| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Department of Molecular Pharmacology, Physiology and Biotechnology, Center for Biomedical Engineering, Brown University, Providence, Rhode Island, USA
1Correspondence: Brown University, G-B 393, Biomed Center, 171 Meeting St., Providence, RI 02912, USA. E-mail: jeffrey_morgan{at}brown.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Key Words: aggregation • spheroid
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
2
3)
. Several techniques rely on self-assembly, a natural phenomenon whereby mono-dispersed cells will spontaneously form a spheroid structure. It is thought that self-assembly mimics natural processes that occur during embryogenesis, morphogenesis, and organogenesis (4
5
6)
. Numerous studies have established that cells differ in their ability to self-assemble (4
, 7)
, that self-assembly is driven by cell-cell adhesion or surface tension, that intercellular adhesion is dependent on cadherin levels (5
, 8)
, that mixtures of cells will self-segregate as immiscible fluids according to the strengths of the homo- and heterotypic adhesiveness of the cell types, and that the spheroid is the energy- and surface area-minimized terminal structure (9
, 10)
.
A variety of methods are used for the self-assembly of spheroids from mono-dispersed cells, including hanging drop (11)
, spinner culture (12
, 13)
, agitation in micromilled recesses (14)
, and pelleted suspensions of cells (15)
, and a wide variety of cell types has been shown to self-assemble as spheroids. Examples of self-assembled spheroids include those of cuboidal hepatocytes with bile canaliculi (16)
, beating cardiomyocytes (2)
, vascularized smooth muscle cells and fibroblasts (15)
, and fibroblasts and dorsal root ganglion cells penetrated by axon-like projections (17)
. Moreover, the spheroid microtissue has been proposed as a building block for construction of even larger in vitro engineered tissues (18)
. Spheroids have been assembled into a millimeter-sized macrotissue (19)
and have been embedded into biocompatible gels, where they fuse to form a metastable millimeter-sized toroid structure (20)
.
Using micromolded nonadhesive agarose gels recently developed by our lab (21)
, we observed via horizontal microscopy that H35 spheroids sag more than do NHF spheroids. We then asked whether, in part because of this sagging, NHFs and H35s differed in their abilities to undergo what we term "directed self-assembly," the guidance of mono-dispersed cells by micromolded features in nonadhesive hydrogels, to form structures with geometries more complex than a spheroid. NHFs and H35s were tested for their ability to self-assemble rods, tori, and honeycombs. We investigated the stability and morphological changes of these structures over time and found significant differences between cell types. H35s readily formed stable rod structures whereas NHFs contracted to spheroids. H35s and NHFs formed tori but differed significantly in their stability, with H35s forming the most stable tori. Last, stable honeycomb structures with patent lumens could be self-assembled by H35s but not by NHFs. The kinetics of directed self-assembly revealed significant differences between the two cell types. Self-segregation of cells was evident in structures generated by a mixture of H35s and NHFs, and this segregation was influenced by structure geometry.
These results demonstrate that significant differences exist between cell types in their self-assembly properties and that these properties can be measured and quantified using micromolds of nonadhesive agarose. These data also show that the geometry of self-assembled structures is not limited to the spheroid, but that complex-shaped and even branched tissue constructs are feasible end products of directed self-assembly of mono-dispersed cells.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Fig. 4
, which used 1600 µm-long troughs. Wells were 800 µm deep with a hemispherical bottom and a 200 µm radius. Each gel had a staggered array of 822 wells.
|
|
|
|
Agarose gels were cast directly from wax molds. Aliquots (3 g) of Ultrapure© Agarose (Invitrogen, Carlsbad, CA, USA) were autoclaved as a powder, then 100 ml of sterile dH2O was added and the agarose was dissolved by heating and mixing on a hot plate. The solution was cooled, then 2.75 ml was pipetted into each wax mold in a sterile dish and degassed briefly to remove air bubbles. After setting, gels were separated from the mold using a spatula, transferred to a 6-well plate, and each gel was equilibrated overnight with 6 ml of tissue culture medium.
