Learning from the Extraordinary: How the Highly Derived Larval Eyes of the Sunburst Diving Beetle Can Give Insights Into Aspects Of

Home , Abedus, Thermonectus, Thermonectus marmoratus

Learning from the extraordinary: How the highly derived larval eyes of the Sunburst Diving Beetle can give insights into aspects of

holometabolous insect visual systems

A dissertation submitted to the

Division of Research and Advanced Studies

of the University of Cincinnati

In partial fulfillment of the requirements for the degree of

Doctorate of Philosophy (Ph.D.)

In the department of Biological Sciences

of the College of Arts and Sciences

2011

Nadine Stecher

B.S., University of Rostock, 2001

M.S., University of Rostock, 2005

Committee Chair: Elke K. Buschbeck, Ph.D.

Abstract

Stemmata, the eyes of holometabolous insect larvae, vary greatly in number, structure and task.

The stemmata of the Sunburst Diving Beetle, Thermonectus marmoratus, are among the most sophisticated. The predatory larvae have six eyes and a potentially light-sensitive spot (eye spot) adjacent to the stemmata. The forward-pointing tubular eyes Eye 1 (E1) and Eye 2 (E2) are involved in prey capture, and possess a biconvex lens, a cellular crystalline cone-like structure, and tiered retinal tissue. A distal and a proximal retina can be distinguished, which differ not only in morphology but possibly also in function. E1 has an additional retina which runs medially alongside the crystalline cone-like structure. Using transmission electron microscopic preparations, I described the ultrastructure of the retinas of the principal eyes E1 and E2. The proximal retinas are composed of photoreceptors with predominantly parallel microvilli, and neighboring rhabdomeres are oriented approximately orthogonally to each another. This rhabdomeric arrangement is typical for eyes that are polarization sensitive. A similar organization is observed in a portion of the medial retina of E1, but not in either of the distal retinas. Since Thermonectus marmoratus larvae are visually-guided predators, polarization sensitivity would perhaps improve their ability to detect prey with polarization features.

Measuring hunting performance of larvae under polarized or unpolarized illumination conditions, I have shown that polarized illumination decreased the latency to prey detection and improves capturing success. Although polarization-sensitivity is expected to be common among holometabolous insect larvae due to the rhabdomeric nature of their stemmata, no larvae have been named that possess polarization sensitivity that is involved in prey capture. Moreover, while many invertebrates that are polarization sensitive have polarization-specific regions in

their visual system, the eyes of Thermonectus marmoratus larvae potentially have a division of labor within the same stemma. The interesting question arises as to how these stemmata develop.

Although they are structurally very unlike each other, stemmata are considered to be homologous to adult compound eyes. It is perhaps in their development that one can find evidence for a common ancestry. Using basic histological methods, I observed stemmatal development in Thermonectus marmoratus embryos, and contrasted my finding to structural changes during compound eye development. The latter is described best in Drosophila. Similar to Drosophila ommatidia, the stemmata originate from a proliferative, pseudostratified epithelium. The photoreceptor cells differentiate in sequence, and they appear before the accessory cells differentiate. In Thermonectus marmoratus, the photoreceptor cells that are located in the proximal layer in the fully-developed stemma assume their position earlier than the prospective distal photoreceptors. In contrast to the Drosophila compound eye, which is characterized by distinct periods of high mitotic activity, cell proliferation in T. marmoratus stemmatal development appears to be a continuous but low-level process. Moreover, I did not observe a morphogenetic furrow-like differentiation process such as it is typical for Drosophila ommatidial development. Based on the morphological study, it will be possible to investigate molecular similarities between the development of compound eyes in Drosophila and the stemmata in Thermonectus marmoratus.

iii

The copyright to th article entitled “Retinal ultrastructure may mediate polarization sensitivity in larvae of the Sunburst diving beetle, Thermonectus marmoratus (Coleoptera: Dytiscidae)” is transferred to Springer-Verlag (respective to owner if other than Springer and for U.S. government employees: to the extent transferable) effective if and when the article is accepted for publication. The author warrants that his/her contribution is original and that he/she has full power to make this grant. The author signs for and accepts responsibility for releasing this material on behalf of any and all co-authors. The copyright transfer covers the exclusive right to reproduce and distribute the article, including reprints, translations, photographic reproductions, microform, electronic form (offline, online) or any other reproductions of similar nature. An author may self-archive an author-created version of his/her article on his/her own website and or in his/her institutional repository. He/she may also deposit this version on his/her funder’s or funder’s designated repository at the funder’s request or as a result of a legal obligation, provided it is not made publicly available until 12 months after official publication. He/ she may not use the publisher’s PDF version, which is posted on www.springerlink.com, for the purpose of self- archiving or deposit. Furthermore, the author may only post his/her version provided acknowledgement is given to the original source of publication and a link is inserted to the published article on Springer’s website. The link must be accompanied by the following text: “The final publication is available at www.springerlink.com”. Prior versions of the article published on non-commercial pre-print servers like arXiv. gov can remain on these servers and/or can be updated with the author’s accepted version. The final published version (in pdf or html/xml format) cannot be used for this purpose. Acknowledgement needs to be given to the final publication and a link should be inserted to the published article on Springer’s website, accompanied by the text “The final publication is available at springerlink.com”. The author retains the right to use his/her article for his/her further scientific career by including the final published journal article in other publications such as dissertations and postdoctoral qualifications provided acknowledgement is given to the original source of publication. The author is requested to use the appropriate DOI for the article. Articles disseminated via www.springerlink.com are indexed, abstracted and referenced by many abstracting and information services, bibliographic networks, subscription agencies, library networks, and consortia. The copyright of the chapters entitled “The influence of polarized light illumination on the hunting success in the first instar larvae of the Sunburst Diving Beetle, Thermonectus marmoratus (Insecta, Dytiscidae)” and “Development of the larval eyes of the Sunburst Diving Beetle, Thermonectus marmoratus (Insecta: Dytiscidae)” belongs to the author, Nadine Stecher.

Acknowledgements

I would like to thank my advisor Elke Buschbeck for her generous support and interest in my projects, on which I was able to work under her supervision but still independently. Elke always manages to give each single student special attention in their time of need.

I would like to thank my committee members: George Uetz for his help with my statisctics. Ed

Griff for asking those questions I never would have have thought of asking, and also for the mentoring I received being his TA. Tiffany Cook for sharing her expertise on eye development, and John Layne for his help with computational analyses.

Thanks to Guy Cameron for giving me a UGS for the home stretch of my thesis work when I was really no longer elibible for funding.

Randy Morgan and the Insectarium of the Cincinnati Zoo & Botanical Garden provided the initial culture of Sunburst diving beetles.

This research was funded by the National Science Foundation (IOB-545978).

Special thank you hugs go out to my friends and emotional support group, featuring Prem

Rajkumar, Sri Pratima Nandamuri, Shannon Werner, Jessie Ebie, Shira Gordon, K Marie Hoefer, and Srdjan Maksimovic. Having these guys in my life really helped me through grad school and my personal life as they would always pick me up when I was feeling low. They also shared many happy days with me, of which there were plenty, and always let me believe that my jokes were funny. We talked a lot about science, too.

I am eternally grateful to my parents, who have always believed in me and helped me out of any bad situation. I know I am very lucky having a family that is as caring and supportive as mine.

And most of all, my love Suman, who has been my anchor for the past few years.

