the optical setup is shown in Fig. 1. To maximize the near infra-red throughput, the standard galvanometer scanning mirrors were replaced with silver galvanometric mirrors and the tube lens was replaced with a MgF₂ coated lens. A 60×, 1.2 NA water immersion objective (UPlanS Apo, Olympus UK) was used to focus the laser excitation into the sample. The nonlinear intensity-dependence of multiphoton excitation confines the signal to the focus where the photon flux is highest, obviating the need for confocal detection. The scanning confocal dichroic was replaced by a silver mirror with high reflectivity throughout the visible and near infra-red (21010; Chroma Technologies, USA). Laser excitation was directed into the scan unit and femtosecond and picosecond beams selected by a flip mirror.

Second harmonic generation was detected in the forwards direction, collected by an air condenser (NA = 0.55) and directed via collimating lenses onto a dichroic mirror, which removed the fundamental wavelength and separated the SHG from the CARS signal. The SHG signal was then isolated by a band pass filter (F10-400-5-QBL; CVI Melles Griot UK, UK) and focused onto a red-sensitive photomultiplier tube (R3896; Hamamatsu Photonic UK, UK). TPF was measured in the epi-direction, after spectral separation from the excitation wavelength by a dichroic mirror (670dcxr; Chroma Technologies) and bandpass filters (CG-BG-39-1.00-1 and F70-500-3-PFU; CVI Melles Griot UK), using a photomultiplier tube (R3896; Hamamatsu Photonic UK) mounted at the back port of the microscope. The forward-CARS signal was collected by the air condenser, transmitted by the dichroic mirror and directed onto a red-sensitive photomultiplier tube (R3896; Hamamatsu Photonic UK). The epi-CARS signal was collected (data not shown in this report) using the objective lens and separated from the pump and Stokes beams by a long-wave pass dichroic mirror (z850rdc-xr; Chroma Technologies) and directed onto a second R3896 photomultiplier tube at the rear microscope port. The anti-Stokes signal was isolated at each photodetector by a single band-pass filter centered at 750 nm (HQ750/210; Chroma Technologies).

Three-dimensional images were constructed from stacks of two-dimensional scans in the x–y plane, increments in the z-direction being achieved by alteration of the objective focus.

Visualization of the three-dimensional datasets was performed using ImageJ (Rasband, 1997–2009) and velocity 3.6.1 (Improvision, UK). Lipid contrast was enhanced by subtracting the 2965 cm⁻¹ CARS images from the corresponding 2848 cm⁻¹ CARS images. The SHG, TPF and CARS images were overlaid and colour-coded in blue, green and red, respectively.

For quantitative analysis of the fibre networks, macros were written within ImageJ utilizing the oval profile plugin (O’Connell, 2002–2009). The background was subtracted using a rolling ball technique with a 7-μm radius ball for elastin fibres, and a 20-μm radius ball for the collagen. Auto-correlation of the individual two-dimensional images in the stacks was performed to find the predominant fibre directions. Line profiles were taken perpendicular to the fibre directions to detect periodicity.

Results

The walls of vessels approximately 2 mm in diameter were sufficiently thin (approximately 30 μm) that z-scan along a radius from the outer surface of the vessel encompassed the whole thickness of the wall. SHG images of the collagen fibres acquired in the forward- and back-scattered configurations revealed essentially similar structures. There were small differences attributable to local variations in the scattering and attenuation properties of the tissue, but these were insignificant to the structures revealed and we present only results from forward scattering. The images showed excellent signal-to-noise and the only caveat to be borne in mind in viewing the images presented below is that collagen fibres running in the z-direction give no signal. This can result in a false impression of discontinuities in collagen networks. Back-scattered TPF images gave excellent delineation of elastin fibres, but autofluorescence spill-over gave rise also to a signal from collagen fibres. However, this was much weaker and caused no problems in interpretation; it was, in fact, used to advantage in ensuring the co-registration of images.

In total, seven vessels were studied. At relaxed length and zero transmural pressure, two collagen fibre networks were evident (Fig. 2). One consisted of fine fibre bundles (typically 2–5 μm in diameter), which were so densely packed as to be barely resolvable individually. They presented a wavy appearance along the vessel in a single cross-section (Fig. 2A), but three-dimensional reconstructions revealed that they actually spiralled around the circumference of the vessels with a pitch of 30 μm (Fig. 3A–C). These structures occurred predominantly towards the adventitial surface of the vessel. Bundles of thicker fibres (15–25 μm in diameter) were found primarily closer to the luminal surface (Fig. 2E–H). These fibres were interwoven and also spiralled along the long-axis of the vessel.

Fig. 2.