Horizontal-view gels were 13% polyacrylamide; 400 µl of polymer solution was added to wax molds. After 10 min, gels were removed and equilibrated as above.
Cell culture and production of self-assembled structures
Unless stated otherwise, all chemicals were purchased from Sigma-Aldrich (St. Louis, MO, USA). Normal human fibroblasts (NHF) derived from neonatal foreskins and rat hepatoma cells (H35) (generously provided by Martin Yarmush, Massachusetts General Hospital, Boston, MA, USA) were expanded in DMEM supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin solution. NHFs were incubated in a 10% CO2 atmosphere; H35s and a 1:1 mix were incubated in 5% CO2. Cells used in experiments were between passage 3 and passage 9.
Seeding of the gels has been described elsewhere (21)
. Briefly, excess medium was aspirated from the wells and 200 µl of a single-cell suspension was pipetted into the rectangular recess of each gel. Medium was added to the wells after a 1.5 h incubation period wherein cells settled into the recesses and began directed self-assembly. Medium was exchanged every other day.
Microscopy, image and data analysis
Bright-field and fluorescent images were obtained using an Olympus IX70 microscope equipped with an AxioCam MRc digital camera. Time-lapse images were obtained with a Nikon TE2000-S microscope using a Hamamatsu Orca-ER camera, outputting to Openlab v4.0.2 (Improvision, Lexington, MA, USA). Cells were maintained at 37°C and % CO2 matching above incubation conditions for each cell type or 10% for hybrids. A z-stack of images from the x4 objective with 10–50 µm between focal planes was collected at each time point using custom-designed Openlab 4.0.2 automations. Images were analyzed using Volocity 3.1 (Improvision). Medium was carefully added 20–30 min postseeding before gels were transferred to the microscope.
For horizontal-view microscopy, spheroids were harvested by inverting the gels in 6-well plates and centrifuging at 800 rpm for 6 min. Spheroids were resuspended in medium and 30 µl was added to each viewing gel. Using a Mitutoyo FS-110 microscope modified to lie horizontally, bright-field images of the front face of the spheroids were taken. Harvesting and microscopy were performed 2 days after cell seeding.
Morphological changes were measured with ImageJ software (Rasband, W.S., ImageJ, U.S. National Institutes of Health, Bethesda, MD, USA, http://rsb.info.nih.gov/ij/, 1997–2006). Spheroid aspect ratios were computed from spheroid height and width was obtained from horizontal-view images. The length of a rod was the length of a line drawn from end to end of the structure (long axis length). The core circumference of tori was measured as a continuous circumferential line located at the estimated midpoint of the perpendicular width of the toroid.
Cell viability and staining
Cell viability was determined using the LIVE/DEAD® Viability/Cytotoxicity Kit (Invitrogen), which showed that assembled structures were viable whereas debris at the bottom of the recesses was not. After every experiment, cell aggregates were harvested by inversion and agitation, then allowed to settle and attach to the plastic. Initial and subsequent daily images were obtained until complete deconstruction of original aggregate form had occurred (up to 10 days).