Contents Introduction ...... 1

Chapter 1

Retinal ultrastructure may mediate polarization sensitivity in larvae of the Sunburst diving beetle, Thermonectus marmoratus (Coleoptera: Dytiscidae)...... 6

Abstract ...... 7

Introduction ...... 8

Materials & Methods ...... 12

Animal care ...... 12

Ethyl gallate staining...... 12

Tissue preparation and TEM ...... 13

Tissue analysis ...... 14

Mosquito picture ...... 15

Results ...... 16

Medial retina (MRe) ...... 16

Distal retina (DRe) ...... 18

Proximal retina (PRe) ...... 21

Consistency of microvillar organization ...... 22

Discussion ...... 25

1) Potential for polarization sensitivity in the proximal and medial retinas of the principal

eyes ...... 25

2) The distal retinas and the central region of the medial retina are unlikely to mediate

polarization sensitivity ...... 28

3) The ecological importance of polarization sensitivity ...... 28

Chapter 2

The influence of polarized light illumination on the hunting success in the first instar larvae of the Sunburst Diving Beetle, Thermonectus marmoratus (Insecta, Dytiscidae)...... 32

Introduction ...... 34

Materials and Methods ...... 38

Animal care ...... 38

Analysis of polarization features in mosquito larvae ...... 39

Test chamber ...... 39

Behavior ...... 42

Data analysis ...... 44

Analysis of light ...... 44

Results ...... 49

Behavior ...... 49

Properties of the illumination...... 52

Discussion ...... 57

properties of the illumination ...... 59

vii

Chapter 3

Development of the larval eyes of the Sunburst Diving Beetle, Thermonectus marmoratus

(Insecta: Dytiscidae)...... 64

Abstract ...... 65

Introduction ...... 67

Materials and Methods ...... 72

Beetle husbandry ...... 72

Ethyl gallate staining...... 73

Phalloidin and DAPI staining ...... 73

Images ...... 75

Three-dimensional reconstruction ...... 76

Results ...... 77

The origin of the stemmata ...... 80

Stemmata Eye 1 and Eye 2 ...... 81

The medial retina and the eye patch ...... 90

Discussion ...... 93

The prospective eye tissue ...... 94

Photoreceptor cells and accessory cells ...... 96

Lens and crystalline cone-like structure ...... 98

Eye patch (EP) ...... 101

viii

Ensuing research projects ...... 102

Conclusion ...... 105

Bibliography ...... 108

List of tables and figures

Figure 1 a SEM image of the head of first instar larva of Thermonectus marmoratus. White lines

show the section level of the light microscopy image depicted in b. b semithin horizontal

section through left Eye1 (bottom), and 3-D reconstruction (top) to represent the gross

anatomy of the left E1. The 3-D image also indicates the orientation and position of

histological sections c-e. Dorsal (d), frontal (f), lateral (l), medial (m), ventral (v). c ethyl

gallate-stained horizontal section of the medial retina (MRe) of E1, with the retinula cells of

the medial retina (MRC) and the rhabdoms of the medial retina (MRh). The arrow points at

a single rhabdom (Rh), which is visible as a darker streak. d ethyl gallate-stained sagittal

section of the distal retina (DRe) of left E1. e ethyl gallate-stained frontal (cross) section of

proximal retina (PRe) of E1. Crystalline cone-like structure (CC), retinula cells of distal

retina (DRC), rhabdoms of distal retina (DRh), Eye1 (E1), Eye2 (E2), lens (L), retinula cells

of proximal retina (PRC), rhabdoms of proximal retina (PRh)...... 9

Figure 2 Medial retina (MRe). a ethyl gallate-stained frontal section of E1, showing the

crystalline cone-like structure (CC) and the different regions of the medial retina: retinula

cells of the central region (CMRC), rhabdoms of the central (diffuse) region (CMRh),

retinula cells of the peripheral region (PMRC), rhabdoms of the (ordered) peripheral region

(PMRh). b TEM image of the central area of medial retina, frontal section. c TEM image of

the medial retina, frontal section, peripheral area with the two different retinal cells types

(PMT1 and PMT2) with organized rhabdomeres. Zonula adherens (ZA), dorsal (d), lateral

(l), medial (m), ventral (v)...... 17

Figure 3 Distal retina. a TEM image, frontal section, showing the tree-like branching of the

microvilli and the extended cell body (“trunk”). dorsal (d), lateral (l), medial (m), ventral

(v). b Composite TEM image of the distal retina, sagittal section, showing the retinula cell

bodies (DRC), and the distal rhabdoms (DRh). Crystalline cone-like structure (CC),

proximal retina (PRe)...... 20

Table 1 Angular distribution of the microvilli of proximal retina cell type 1 (PT1), cell type 2

(PT2) and cell type 3 (PT3) of Eye1 (E1) and Eye2 (E2). 0° corresponds to horizontal

(mediolateral) microvillar alignment, and 90° corresponds to vertical (dorsoventral)

microvillar alignment. The rhabdoms of PT2 are oriented approximately horizontally,

whereas the rhabdoms of PT1 and PT3 are oriented predominantly vertically (E1: p=0.737*,

E2: p=0.945^)...... 24

Figure 5 Visualization of polarization contrast in the larva of mosquito Aedes aegypti (4 mm

body length). The mosquito is illuminated from behind with linearly polarized light.

Visualization with a polarization analyzer that is oriented orthogonally to the background

polarization (crossed analyzer)...... 30

Figure 6 Aedes aegypti (4mm body length) imaged against a linearly polarized background. a

viewed without a polarization analyzer. b viewed with a polarization analyzer. Scale bars

1mm...... 38

Figure 7 Experimental setup used to test hunting performance in Thermonectus marmoratus.

The scale in the image does not represent the actual dimensions of the components, except

that the filter sandwich completely covered the top surface of the arena, as shown here. The

position of the polarizer and diffuser (=depolarizer) was switched to create linearly

polarized and unpolarized experimental illumination. For example, the order of the filters

displayed in this image would create POL conditions...... 42

Figure 8 Experimental results of testing hunting performance of Thermonectus marmoratus

under linearly polarized (blue bars) and unpolarized (yellow bars) illumination. a Mean

detection latency: time elapsed first to first overall strike, including all failed attempts. b

Mean capture latency: time elapsed to first sucessful strike, which may or may not have

been preceded by a failed attempt. c Mean umber of strikes per trial, including successful

and failed attacks, as well as `no attack`. d Mean probability of capture, as a fraction of

unsuccessful attacks to successful strikes. e Distribution of total number of strikes per trial.

Zero signifies that the Thermonectus marmoratus larva did not attempt to catch the prey

item. Per trial, one mosquito larva was available as prey...... 50

Figure 9 a absolute irradiance measurement of the down-dwelling light under POL (blue line)

and unPOL (yellow line) illumination. b percent difference between the absolute irradiance

measurements of the down-dwelling light of the POL and the unPOL illumination

conditions. c Relative number of photons absorbed in 2150 ms by an idealized

photoreceptor in the proximal retina in the POL setup (blue line) and the unPOL setup

(yellow line) per nm wavelength. The values were calculated from the absolute irradiance

measurements from a multiplied by the relative absorption curve for the TmUVII opsin (red

dotted line) from (Maksimovic et al., in press). TmUVII is expressed in the proximal retina,

and its spectral sensitivity ranges from 300-450 nm, with the absorption maximum at

360nm. This graph shows the potential relevance of brightness differences to the respective

opsin. Within the spectral range of the TmUVII opsin, an idealized photoreceptor cell in the

proximal retina would perceive the light under POL conditions to be approximately 11%

brighter than with unPOL illumination...... 51

xii

Figure 10 absolute irradiance measurements taken from below the experimental setup (a+b) and

through the front glass (c+d) under polarized (a+c, marked by a blue bar on the right) and

unpolarized (b+d, marked by a yellow bar on the right) illumination. Measurements were

taken through a polarization analyzer that was oriented parallel (blue dotted line), in 45°

(red solid line), and orthogonal (green dashed line) to the polarizer above the experimental

arena (see Figure 7 for reference)...... 54

Figure 11 Degree of polarization Ɵ (a+c, solid line in blue for POL, yellow for unPOL) and e-

vector p (b+d, dashed line in blue for POL, yellow for unPOL) of the (un)polarized

experimental conditions. For b and d, an angle of 0° (or 180°) represents a horizontal e-

vector, whereas 90° refers to a vertical e-vector. The values for p and Ɵ were calculated

based on the absolute irradiance measurements carried out under POL, and unPOL,

illumination and a polarization analyzer (see Figure 10) and the formulas given in Sabbah &

Shashar (2006). See Methods & Materials for details on the equations...... 56

Figure 12 External morphology of Thermonectus marmoratus embryos. a Embryo 8 hours post-

oviposition. The egg appears uniformly white. b Embryo 27 hours post-oviposition. The

embryo (ventral, dashed outline) is discernable from the egg yolk (dorsal). Arrow points to

approximate location of developing head. c Embryo 56 hours post-oviposition, the

stemmata 1 through 6 are clearly visible externally as separate units with yellow-light

brown pigmentation. Insert: ellipse indicates the position of the medial retina (mre) of E1.