Second harmonic generation images of collagen in intact bovine collecting vessel wall at different luminal pressures and longitudinal tensions: tangential images with the long axis of the vessel running from left to right, field of view 240 × 240 μm. The top images show fine collagen fibres in the outer wall and the lower panel shows the coarser fibres in the inner wall of the same vessel. (A,E) Relaxed length, zero pressure. (B,F) In situ length, zero pressure. (C,G) In situ length 15 cmH₂O luminal pressure. (D) In situ length, 30 cmH₂O luminal pressure.

Fig. 3.

Three-dimensional reconstructions of TPF image of elastin fibres in relaxed vessel. The collagen autofluorescence has been enhanced to demonstrate the relationship of elastin to the large spiral collagen fibre bundles. (A–C) Relaxed vessel, zero pressure. (D–F) In situ strain, 15 cmH₂O luminal pressure. 25-μm-grid squares.

Quantitative analysis of fibre orientation that showed that the principal direction of these larger fibre bundles rotated through 45° with depth through the vessel (Fig. 5A), confirming the woven spiral structure. Auto-correlation analysis revealed that the larger collagen fibres bundles had a spacing of approximately 35 μm spacing with a 40% standard deviation, indicating a lack of long-range ordering.

Fig. 5.

Principal directions of collagen (□) and elastin (○) fibres as a function of luminal pressure at different depths through the vessel wall. 180° represents the direction of the longitudinal axis of the vessel.

Elastin fibres were present throughout the tissue but were more numerous near the luminal side of the wall. The fibres were of quite uniform diameter (approximately 1 μm) and generally occurred singly. Some fibres branched into two fibres of equal size and there were also junctions with four branches which appeared to involve mechanical connections rather than crossing of fibres (Fig. 3). In the relaxed tissue the fibres were aligned primarily parallel to the long axis of the vessel. The majority of the fibres were strikingly straight (Fig. 4). Single elastin fibres of similar aspect ratio isolated from nuchal ligament have a low bending modulus and thus in a relaxed state follow a tortuous path (unpublished observations). It therefore appears probable that even in the relaxed vessel the elastin fibres are under tension. Autocorrelation analysis gave a mean fibre spacing of 40 μm, which in every sample was larger than that of collagen bundles, although again the standard deviation was, at 40%, indicative of an absence of long-range order.

Fig. 4.

TPF images of an intact vessel wall at different luminal pressures and longitudinal tensions: Image orientation, conditions and magnification as in Fig. 2A–D. No background subtraction has been performed, enabling thin, intensely fluorescent elastin fibres to be visualized against a background of broad collagen fibre bundles.

Longitudinal strain and luminal pressure both affected fibre orientations. When vessels were returned to their length before excision from the mesentery (approximately 100% greater than the relaxed length) but luminal pressure remaining zero, the elastin fibres re-orientated slightly in the direction of strain and the collagen fibres did so to a greater extent (Fig. 2B,F). The quantitative analysis shown in Fig. 5B is, for collagen, dominated by the large fibre population: the small fibres retained their spiral configuration.

When the luminal pressure in the extended vessels was increased through the physiological range the total wall thickness halved (Fig. 3). At a pressure of 15 cmH₂O, both the collagen and elastin fibres re-orientated away from the vessel axis to form two populations of fibres at approximately ± 40° to the axis, the angle of the collagen fibres being slightly greater compared with the elastin (Fig. 5C). Increasing the pressure further to 30 cmH₂O, caused little further increase in angle. Three-dimensional images of the pressurized vessel clearly revealed elastin fibres weaving between the collagen (Fig. 3). When a longitudinally relaxed vessel was held at 30 cmH₂O pressure a proportion of the collagen and elastin fibres turned further from the vessel axis to assume an almost circumferential orientation.

Cellular material, probably plasma membranes (Mansfield et al. 2009), were readily identified in the CARS images (Fig. 6). Where they had survived the dissection procedure, adventitial fat cells gave a particularly strong signal (Fig. 6A). On their luminal surface they were closely associated with the outer collagen and elastin networks, though it was not possible to establish whether the cells were simply entangled in the matrix or formed point attachments to it. There was also a strong signal from the intact monolayer of endothelial cells. These cells were attached to an elastin-rich matrix, whose fibres ran predominantly along the vessel axis. A second similar layer overlaid the cell-rich region in the middle of the wall. The boundaries of individual muscle cells were difficult to resolve because of the close apposition of neighbouring cells. Only occasional elastin fibres were observed, following a tortuous path between the cells. In relaxed tissue the cells were spindle-shaped and randomly orientated, as confirmed by smooth muscle actin staining of tissue rings and previously reported from electron microscopy (McCloskey et al. 2002). As the vessel was pressurized, no changes in the orientations of individual cells could be resolved but the thickness of the muscle cell layer was approximately in proportion to the change in total thickness. The adventitial fat cells remained unaffected.

Fig. 6.