NHFs and H35s were stained with fluorescent dyes CellTrackerTM Red CMTPX and CellTrackerTM Green CMFDA (Invitrogen), respectively, to visualize their positions within in complex shapes. One day before seeding, subconfluent flasks were incubated with serum-free DMEM containing 2.5 µM CellTrackerTM for 45 min at 37°C. The DMEM was discarded and replaced with normal, dye-free medium (with serum). Stained cells were trypsinized and seeded into hydrogels with a NHF:H35 ratio of 1:1 and imaged by fluorescence microscopy.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
). Despite a larger mean cell diameter than H35s (NHF=20±3 µM, n=20; H35=17±2 µM), NHF spheroids were consistently smaller in width and height, and therefore in cross-sectional areas. The aspect ratio of the NHF spheroids was significantly lower than that of H35 (NHF=1.18±0.14, n=89; H35=1.44±0.25, n=80; Student’s t test, P<0.05) regardless of spheroid size, and more variability was apparent among larger H35 spheroids
Using micromolded nonadhesive agarose hydrogels, we determined whether these two cell types differed in their abilities to self-assemble into shapes more complex than a simple spheroid. NHFS, H35s, and their 1:1 mixture were seeded onto gels containing toroid or trough-shaped recesses (Fig. 2
). Within minutes after seeding, cells settled into and coated the curved recess bottoms with several layers of cells but did not fill the recesses. Directed self-assembly began immediately, with cells aggregating to form smooth-surfaced 3-D structures. In trough-shaped recesses, cells self-assembled into round-ended cylindrical rods and shrank along their long axes toward the spheroid morphology. NHF and hybrid rods became spheroids in as few as 15 h, but H35s maintained the rod morphology. NHFs, H35s, and hybrids all formed stable tori that conformed around the central agarose peg. Consistent with side-view images of spheroids, NHF tori were thinner than those of H35. Hybrid thickness was less regular, with foci of higher cell density connected by thinner regions. Whereas both NHF and hybrid tori could be formed with an initial seeding density of 2 x 106 cells/gel, H35 tori were stable even with a seeding density of 1 x 106cells/gel (data not shown).
To examine the kinetics of rod and toroid formation, we measured stability and morphological changes over time (Fig. 3
). Percentages of original tori either intact around the pegs or off the pegs but with a patent center are plotted. On day 5 there were significantly more intact H35 (88%) than NHF (30%) or hybrid (60%) tori. Notably, stability of hybrid tori was intermediate between that of H35 and NHF.
To quantify the changes that occurred during directed self-assembly, we measured the length of rods and the core circumference of tori over time (Fig. 3)
. The length of H35 rods changed at a significantly slower rate than those of NHFs and hybrids. NHF and hybrid rods became spheroids overnight, whereas H35 rods reached a final stability of 49 ± 3% of their original length by day 5. Data on rods formed in troughs of increasing length suggest that this percentage remains constant regardless of maximal length. Whereas NHFs and hybrids (tori and rods) progressed to their final morphology overnight, H35 tori reached their final toroidal morphology in 2 days whereas H35 rods required 5 days.
Core circumference was maximal at time 0, when cells coated the bottom of the recesses. As self-assembly proceeded, the core circumference decreased until the tori conformed to the peg. This process occurred at strikingly different rates between H35s and NHFs and hybrids. Steady state was achieved by day 2 in H35s, but in less than a day for both NHFs and hybrids. These data show that stable tori have different core circumferences depending on cell type, NHFs being the shortest, H35s the longest, and hybrids nearly midway between the two. Note that the rate of change in core circumference for hybrids and NHFs was similar but that hybrids formed more stable tori than NHFs. Only H35s differed in their kinetics of rod and toroid changes, with core circumferences progressing at a slower rate than rod lengths (data not shown).
Time-lapse microscopy was used to investigate the early kinetics of rod and toroid-directed self-assembly (Fig. 4
and Supplemental Figs. 1 and 2). Photos of rods in 1.6 mm troughs were taken at 10-min intervals and the long axis was plotted as a function of time. When NHF rods began to self-assemble, a necking or pinching in the middle and across the short axis was often observed. The initial length of NHF rods remained nearly static for 20–30 min before rapidly shrinking by 40% over 50 min, followed by a transition to a slower period of shrinkage (15%) over 1120 min. The width of the NHF rods narrowed over the first 50 min, then slowly increased to match the minimum length. Final length and width were nearly identical, indicating spheroid formation. Hybrid rods also progressed to a spheroid with exponential kinetics similar to NHFs (hybrid tau=347 min; NHF tau=202 min). In contrast, there was no lag period for H35 rods, and directed self-assembly proceeded with linear kinetics at a much slower rate.