Arrowhead points at the anteroposterior cleft. d Same stage as c, head in dorsal view. e

Embryo 80 hours post-oviposition. The stemmata are wider than in earlier stages (c&d) and

their pigmentation darkens. Insert highlights location of mre. f Same stage as e, head in

dorsal view, box indicated position of eye patch (EP). g Embryo 104 hours oviposition. The

xiii

stemmata possess lenses (arrow, L) and dark-brown pigmentation. Insert: EP is

unpigmented but separated from E1. h Same stage as g, head in dorsal view. Orientation for

specimens a-c, e and g (incl. inserts) given in a, unless otherwise noted. Orientation for

specimens d, f, and h given in d. ant anterior, dor dorsal, lat lateral, post posterior. i

Diagram of an embryo in lateral view, defining the line of reference used to describe

positions of structures in relation to the body axis. k Diagram of an embryo in lateral view

of the head, defining the line of reference used to describe positions of structures in relation

to the head axis. Scale bars 1mm...... 79

Figure 13 a&b Sagittal section of embryo 12 hours post-oviposition. Nuclei of imaginal disc

epithelia are stained with DAPI (colored in blue). Mitosis-active cells are marked with anti-

histone H3 (colored in red). a Approximate location of imaginal disc anlagen. b Higher

magnification of cephalic imaginal disc as framed in a, showing low degree of mitosis

activity. c Embryo at 8 hours, demonstrating the section levels of a&b. d Embryo at 27

hours, demonstrating the section level of e. e Ethyl gallate-stained sagittal section of the

pseudostratified epithelium of the eye disc at 30 hours, highlighting one of the developing

principal eyes. Scale bars 100µm...... 80

Figure 14 a Ethyl gallate-stained sagittal section of eye anlagen 36 hours post-oviposition. The

stemmata develop within the outer epidermis of the head. In sagittal sections, the stemmata

appear as a continuous cell layer. Cell populations begin to morphologically differentiate.

Axonal projections (ax) identify the developing photoreceptor cells. b Embryo at 54 hours,

in lateral view demonstrating the section levels of a, c-g. c Ethyl gallate-stained horizontal

section of the cup-shaped principal stemmata E1 (right) and E2 (left). At this stage, cell

populations are more distinctly differentiated morphologically and photoreceptor cells with

xiv

their axonal projections into their respective optic lobes (op lb) are visible. Developing

corneagenous cells (cng) are located on the periphery of each stemmatal cup. The

corneagenous cells of one stemma form a continuum with the corneagenous cells of the

adjacent stemma. The middle of each approximately symmetric stemma is established by

the embryonic fissure (arrowhead) d higher magnification of the area outlined in c,

highlighting the lack of morphological separation between the adjacent populations of

developing corneagenous cells of neighboring stemmata. Scale bars 50µm...... 82

Figure 15 Embryo 42 hours post-oviposition. a-d ethyl gallate-stained horizontal sections of E2

at different levels of depth. In horizontal sections, the location and appearance of the

developing retina (re) and the corneagenous cells (cng) remain relatively unchanged

throughout the long axis of the stemma, suggesting that the photoreceptor cells and

corneagenous cells are oriented in parallel layers, with the corneagenous cells at the

periphery, and the photoreceptor cells centrally. In these sections the stemma appears

approximately symmetric, with the apical embryonic fissure (arrowhead) marking its

middle. Scale bars 50µm. e Embryo at 48 hours, indicating the approximate section levels

of a-d. f three-dimensional reconstruction of E2 based on an ethyl gallate-stained horizontal

series, illustrating the layered appearance of corneagenous cells (cng, colored in purple) and

photoreceptor cells (re, colored in green). Approximate section levels of a-d are indicated.

...... 83

Figure 16 Embryo 54 hours post-oviposition. a Ethyl gallate-stained sagittal section of E1.

Distal (dre) and proximal retinas (pre) are morphologically separated. Distal photoreceptor

cells have secreted a clear, refractive substance at their most dorsal region (ellipse). b

Sagittal section of E1 with DAPI (blue, marked nuclei) and Phalloidin (red, marking actin)

staining. Actin present in the developing rhabdoms of the distal retina (drh) and the

proximal retina (prh). c Ethyl gallate-stained sagittal section of E1, apical-most region. In

this region, corneagenous cells are rich in vesicles that contain a clear substance. d-g Ethyl

gallate-stained horizontal sections of E2 at four different levels of depth; approximate

section levels are indicated in i (left). d The vesicle-rich area of the corneagenous cells as

well as the embryonic fissure are deepest most anteriorly. The anterior part of the stemma

also is dominated by the corneagenous cells. e In a section posterior to d, the vesicle-rich

region of the corneagenous cells is shallower, as is the depth of the embryonic fissure.

Corneagenous cells overlie the distal retina, and the dre is situated above the proximal

retina. f Section level posterior to e. The dre overlies the pre. g Section level posterior to f.

The posterior-most region of the stemma is dominated by proximal retinal tissue. h Three-

dimensional reconstruction of E2 based on an ethyl gallate-stained horizontal section series,

showing the layered appearance of corneagenous cells (colored in purple), distal (colored in

blue) and proximal (colored in yellow) photoreceptor cells. The narrow, proximal cell

extensions of cng and distal photoreceptor cells were omitted in order not to obscure the

proximal retina. i Embryo at 54 hours, indicating section levels of d-g (left) and k-n (right).

k-n Ethyl gallate-stained series of sagittal sections; approximate section levels indicated in i

(right). Confirming the rearrangement of the tissue layers compared to 42 hours post-

oviposition (Figure 4). l Example of how narrow cell extensions of the cng (purple arrow)

and dre (blue arrow) reach the base of the stemma. Scale bars 50µm...... 85

Figure 17 Embryo 72 hours post-oviposition. a-d Ethyl gallate-stained sagittal sections of E2 at

different section levels as depicted in e and f. a-c Corneagenous cells (cng), distal retina

(dre) and proximal retina (pre) assumed their final, stacked, relative positions as they are

xvi

found in the fully-developed larval stemma. b Distal photoreceptor cells have thin

extensions (blue arrow) that reach the basal membrane and enclose the proximal retina. The

distal retina is surrounded by narrow proximal extension of the corneagenous cells (purple

arrow). d Posterior-most region of eye is dominated by the tissue of the pre. e Three-

dimensional reconstruction of E2 based on a series of ethyl gallate-stained sagittal sections,

in side view. In order not to occlude the dre and pre, narrow cell extensions of cng and dre

were omitted from the reconstruction. Scale bars 50 µm...... 87

Figure 18 a Embryo at 72 hours, sagittal section of E1. Centrally, the apical region of the

corneagenous cells provides less contrast (arrow) with ethyl gallate staining than the apical

region of peripherally located cng. As seen in earlier stages (Figure 3f), a clear, refracting

substance fills the extracellular space above the central region of the distal retina (ellipse),

and it also secreted by distal photoreceptor cells. b Embryo at 84 hours; ethyl gallate-stained

horizontal section of E2. Corneagenous cells begin to secrete a uniformly thin lens (l) above

the vesicle-rich area (ves). The region of the cng that is just proximal to the vesicle-rich

region stains darkly with ethyl gallate, which causes poor resolution on a cellular level. c