Cells of the vessel wall (CARS, red) with superimposed images of elastin TPF (green) and collagen (SHG, blue). (A) Adventitial fat cells embedded in the outer layers of elastin and collagen. The vessel axis runs from right to left (vessel fixed to better preserve the integrity of the fatty tissue). (B) Luminal surface, showing endothelial layer. Fresh tissue, orientation as in A. (C) Transverse section of the wall, the lumen is on the right.

Discussion

There have been several TPF and SHG studies in blood vessels (Zoumi et al. 2004; Megens et al. 2007). Megens et al. (2007) imaged an intact blood vessel but used conventional staining for the elastin. Boulesteix et al. (2006) used TPF and Wang et al. (2008) used CARS in combination with TPF to image transverse tissue sections. Our study combines these imaging techniques and we believe it to be the first on lymphatic vessels.

The CARS images themselves are difficult to interpret. In light of previous work on articular cartilage (Mansfield et al. 2009), the likely candidate for the majority of the CH signal, apart from the collagen and elastin fibres, is the lipid from the plasma membranes. It should be noted, however, that there would be contributions from the membrane and cortical cytoskeleton protein CH groups, which we might ultimately be able to distinguish spectroscopically with developing techniques such as stimulated Raman spectroscopy. However, this distinction does not detract from the current imaging work.

The lymphatic vessel proved a difficult tissue to study in comparison with blood vessels because of the density and complexity of the cellular organization and the rich structure of the extracellular matrix. Future refinement of the optics, for example adding in adaptive optics (Wright et al. 2007) to minimize the scattering effects of depth, and to optimize the conflicting requirements of the three modalities, together with better methods of background elimination or improved spectral resolution in the CARS images should improve image quality in the future. The analysis of the data stacks was not simple. The spiral pattern of the collagen caused particular problems due to loss of signal from fibres orientated perpendicular to the beam direction. The 3-D structure of the elastin network was also difficult to analyse using the 2-D auto-correlation technique due to the short correlation length in the x–y direction in the relaxed samples. The curvature and size of the vessel made it non trivial to align the individual two-dimensional images with equal radius, especially with the unpressurized samples.

There were also some physiological limitations to the present study. One source of uncertainty was in the determination of the in situ length of the vessel. The need to excise the vessel before the peripheral fat solidified, limited the accuracy of the initial length measurement. Another source of uncertainty was the state of smooth muscle tone. Experience suggests that the vessels were in a relaxed state, but a detailed analysis of the structural effects of changing tone would be a valuable extension of the study. Our dissection technique preserved the integrity of the endothelium, although intracellular structures could not be resolved through the thickness of the underlying tissue. The adventitial fat cells were clearly visualized. There was evidence that they were coupled to the vessel and the surrounding tissue through an elastin-rich fibre network, but this was easily disrupted during tissue dissection and an alternative preparation would be required for a more detailed examination of its structure and mechanical properties. This layer warrants further consideration, as it may be important in mechanically coupling the vessel to the supporting tissue.

The medial structure of the collecting lymphatic differs in a number of ways from that of blood vessels. Neither the two collagen fibre networks nor the elastin fibre networks consisting of individual fibres appear to have analogues in the vascular system. Blood vessels are exposed mainly to radial variations in pressure and the principal objective of smooth muscle is to effect changes in radius, and the structural adaptations to these requirements have been extensively discussed (e.g. Dobrin, 1978). The function of the wall of the collecting lymphatic is primarily to propagate a contraction wave to generate flow. This clearly generates radial, longitudinal and perhaps torsional stresses and the ways in which these are accommodated by the organization of cells and matrix are still poorly understood. Data such as those reported here, perhaps extended to include gated image acquisition during contraction, should provide a basis for structurally based modelling of wall mechanics.

The available data on the macroscopic mechanical properties of the lymphatic wall relate predominantly to the quasi-static stress–strain characteristics. These are highly non-linear and it has been speculated that the initial compliance reflects the contribution from the elastin network and the stiffening arises from the transfer of loads to collagen at larger strains (MacDonald et al. 2008). The present results refine this interpretation. Physiological longitudinal strains are borne by mainly thick collagen bundles to impede stretch. Luminal pressurization initially causes reorientation of both collagen fibres to support radial and tangential stresses. At physiological pressure the network is relatively isotropic. Further increase in pressure, which produces only a small increase in strain, causes little further rearrangement of the matrix. The smaller collagen bundles do not become straight under these measured pressures. They may be involved in localized radial deformations of the matrix during pumping. The elastin network is sparse, though more evenly spread through the depth than in blood vessels, and its functions may relate more to maintaining the organization of the collagen networks than to supporting stresses directly. This view is consistent both with the inter-relationships of the two networks, and with the wide variations in strain we inferred in different elements of the elastin network. At any particular strain, some elastin fibres were straight, as if held in tension, whilst others followed highly tortuous paths. Often fibres of these two types came together at a single point, suggesting that points are nodes of a network rather than overlaying fibres. These details will have to be considered in structurally based models.