To evaluate cell sorting within tori, cells were labeled with fluorescent dyes (H35s, green; NHFs, red) and allowed to self-assemble for 2 days (Fig. 5
). Sorting occurred within 24 h, in parallel with changes in core circumference. NHFs were biased inward toward the peg, particularly in thicker regions of the toroid. Their morphologies were stretched, most notably in areas of high stress, but rounded in relaxed tori that had been freed from the pegs. In contrast, H35s were rounded regardless of position within the toroid and occupied the outermost regions away from the peg. Their morphologies remained rounded in relaxed tori. Time lapse microscopy was used to monitor directed self-assembly of honeycomb structures of cell types and hybrids (Fig. 6
). As self-assembly proceeded, tension was generated across the entire honeycomb structure so that contact occurred only at the outer edge of the outermost six pegs. Contact points were also the sites of the narrowest cellularity across the honeycombs. In the case of NHFs and hybrids, enough radial tension was generated over 12–15 h to pull the honeycomb structures up and off the outside pegs, although a larger percentage of hybrid honeycombs remained at least partially within their recesses and wrapped around several pegs (NHF: 0/9 original structures; hybrid: 3/9 at 15 h). Even without the pegs, the architectural integrity of the resulting free-floating NHF and hybrid honeycombs was preserved as the structures contracted. Patency of the lumens was lost during this contraction, but the structures did not progress to spheroids. Consistent with their ability to self-assemble rods and stable tori, H35s formed intact and stable honeycombs (n=9/9), which remained in the mold anchored to the outer edge of the six outermost pegs. When removed from the mold, H35s maintained their honeycomb structure for days (data not shown).
|
|
To assay self-segregation in this complex structure, labeled H35s and NHFs were seeded in the honeycomb mold and viewed 1 and 2 days later (Fig. 7
). As in labeled tori, sorting occurred within 24 h, with NHFs centrally located and H35s on the periphery coating the entire structure. NHFs had a stretched and smooth cell morphology and H35s were more rounded, with visible cell boundaries. Self-segregation varied with position in the honeycomb. Along the outer edge of the six outermost pegs, where tension was highest and the structure was contacting the pegs, the fibroblasts were located less centrally and were closer to the peg, as was the case for tori. At locations in the interior of the honeycomb where tension was more equally distributed and not localized, NHFs were centrally located within the H35 coating.
|
We also evaluated the cell position and morphology of hybrid honeycombs that freed themselves from the pegs. In these relaxed structures, relative cell positions remained unchanged, but there was a significant change in fibroblast morphology. Instead of the indistinct cell boundaries and elongated spindle-like morphologies found in taut structures, NHFs in relaxed honeycombs were rounded and cell boundaries were easily discernible. This was in contrast to H35s, which remained round and distinct from neighbors in both stretched and relaxed conformations.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
, we have shown for the first time that mono-dispersed cells can be directed to self-assemble into microtissues with complex geometries (rods, tori, and lumen-containing honeycombs) and that their ability to do so varies with cell type. These data expand our understanding of the rules governing self-assembly by demonstrating that the generation of self-assembled, complex, and even branched structures is feasible.
Using horizontal-view microscopy, we show that NHF spheroids sag less than those of H35. In experiments on the thermodynamic properties of equilibrium state aggregates, Phillips and Steinberg demonstrated that the ability of an aggregate to round up within a centrifugal field that tends to flatten it depends on intercellular cohesiveness, and therefore higher surface tension (22)
. Together with our own, these findings would suggest that NHF spheroids have a higher surface tension than H35 spheroids.
A fundamental finding of this study was that in troughs of up to 2.2 mm long, H35s self-assembled rod-like structures and did not progress to spheroids, as would be predicted by previous studies and as did both NHFs and a mix of NHFs and H35s in our study (9
, 10
, 17)
. The differential adhesion hypothesis (DAH), first proposed in 1962 to explain self-assembly and self-segregation within a spheroid, is based on three assumptions: cells are discrete units within a spheroid; cells are mobile; and different cell types are differently adhesive and cohesive (23
24
25
26)
. Modeled after the behavior of immiscible fluids, cellular aggregates assume a spherical geometry to maximize adhesion and minimize energy; self-segregation occurs due to differences in cell-cell adhesion or apparent surface tension, with cells of highest cohesion on the inside and those with lower cohesion on the outside (10)
. Interpreted in light of the DAH, our data suggest that the strength of intercellular cohesion driving the self-assembly of NHFs is sufficient to drive spheroid formation by NHFs, but not by H35s. This assertion is further supported by the hybrid sorting data in which H35s coated NHFs.