Embryo 90 hours post-oviposition; ethyl gallate-stained sagittal section of E2. The center of

the lens is thicker than its periphery. The curvature of the lens is more flat on the dorsal

surface than its back surface. The apical regions of the cnc form the crystalline cone-like

structure (cc), which appears separated from the proximal end of the cnc by a small segment

that stains darkly with ethyl gallate (arrow). d Embryo at 102 hours; ethyl gallate-stained

sagittal section of E2. The back surface of the lens has an increased curvature compared to a

lens at 90 hours (c). The drh appears approximately triangular in shape. The rhabdomeric

region of distal photoreceptor cells is extended centrally, except in the distal-most area, in

xvii

which the rhabdomeres are located distally (arrow). Extracellular material is located in

small area between the cc and the drh (circle). Scale bars 50 µm...... 89

Figure 19 Development of the medial retina (mre) and the eye patch (EP). a-c Embryo at 42

hours. Three consecutive ethyl gallate-stained horizontal sections of the anterior end of E1

(a is most anterior). The EP develops in close physical proximity to E1. Embryonic fissure

in E1 marked with arrowhead. d Embryo at 54 hours. Ethyl gallate-stained sagittal section

of the medial portion of E1. Both the anteromedial photoreceptor cells of E1and the EP

contact the epidermis. The upwards migration of the EP occurs in parallel with the

downwards movements of the photoreceptor cells. In E1, the anteromedial-most

photoreceptor cells of E1 do not fully migrate ventrally as in E2 (illustrated in Figure 5), but

instead establish the medial retina. e&f Embryo at 72 hours. e Ethyl gallate-stained sagittal

sections of medial portion of E1. The EP projects axons (ax). f Ethyl gallate-stained sagittal

section of EP, section level medial to e. EP somewhat detaches from E1 to a position more

medial and posterior to E1. g&h Embryo 84 hours post-oviposition. Medial retina

positioned in the medioanterior region of E1. g Ethyl gallate-stained horizontal section of

E1 and EP. Although they lie in close proximity, the EP and E1 appear completely

detached. h Ethyl gallate-stained horizontal section of far-anterior region of E1. Unless

otherwise noted, the orientation of all structures in each line is presented in the respective

left-most image. Scale bars 50µm...... 91

xviii

Introduction

Scientists have long been fascinated by visual systems. Naturally, the research was biased towards the understanding of the human eye and that of mammalian vertebrates, and advances in optics and the description of the physics of light contributed much to the understanding of vision.

For example, more than 400 years ago, Johannes Kepler first explained the optics of the eye by combining geometric optics and physical concepts of light, and suggested that light is focused onto the retina by the lens, creating a real, inverted image. Although there have been earlier attempts to characterized the retina, the first detailed description of the mammalian retina was given in 1819 by Arthur Jacob. Slowly, the attention of vision scientists was directed towards the compound eye as well, but it appears it was not until the mid-20th century that invertebrate eyes were considered a worthwhile study subject. Largely responsible for sparking interest in rhabdomeric photoreceptors was the honey bee Apis mellifera and its sensitivity to the polarization of light, which was demonstrated through behavioral experiments. Bees are able to use the polarization pattern of the sky to determine the position of the sun, even when the sun itself is occluded by clouds (von Frisch, 1949; von Frisch and Lindauer, 1956).

The honey bee was not the first animal that was known to be able to distinguish between polarized and unpolarized light, but it pioneered in show ing that polarization sensitivity can be exploited as a tool for navigation by an arthropod. This discovery was as puzzling as it was challenging because it was not yet known where and how photoreception took place in compound eye ommatidia, let alone how they would be able to ‘read’ the polarization pattern of the sky. Subsequent research revealed the microvilli of the retinula cells as the site of light

absorption (Goldsmith and Philpott, 1957), and that the orientation of the photopigment, which is embedded in the microvillar membrane, determines the grade to which these retinula cells are polarization-sensitive (Eguchi and Waterman, 1967). A more detailed description of this mechanism is given in Chapter 1 of this thesis.

Moreover, the example of the honey bee triggered an extensive search that was aimed at finding additional species that were polarization-sensitive. As it turns out, polarization sensitivity is not at all a rare phenomenon among invertebrates, as it was found in numerous insects, crustaceans, spiders and cephalopods. Furthermore, polarization sensitivity is not only utilized for navigation, but it can also play a role in intraspecific communication, object recognition and contrast enhancement. Chapter 2 of this thesis will elaborate on polarization sensitivity and its exploits.

There are a number of other invertebrates that had a significant impact on vision research.

For example, one of the discoveries the horseshoe crab Limulus polyphemus became most famous for is the occurrence of lateral inhibition. Limulus possesses compound eyes that are composed of approximately 800 ommatidia. Illuminating one isolated ommatidia decreases the neuronal activity of the neighboring ommatidia. This antagonistic process enhances responses to edges while reducing the response to a constant surface (Hartline, 1949; Hartline et al., 1956).

Even though the physiology of interactions between neurons had been well described for the vertebrate retina, finding such a ‘sophisticated’ feature as lateral inhibition in such a ‘simple’ eye as that of Limulus was quite spectacular. The realization that invertebrate visual systems might reveal mechanisms that could be relevant to the understanding of the visual process in

vertebrates was perhaps greatest influence Limulus had on vision science. The 1967 Nobel Prize of Medicine was awarded in part for studies performed on the compound eyes of Limulus.

Also in the 1960s, headlines were made by jumping spiders. Like most spiders, salticids have eight pairs of eyes, but the frontal-most pair is especially prominent. What was astonishing is that these principal eyes possess a spatial resolution that is almost comparable to that of the human eye. Surprisingly, the retina of the principal eyes is organized into four tiers of photoreceptor cells (Land, 1969b), each with its own spectral sensitivity (Blest et al., 1981;

Land, 1969b), and each with its own area of high photoreceptor density similar to the fovea in the mammalian retina (Land, 1969b). The layered structure possibly compensates for chromatic aberration, a phenomenon of optic systems that causes light of a short wavelength to be converged at a point closer to the lens than light of a longer wavelength. Hence, light with a shorter wavelength (e.g. UV) is focused on the distal retinal tier while light with a longer wavelength (e.g. green) focuses on a proximal layer. In addition, the tiered retina might serve as a distance detector (Williams and McIntyre, 1980). Last but not least, the principal eyes can be moved within the head alone, which causes the retinas to be translated without moving the head.

The spider thus is able to scan across and track an object while holding its body perfectly still

(Land, 1969a).

Famous representatives of invertebrate visual systems are not only outstanding because of their physiology but also due to what is known about their embryonic development. The fruit fly Drosophila sp. is probably the best-studied representative for compound eye development.

Drosophila is a popular experimental animal because it can be cultured effortlessly and the

generation times are short. Furthermore, the genome of some Drosophila species has been fully or partially sequenced, which makes those species more accessible to genetic manipulation. Its compound eye development has been detailed through basic histological studies, which laid the groundwork for advanced methods on a molecular basis. Nowadays, many specification and differentiation events, and contributing molecular factors in compound eye development in

Drosophila are well understood, and are being used to make comparisons not only to other invertebrate visual systems but also to vertebrate eyes. Since compound eyes and the larval eyes of holometabolous insects are considered to be homologous (Liu and Friedrich, 2004; Paulus,

2000; Paulus and Schmidt, 1978), potential similarities in morphological changes and developmental pathways between the ommatidial patterning during Drosophila compound eye development and the larval eye formation in holometabolous insects might provide evidence of a common ancestor. Most importantly, this provides the basis of studies in invertebrates, the results of which then can help understand (and potentially cure) vertebrate eye diseases.

The highly-derived visual system of the holometabolous larvae of Thermonectus marmoratus (Insecta: Dytiscidae) has great potential to become an important contributor to vision research. Its original description (Mandapaka et al., 2006) emphasizes the unusualness of the layered arrangement of anatomically distinct retinas in each of the six pairs of stemmata, while the latter themselves differ in shape and position, and also in their visual task. The forward-pointing principal stemma Eye 1 and Eye 2 are involved in prey capture (Buschbeck et al., 2007; Mandapaka et al., 2006), yet their vertical visual field is extremely narrow (Mandapaka et al., 2006), which the larvae compensate for by performing dorsoventral scanning movements

(Buschbeck et al., 2007). Interestingly, the retina which is positioned closest to the lens in each

of the stemma, the distal retina, expresses long-wavelength opsins, while the retina farther away from the lens, the proximal retina, expresses UV opsins (Maksimovic et al., 2009). This is interesting because the phenomenon of chromatic aberration would call for an inverted expression pattern. The explanation for the unexpected opsin expression results is that

Thermonectus marmoratus larvae have bifocal lenses, which presumably allows an image to be focused simultaneously on two separated planes (Stowasser et al., 2010). Thus far, no other extant species is knows that possesses bifocal lenses.

Moreover, the ultrastructure of the retinas suggests that at least one of the retinal layers in the principal eyes of the larvae might be polarization-sensitive. Chapter 1 of this thesis presents a cellular description of the anatomy of the retinas of E1 and 2, and how their ultrastructure might affect polarization sensitivity. Chapter 2 explores what possible advantages might result from this proposed polarization sensitivity in the hunting behavior of the larvae. The highly derived and specialized morphology of the principal eyes leads to the question as to how these eyes develop. Chapter 3 of this thesis examines the development of the principal stemma in the embryo of Thermonectus marmoratus, and investigates structural similarities to compound eye development.

Chapter 1

Retinal ultrastructure may mediate polarization sensitivity in larvae of the Sunburst diving beetle, Thermonectus marmoratus (Coleoptera: Dytiscidae).

Nadine Stecher(1), Randy Morgan(2) and Elke K. Buschbeck(1)*

(1)Department of Biological Sciences, University of Cincinnati, Cincinnati, OH 45221, USA

Corresponding author: phone (+1) 513-556-9747; [email protected]

(2) Insectarium; Cincinnati Zoo & Botanical Garden; 3400 Vine Street; Cincinnati, OH 45220-

1399

Published in Zoomorphology, Volume 129, Issue 3 (2010) pp. 141-152 doi: 10.1007/s00435-010-0107-7

The final publication is available at www.springerlink.com.

Abstract

A number of invertebrates are known to be sensitive to the polarization of light, and use this trait in orientation, communication, or prey detection. In these animals polarization sensitivity tends to originate in rhabdomeric photoreceptors which are more or less uniformly straight and parallel. Typically polarization sensitivity is based on paired sets of photoreceptors with orthogonal orientation of their rhabdomeres. Sunburst diving beetle larvae are active swimmers and highly visual hunters which could potentially profit from polarization sensitivity. These larvae, like those of most Dytiscids, have a cluster of six lens eyes or stemmata (designated E1 through E6) on each side of the head capsule. We examined the ultrastructure of the photoreceptor cells of the principal eyes (E1 and E2) of first instar larvae to determine if their rhabdomeric organization could support polarization sensitivity. A detailed electron microscopical study shows that the proximal retinas of E1 and E2 are in fact composed of photoreceptors with predominantly parallel microvilli, and that neighboring rhabdomeres are oriented approximately perpendicularly to one another. A similar organization is observed in the medial retina of E1, but not in the distal retinas of E1&2. Our findings suggest that

Thermonectus marmoratus larvae might be able to analyze polarized light. If so, this could be used by freshly hatched larvae to find water, or within the water to break the camouflage of common prey items such as mosquito larvae. Physiological and behavioral tests are planned to determine if larvae of Thermonectus marmoratus can actually detect and exploit polarization signals.

Introduction

Polarization sensitivity is widely utilized among invertebrates for a variety of visual tasks. For example, honey bees exploit the celestial polarization pattern for navigation (Glas, 1977; Rossel,

1993; Rossel and Wehner, 1982; von Frisch, 1949; von Frisch and Lindauer, 1956). Some flying insects living in or near water locate potential habitats through the polarized reflection off the water surface (Schwind, 1984, 1991). Polarization sensitivity can also be used for intraspecific communication in aquatic animals, as has been observed for cuttlefish (Boal et al., 2004) and stomatopod crustaceans (Marshall et al., 1999). Interestingly, some predatory cephalopods utilize polarization sensitivity to detect prey which are otherwise camouflaged (Shashar et al., 2000;

Shashar et al., 1998). Underwater contrast is generally poor due to the scattering of light within the water. Therefore contrast enhancement based on differential polarization signals would be advantageous for a variety of aquatic animals (Wehner and Labhart, 2006). However, thus far only a few are known to exploit polarization contrast. In this study we examined the photoreceptor ultrastructural organization in the aquatic predatory larvae (Figure 1a) of the

Sunburst diving beetle Thermonectus marmoratus (Coleoptera: Dytiscidae). This species is native to much of the extreme southwest United States and adjacent Mexico. Both larvae and adults tend to prefer pool habitats (Velasco and Millan, 1998) but are also often found in slow- moving, seasonal streams with rocky bottoms covered with little or no vegetation (Evans, 2006;

Morgan, 1992, 1995). Larval forms have unusual and highly specialized eyes that are important for prey capture (Buschbeck et al., 2007; Mandapaka et al., 2006). Long wavelength and UV opsins in their photoreceptor cells may allow them to see colors (Maksimovic et al., 2009).

Figure 1 a SEM image of the head of first instar larva of Thermonectus marmoratus. White lines show the section level of the light microscopy image depicted in b. b semithin horizontal section through left Eye1 (bottom), and 3-D reconstruction (top) to represent the gross anatomy of the left E1. The 3-D image also indicates the orientation and position of histological sections c-e. Dorsal (d), frontal (f), lateral (l), medial (m), ventral (v). c ethyl gallate-stained horizontal section of the medial retina (MRe) of E1, with the retinula cells of the medial retina (MRC) and the rhabdoms of the medial retina (MRh). The arrow points at a single rhabdom (Rh), which is visible as a darker streak. d ethyl gallate-stained sagittal section of the distal retina (DRe) of left E1. e ethyl gallate-stained frontal (cross) section of proximal retina (PRe) of E1. Crystalline cone-like structure (CC), retinula cells of distal retina (DRC), rhabdoms of distal retina (DRh), Eye1 (E1), Eye2 (E2), lens (L), retinula cells of proximal retina (PRC), rhabdoms of proximal retina (PRh).