From the viewpoint of energy minimization, the finding that H35s form rods may be explained as follows. Both NHFs and H35s should form spheroids and are shown to do so in this study. To form spheroids, cells must rearrange. This rearrangement comes at an energetic cost to the system. Provided that this cost results in net energy minimization, spheroid morphology will be obtained. However, if this is not the case, an alternative morphology will be maintained. Interpreted in light of these principles, our data suggest that the design of nonadhesive gel recess architecture can be used to bring out energetic differences in cell types so that microtissue shapes more complex than the spheroid can be maintained. In the case of H35 rods, the rearrangement required from a sheet that forms along trough bottoms at initial seeding to a spheroid is more costly energetically and thermodynamically than it is to elongated aggregates, or "rods." In hybrids, the addition of NHFs to the mix therefore must represent a net gain of energy with which the system may overcome this energetic barrier. Further, it is likely there is a point in increasing trough length after which NHFs alone would be unable to generate spheroids and would therefore remain as rods. Further experimentation is required to confirm these assertions.
The kinetics of directed self-assembly revealed interesting differences between NHFs and H35s as well as their mix. After a slight delay, NHF rod lengths decreased rapidly with exponential kinetics, then decreased more slowly while approaching spheroid morphology by 1100 min. In contrast, H35 rod lengths decreased with almost linear kinetics during the first 24 h, then with slower kinetics over the ensuing 4 days. Clearly, the forces driving self-assembly of NHFs act more rapidly than those driving H35s. The kinetics of the 1:1 mix of NHFs and H35s was not intermediate and was more similar to the kinetics of NHFs rather than H35s. The mix had the same slight delay, followed by an exponential decrease in rod length, with a curve that was nearly identical to that of NHFs but delayed by 
2 h. Since the mix contains half the number of NHFs, it is not too surprising that it self-assembles at a slower rate. However, the mix also contains an equal number of H35s randomly interspersed among the NHFs. These H35s are spatial obstacles to the cell surface interactions involved in self-assembly, and so might be expected to slow the reaction even further so that it more closely approximates the kinetics of pure H35s. That the mix proceeds with kinetics similar to that of pure NHFs suggests that either the mix self-segregates in fewer than 2 h or that the forces driving self-assembly of NHFs are only weakly affected by the presence of H35s. Indeed, the final shape of the mix is not intermediate between the NHF spheroid and the H35 rod, but is a spheroid.
Another interesting observation of the mix is that the number of stable tori on day 5 of an equal mix of NHFs and H35s was intermediate between pure H35s and NHFs. Unlike the troughs that give rise to unconstrained structures such as rods and spheroids, self-assembled tori are constrained structures that constrict around the central peg of agarose. The kinetics of constriction of the mix, as measured by a change in core circumference, are similar to the kinetics of rod-directed self-assembly. Pure NHFs proceed more rapidly than pure H35s, and the kinetics of the mix are similar to that of pure NHFs. The fact that the stability of mixed structures is intermediate between pure NHFs and pure H35s suggests that the presence of H35s and/or a slowing of constriction of a constrained structure significantly increases stability, perhaps by facilitating the development of more balanced tension in the toroid structure.
The self-assembled honeycomb structures further demonstrate that tension can be balanced and distributed throughout a complex self-assembled structure. Honeycombs of NHFs, H35s, and a mix were each anchored by contact with the outer edge of the six outer pegs. The structure had little if any contact with the inner seven pegs, presumably because the constrictive forces of a single honeycomb cell were distributed and balanced by the forces of neighboring honeycomb cells. In fact, the honeycomb shape spontaneously self-assembled as a result of the balance and distribution of this tension. Our time-lapse experiments clearly show that the entire honeycomb structure is under tension and that the presence of NHFs significantly increases this tension. Honeycombs of NHFs and the NHF/H35 mix spontaneously popped off the outer pegs and the structures quickly contracted. These data introduce the possibility that cellular contraction, in this case by NHFs more than H35s, may influence directed self-assembly. The observations that H35s maintained constant cuboidal morphology in taut vs. relaxed structures, whereas NHFs changed from elongated to spherical, support this possibility as do the well-known contractile properties of fibroblasts (27
28
29)
. Because actin stress fibers assemble and connect with intercellular adhesions, it is likely that contraction could account in part for increased kinetics of morphological changes and decreased microtissue stability during directed self-assembly, as seen with NHFs (30)
.