Our primary goal was to determine if the microvillar organization of photoreceptor cells in

Thermonectus marmoratus could potentially support polarization sensitivity. Generally, invertebrate eyes which are sensitive to the polarization of light possess rhabdomeres that are oriented orthogonally to each other (Wehner and Labhart, 2006). Polarization sensitivity is a feature that is frequently inherent to rhabdomeric photoreceptors, due to the predominantly

parallel orientation of the visual pigment to the long axis of the microvillus, and the parallel stacking of the microvilli within the rhabdomere (Eguchi and Waterman, 1968; Goldsmith, 1975;

Goldsmith, 1977; Waterman, 1975). Light is absorbed maximally when its predominant plane of polarization (e-vector) is parallel to the long axis of a microvillus (Eguchi and Waterman, 1968;

Goldsmith, 1975; Goldsmith, 1977; Wehner and Labhart, 2006). To distinguish between different planes of polarization, two sets of orthogonally oriented photoreceptors are sufficient

(Wehner and Labhart, 2006). Moreover, to avoid confusing polarization information with spectral features (Bernard and Wehner, 1977), a polarization detection system should include only a single spectral class of photoreceptors (Wehner and Labhart, 2006), an adaptation seen in some cephalopods (Brown and Brown, 1958; Marshall and Messenger, 1996). Conversely, a color detection system needs to be minimally influenced by polarization patterns. One way to separate color from polarization features (and to gain color information) is to render the polarization analyzer dysfunctional by rotating the rhabdoms (Smola and Tscharntke, 1979;

Smola and Wunderer, 1981; Snyder and McIntyre, 1975; Wehner and Bernard, 1993; Wehner et al., 1975). Indeed, the compound eyes of many polarization-sensitive insects possess twisted rhabdoms in all ommatidia but the specialized dorsal rim area of the compound eye (POL area,

(Labhart, 1980; Wehner and Strasser, 1985) used for the analysis of polarization characteristics.

Therefore, an important component of our study is to systematically determine the orientation of specific larval photoreceptor cells.

Larvae of Thermonectus marmoratus are highly efficient visually-guided predators

(Buschbeck et al., 2007; Mandapaka et al., 2006; Morgan, 1992). They bear six stemmata

(designated E1 to E6) on each side of their head capsule (Mandapaka et al., 2006). The principal

eyes (E1 and E2) are particularly large, tubular and forwardly-directed and utilized during prey approach (Buschbeck et al., 2007). The anatomical organization of these eyes indicates their sophistication and unusual characteristics. During prey approach their very narrow vertical visual field (Mandapaka et al., 2006) is enlarged by whole body dorsoventral scanning movements

(Buschbeck et al., 2007). E1 (Figure 1b-e) and E2 are characterized by multiple, distinct retinas,and in both eyes a distal and a proximal retina underlies the crystalline cone-like structure. In E1 there is an additional medial retina (Figure 1c) that extends along the length of the crystalline cone-like structure (this retina was initially termed “lateral retina” in Mandapaka et al. (2006), and then “medial retina” in Maksimovic et al. (2009)). Within the proximal retina

(of E1 and E2), Mandapaka et al. (2006) described the presence of “horseshoe-shaped” and

“oval” rhabdomeric units. Similarly, “horseshoe-shaped” rhabdoms can also be identified in the medial retina.

The proximal retinas of E1 and E2, and the medial retina of E1 could be formed by orthogonally-oriented neighboring rhabdomeres (Mandapaka et al., 2006). We further investigated this notion in larval Thermonectus marmoratus to determine if E1 and E2 could meditate polarization vision. If so, we would expect to find certain anatomical features (adapted from (Waterman, 1975)): 1) regularly-structured rhabdomeres with approximately parallel microvilli, 2) at least two cells that have approximately orthogonally-oriented microvilli, and 3) consistent microvillar orientations throughout the length of all rhabdoms. The latter requires rhabdoms to be relatively straight.

Materials & Methods

Animal care

Adult Thermonectus marmoratus were initially provided by the Insectarium of the Cincinnati

Zoo & Botanical Garden, and later collected near Tucson, AZ (USA) between 2004 and 2008.

Beetles were kept in a fresh water aquarium provided with moist terrestrial oviposition sites, maintained on a 16h:8h light:dark photoperiod, and fed freshly-killed or frozen-thawed crickets.

Hatchling larvae were reared in isolation (to limit cannibalism) in 240ml opaque plastic cups at

28°C, and fed live mosquito larvae and frozen-thawed blood worms. The study was carried out exclusively on first instar larvae because their retinal anatomy is best described among the three larval stages (Mandapaka et al., 2006).

Ethyl gallate staining

This procedure was modified from (Strausfeld and Seyan, 1985). First instar larvae were anesthetized on ice and subsequently decapitated. Tissue was fixed in a solution of 4%

Paraformaldehyde (Electron Microscopy Sciences) in Sørensen’s phosphate buffer at pH 7.4

(Electron Microscopy Sciences). The tissue was washed twice in buffer, then postfixed in 1%

osmium tetroxide (OsO4) solution (Electron Microscopy Sciences), and held for 2h at 4 °C followed by 2h at room temperature. Thereafter the tissue was thoroughly washed and transferred to a cold saturated ethyl gallate (Fluka, Buchs, Switzerland) solution, again for 2h at

4°C and 1h at room temperature. Tissue was washed again and dehydrated through a graded series of alcohol and acetone, and embedded in Ultra-Low Viscosity Embedding Media

(Polysciences, Warrington, PA, USA). Ethyl gallate stained sections were previously used for 3D

reconstructions of the primary larval eyes E1 and E2 (Mandapaka et al., 2006). E1 is re- illustrated here to show its basic anatomical organization (Figure 1b).

Tissue preparation and TEM

First instar larvae were anaesthetized on ice and subsequently decapitated. The head was parted medially and fixed in 2.5% glutaraldehyde and 4% formaldehyde in Sørensen’s Phosphate

Buffer (Electron Microscopy Sciences) for 14-18h at 4°C. The head halves were postfixed in 1%

osmium tetroxide (OsO4) solution (Electron Microscopy Sciences), dehydrated and embedded in ultra-low viscosity medium (Electron Microscopy Sciences) at an upward angle of 35° to ensure transverse sections through the primary eyes (Mandapaka et al., 2006b). All sections were prepared using an Ultracut E Microtome (Reichert-Jung). A histo diamond knife (size 8,

Diatome-U.S.) was used to cut each of the semithin sections (thickness: 1µm) and was then substituted with an ultra35° diamond knife (size 1.5, Diatome-U.S.) to acquire each of the ultra- thin sections (thickness was adjusted between 50µm and 70µm). Sections were taken at a knife angle of 6°.

The specimens were sectioned in series to obtain sequential samples at intervals of either

10 µm (1 specimen each) or 20 µm (2 specimens each) along the length of the entire eye, starting with the onset of the medial retina. Five to ten ultra-thin sections were taken at specified levels, followed by ten or twenty semi-thin sections (to reach the next sample interval). The last two semi-thin sections of each interval served as a reference for the section level.

For the measurement of the angular distribution of microvilli in the proximal retina of E1 and E2, the onset of the proximal retina is defined as “0µm”. Since samples were taken at

intervals of 10µm or 20µm, all subsequent section levels are identified by their distance to the onset of the proximal retina: “+10µm”, “+20µm”, etc.

The sections were placed onto Formvar and Carbon-coated copper grids (150 mesh,

Electron Microscopy Sciences) and stained with 2% uranyl acetate for 15 minutes, followed by

7.5 minutes in Reynold’s lead citrate. The sections were examined with a Transmission Electron

Microscope (JOEL JEM-1230). Digital images were taken with a Megaplus ES 4.0 camera.

Tissue analysis

Contrast and brightness of the TEM images were adjusted in Adobe Photoshop CS2. Cell characteristics, such as the orientation of microvilli, were measured using ImageJ software

(NIH). For each photoreceptor cell type of the proximal retina, and at each sample level, five cells were chosen at random. Most cells originated from different sections, but if they were obtained from the same section, non-adjacent cells were chosen. Quantified cell characteristics are: 1) width of the photoreceptor cell at the distal tip of its rhabdomere, 2) length of rhabdomere

(long axis of microvilli), 3) width of rhabdomere, 4) total number of microvilli, and 5) microvillar angular distribution. To determine the latter, a straight line representing “horizontal”

(0°) was positioned in the cleft between the two retinal layers visible in frontal sections of the proximal retina (see “horizontal cleft” in Figure 4b). This reference line was used to measure angles from all microvilli of subject photoreceptor cells, or in case of large rhabdomeres

(photoreceptor cell type 2), 21 evenly-distributed microvilli were included. Results from three specimens of each E1 and E2 were pooled. Averages and standard deviations were computed using Microsoft Excel 2003. Angular distributions of microvilli were statistically analyzed using the Mann-Whitney Rank Sum Test in SigmaPlot for Windows (Version 11.0).

In some of our samples we observed deformations of cells (data not shown). More specifically, the rhabdomeres of these cells were bent to one side or split into two parts. This phenomenon seemed to be locally restricted and usually affected only one or a few neighboring cells. In other samples the microvilli of the proximal retina appeared somewhat contorted instead of straight and parallel. However, we believe that these deformations were preparation artifacts and therefore affected cells were not included in the analysis.

Mosquito picture

Mosquito larvae are likely among the most common natural prey of Thermonectus marmoratus larvae (Morgan, 1992b). To visualize the extent of polarization contrast in a natural prey, larvae of Aedes aegypti were imaged against a background of linearly polarized light. Images were taken using an Olympus BKH022083 camera on an Olympus BX51 microscope with a polarization analyzer (Figure 5) in cross-polarization configuration.

Results

In larval Thermonectus marmoratus, each of the two principal eyes (E1 and E2) is composed of multiple retinas (for detailed descriptions see Mandapaka et al. (2006). To provide an overview,

E1 is illustrated here with a 3D reconstruction of its core (Figure 1b) and several associated sections (Figure 1b-e). The outer portion of the eye consists of a thick, round lens (Figure 1b,c) and a crystalline cone-like structure comprised of loose tissue. Directly underneath lies the distal retina (DRe; Figure 1b&d), followed by the proximal retina (Pre; Figure 1d&e). Exclusive to E1 is the medial retina (MRe, Figure 1b,c) situated adjacent to the crystalline cone-like structure.

We used criteria adapted from Waterman (1975), outlined in the last paragraph of the

Introduction, to determine if any of the multiple retinas found in E1 or E2 could potentially be polarization sensitive.

Medial retina (MRe)

The medial retina is only present in E1, and consists of a narrow band that begins beneath the lens and continues medially alongside the crystalline cone-like structure (Figure 1b,c). The medial retina is roughly oval in shape when viewed in a frontal (cross) section (Figure 1a), and has two nested layers of cells, here called the central and peripheral regions. The central region consists of a small number of cell bodies which are located dorsally and ventrally, and extend their rhabdomeric portions mostly centrally. Their cell bodies enclose the peripheral region.

In an ethyl-gallate stained frontal section, the rhabdomeric regions of the central photoreceptor cells are visible directly adjacent to the crystalline cone-like region of the eye

(Figure 1a). These cells appear relatively dark and from electron microscopical images we know

that they are densely filled with screening pigment. Electron microscopical images of the microvilli of these central photoreceptor cells reveal that they form a diffuse, rhabdomeric aggregation (Figure 1b). Their microvillar orientation is highly variable, perhaps with a slightly prevalent dorsoventral orientation. Within the

Figure 2 Medial retina (MRe). a ethyl gallate-stained frontal section of E1, showing the crystalline cone-like structure (CC) and the different regions of the medial retina: retinula cells of the central region (CMRC), rhabdoms of the central (diffuse) region (CMRh), retinula cells of the peripheral region (PMRC), rhabdoms of the (ordered) peripheral region (PMRh). b TEM image of the central area of medial retina, frontal section. c TEM image of the medial retina, frontal section, peripheral area with the two different retinal cells types (PMT1 and PMT2) with organized rhabdomeres. Zonula adherens (ZA), dorsal (d), lateral (l), medial (m), ventral (v).

rhabdomeric region, no clear borders between neighboring cells of that type are visible.

This is in contrast to the peripheral region of the medial retina, which is composed of photoreceptor cells that stain slightly lighter in ethyl gallate (Figure 2a). Our TEM data suggest that these photoreceptor cells possess more ordered rhabdomeres, with relatively shorter and narrower microvilli, all of which are aligned approximately parallel for each cell. Based on shape and alignment of the rhabdomeres, we identified two types of photoreceptor cells that alternate in sequence. We call these peripheral medial retina cell type 1 (PMT) and 2 (PMT2; Figure 2c).

The rhabdomeres of these neighboring cells are oriented roughly perpendicular to one another

(Figure 2c, compare microvilli of cells PMT1 and PMT2).

PMT1 cells are relatively narrow and tapered at the rhabdomeric end. Their microvilli are oriented approximately dorsoventrally in all of our samples. At this point it is unclear if this orientation is consistent at all levels of the very long medial retina. At the tip of the rhabdomeric region of the photoreceptor cell, the microvilli are somewhat longer than those closer to the cell body.

PMT2 cells are slightly wider than PMT1 cells and their rhabdomeric region is divided into two portions, one at each side of the cell. The cell’s most apical surface does not form microvilli. The microvilli of PMT2 cells are all nearly equally long and oriented approximately horizontally in all of our samples.

Notably, the rhabdomeres of PMT1 cells are longer (long axis of the microvilli) than those of PMT2 cells. Conversely, the rhabdomeres of PMT2 cells contain relatively more microvilli. The microvilli of both cell types appear similar in width. In Figure 2c the zonula adherens is visible where the PMT1 cell rhabdomer begins. At each cross section the medial retinal is relatively small, only consisting of a few cells of each type.

Distal retina (DRe)

The distal retina (Figure 1b,d) is present in both E1 and E2. It is situated proximal to the crystalline cone-like structure and is formed by at least 12 tiers of receptor cells (Figure 3b).

These cell bodies are long and narrow, and apart from the most proximal cells, are oriented more-or-less perpendicular to the long axis of the eye. These cells contain screening pigment which is densest close to the rhabdom. The more proximally positioned cells within the distal

retina are somewhat tilted caudally, such that their outer portions extend into the region occupied mainly by the proximal retina (Figure 1d).

Centrally positioned, directly below the crystalline cone-like structure, the photoreceptor cells extend their microvilli into a region called the rhabdomeric aggregation (Mandapaka et al.,

2006), which contains numerous rhabdomeres (Figure 3b) and absorbs relatively little ethyl- gallate stain (Figure 1d). The rhabdomeric region of the distal retina appears roughly triangular in a sagittal section (Figure 1d & 3b) and consists almost entirely of extremely long and relatively thick microvilli. In some sections, minor protrusions of the cell bodies appear to extend into the rhabdomeric aggregation, somewhat resembling a tree trunk with the branching microvilli on the sides and top (see frontal section in Figure 3a). This tree-like organization results in individual microvilli of the same cell being nearly orthogonally oriented close to the cell body. The rhabdomeres of the dorsal and ventral regions of the distal retina extend all the way to the middle, where they converge. These microvilli intertwine in a dense meshwork and individual cells are largely indistinguishable from one another. Although the orientation of microvilli is highly variable, a dorsoventral orientation appears predominant (Figure 3b).

Figure 3 Distal retina. a TEM image, frontal section, showing the tree-like branching of the microvilli and the extended cell body (“trunk”). dorsal (d), lateral (l), medial (m), ventral (v). b Composite TEM image of the distal retina, sagittal section, showing the retinula cell bodies (DRC), and the distal rhabdoms (DRh). Crystalline cone-like structure (CC), proximal retina (PRe).

Proximal retina (PRe)

The proximal retina envelops the cup-shaped proximal end of the distal retina (Figure 1b,d; see

Mandapaka et al. (2006) for details on the anatomy). The proximal retina is present in both E1 and E2 (Figure 1b,d& e) but is slightly shorter in E1 (120µm) than in E2 (130µm). It consists of a dorsal and ventral portion (Figure 1e). These cells are narrow horizontally, but extend both dorsoventrally and in the proximal/distal direction, such that their longest axis follows the main axis of the eye. The dorsal and ventral portions are separated by a very narrow horizontal cleft that is filled with screening pigment. Our TEM data show that each portion is composed of photoreceptor cells with ordered rhabdomeres, and that microvilli within each cell are largely parallel to one another (Figure 4). Our frontal sections of the proximal retina at its widest point show approximately 150 cells in each of the two portions, with nuclei located dorsally and ventrally near the base and periphery.