Tension also may play a role in the self-segregation of NHFs and H35s in complex shapes. In spheroids, different cell types self-segregate and envelop one another due to differences in surface adhesion (4)
. This was observed for NHFs and H35s in tori and honeycombs, with NHFs centrally located and H35s coating the periphery. However, in tori the ring of NHFs was not at the midpoint but was located closer to the agarose peg. In honeycombs, the band of NHFs on the outer edges of the outer pegs, an area of high tension, was also closer to the agarose peg, but the band of NHFs in the interior of the honeycomb, where tension was balanced, were more equally spaced. These data show empirically that in addition to differential adhesion, tension can also specify spatial positioning of cells within self-assembled, heterotypic microtissues.
In addition to tension, the agarose pegs themselves imparted further control over spatial positioning by constraining self-assembly without adhering to cells. When self-assembling, hybrid tori or honeycombs did not break around the pegs, patent lumens were formed that were lined by a thin coating of H35s. Therefore, despite being less adherent, H35s occupied an "internal" position within tori and honeycombs around the pegs. This represents an easy way to dictate cell position that could be exploited for tissue engineering purposes.
As freed honeycombs shrank, their lumens narrowed. Shrinkage and narrowing were greatest for honeycombs of pure NHFs, least for pure H35s, and intermediate for their mix. Patency of the lumens was maintained for 2 days for NHFs and for at least 5 days for H35s. Changes to the geometry of the mold, such as increasing the diameter of the agarose pegs, may help to increase patency as well as control lumen size. Maintenance of lumen patency is another significant difference between cell types in their self-assembly properties.
Self-assembled rod-like structures and honeycombs offer interesting new possibilities for the engineering of 3-D microtissues for in vitro and in vivo applications. Unlike a spheroid, whose ultimate size is severely limited by diffusion, the length of a rod structure has no theoretical limit provided its radius remains within the critical diffusion limit needed to maintain cell viability (3
, 31
, 32)
. Such rod-like structures may have applications in bioreactors. The honeycomb, which is essentially a combination of rods and tori in a structural design, is well recognized as an efficient geometrical shape and may also have applications in bioreactors. 3-Dimensional branching microtissues with open lumens more closely approximate the in vivo environment, and lumen structures offer the possibility of endothelializing self-assembled microtissues for transplantation.
|
ACKNOWLEDGMENTS |
|---|
Received for publication April 3, 2007. Accepted for publication June 7, 2007.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
) spheroids were measured from side-view images. The dashed line is a perfect sphere for reference. Representative micrographs are shown (inset: NHF, a; H35, b. Scale bar=100 µm). NHFs spheroids of all sizes were more spherical whereas greater variability is apparent among the larger H35 spheroids.
,
), H35s (
,
) were seeded onto micromolded gels with trough (A) or toroid features (A, B). The number of stable tori with patent centers was measured over time; n = 728 (H35), 511 (NHF), 474 (hybrid). A) The stability of hybrids was intermediate to NHFs and H35s. The kinetics of directed self-assembly was determined by measuring the length of rods [solid lines; n=191 (H35), 210 (NHF), 112 (hybrid)] or core circumference of tori (dashed lines) over time; n = 728 (H35), 511 (NHF), 474 (hybrid). Hybrids self-assembled with kinetics similar to pure NHFs. Histograms of normalized rod lengths (gray bars) or rod length as a fraction of trough length (white bars) on day 5 show that H35 rod lengths increased linearly with trough length but were consistently 49% of trough lengths (C; rods were longer at edges due to edge effects on seeding density). Data are means ± SD.