Based on cell morphology and rhabdomere orientation, three different cell types can be distinguished (Figure 4b). We call these proximal retina cell type 1-3 (PT1-3). Near the base of the rhabdomeric regions of PT2 cells, adjacent photoreceptor cells are linked to one another by short zonulae adherens (Figure 4a). Horizontally, the cells alternate in a repeated ordered sequence as follows: *PT1-PT2-PT3-PT2* (Figure 4b). Hence, PT1 cells as well as PT3 cells are always surrounded by PT2 cells. Since PT1 and PT3 cells generally do not extend to the horizontal cleft, PT2 cells tend to join at their apical end. The dorsal and ventral portions of the proximal retina are mirror-images of one another.

Figure 4 Proximal retina (PRe), frontal sections showing retinal cell types. a TEM image. Zonula adherens (ZA). b schematic of a. proximal retina cell type 1 (PT1), proximal retina cell type 2 (PT2), proximal retina cell type 3 (PT3).

Consistency of microvillar organization

Since consistent microvillar organization is an important criterion for polarization sensitivity

(Waterman 1975), we systematically determined the angular distribution of the microvilli for each of the three cell types using TEM data. For these measurements, the horizontal cleft between the ventral and dorsal portions was used as a reference line for all angular specifications, with horizontal corresponding to 0°.

Cell type 1 (PT1) is about 1.4µm wide at the onset of the rhabdomere. The rhabdomere of

PT1 is oval in shape in a frontal section, approximately 2.0µm long (long axis of the microvilli), measures about 1.8µm at its widest portion, and consists of 23 ± 5 microvilli per section.

Microvilli are oriented approximately dorsoventrally (92.4° ± 7.6° in E1, 89.8° ± 9.8° in E2,

Table 1).

Cell type 2 (PT2) is approximately 1.7µm wide at the level of the onset of the rhabdomere and the apical end of the cell appears roughly rectangular in frontal sections. The rhabdomere is located bilaterally, while the most apical piece of membrane (situated at the cleft) does not possess microvilli. The rhabdomeric portion of PT2 neighboring PT1 cells is about

0.5µm long and 3.7µm wide, and consists of about 52 ± 10 microvilli per section. The rhabdomeric region facing PT3 cells is shorter and stouter, with an average length of 0.3µm and a width of 1.9µm, consisting of 24 ± 13 microvilli per section. The microvilli of PT2 cells are oriented more or less horizontally (13.7° ± 5.1° in E1, 15.2° ± 5.1° in E2, Table 1).

Cell type 3 (PT3) is the slimmest of the three with a width of only 1.0µm at the base of the rhabdomere. It shows some similarities to PT1 since its microvilli have the same orientation

(Table1), though it is smaller. The rhabdomeric region of PT3 is cylindrical in frontal view, only about 1.1µm long, 0.9µm wide. It consists of 12 ± 2 dorsoventrally oriented microvilli per section (92.4° ± 11.0° in E1, 90.4° ± 10.6° in E2, Table 1).

This organization is consistent at all levels of depth (Table 1), and throughout both the ventral and dorsal portions of photoreceptor cells. The only exception is the outermost region

(most medial and lateral), where the cell bodies and rhabdomeres bend sideways slightly as the cleft between the portions narrows. Each rhabdomere of PT1 and PT3 is enclosed by the immediate neighboring lateral rhabdomeric regions of the two adjacent PT2 cells, leading to alternating larger and smaller rhabdoms. Mandapaka et al. (2006) characterized these two seemingly structural units as “horseshoe-shaped” (PT2-PT1-PT2) and “oval rhabdoms” (PT2-

PT3-PT2), respectively. Our findings show that individual PT2 cells contribute to both of these

light-microscopically identified units, suggesting that the latter are not independent in regards to their physiology.

Table 1 Angular distribution of the microvilli of proximal retina cell type 1 (PT1), cell type 2 (PT2) and cell type 3 (PT3) of Eye1 (E1) and Eye2 (E2). 0° corresponds to horizontal (mediolateral) microvillar alignment, and 90° corresponds to vertical (dorsoventral) microvillar alignment. The rhabdoms of PT2 are oriented approximately horizontally, whereas the rhabdoms of PT1 and PT3 are oriented predominantly vertically (E1: p=0.737*, E2: p=0.945^).

E1 E2

PT1 PT2 PT3 PT1 PT2 PT3

0 µm 88.8 ± 21.7 34.3 ± 21.4 90.7 ± 2.1 93.9 ± 18.4 18.8 ± 14.4 94.6 ± 9.2

+10µm 86.0 ± 20.4 16.8 ± 10.7 89.5 ± 11.0 101.4 ± 3.4 12.5 ± 10.2 90.6 ± 17.7

+20µm 92.8 ± 19.7 17.8 ± 13.2 93.1 ± 4.4 92.7 ± 13.2 12.1 ± 11.1 91.0 ± 7.6

+30µm 93.4 ± 8.7 12.4 ± 9.5 89.3 ± 6.3 89.5 ± 13.9 12.6 ± 9.7 82.5 ± 2.8

+40µm 87.4 ± 8.5 15.3 ± 10.7 86.1 ± 7.8 93.2 ± 11.1 14.7 ± 11.4 87.2 ± 5.9

+50µm 92.2 ± 11.9 14.3 ± 10.2 90.7 ± 7.0 88.4 ± 6.1 16.5 ± 12.5 91.1 ± 9.4

+60µm 94.1 ± 13.3 15.5 ± 10.9 89.7 ± 6.8 91.7 ± 12.8 12.6 ± 11.4 88.1 ± 10.4

+70µm 92.4 ± 13.5 14.8 ± 11.2 94.6 ± 14.7 93.7 ± 9.0 10.3 ± 8.6 93.4 ± 9.0

+80µm 85.5 ± 13.3 14.5 ± 10.5 92.4 ± 11.0 93.3 ± 6.9 11.6 ± 10.1 88.9 ± 8.9

+90µm 84.3 ± 14.8 12.7 ± 9.1 91.1 ± 10.7 89.1 ± 3.9 10.5 ± 7.4 92.1 ± 6.1

+100µm 86.6 ± 15.3 14.8 ± 12.3 94.4 ± 22.0 94.2 ± 17.9 12.4 ± 9.7 99.3 ± 16.5

+110µm 82.4 ± 14.1 14.7 ± 0.9 84.9 ± 10.4 88.2 ± 19.5 12.7 ± 8.8 108.4 ± 11.0

+120µm 97.0 ± 20.6 16.6 ± 11.7 88.1 ± 12.1 88.0 ± 25.2 21.6 ± 16.9 100.3 ± 12.6

+130µm 94.7 ± 17.0 12.2 ± 10.3 86.0 ± 22.0 - - -

Average

angular 92.4 ± 7.6 * 12.2 ± 10.3 92.4 ± 11.0 * 89.8 ± 9.8 ^ 15.2 ± 5.1 90.4 ± 10.6 ^ distribution

Discussion

We studied the rhabdomeric organization of the different retinas within the principal eyes (E1 and E2) of first instar larval Sunburst diving beetles, Thermonectus marmoratus. Specifically, we investigated the possibility that the basic anatomy and ultrastructure could support polarization sensitivity in one or more of these retinas.

1) Potential for polarization sensitivity in the proximal and medial retinas of the principal eyes

The proximal retinas (in both E1 and E2) and the peripheral portion of the medial retina

(present only in E1) all show structural organizations that are consistent with polarization sensitivity. The proximal retinas are especially strong candidates fully meeting all criteria; since they 1) are composed of rhabdomeric photoreceptor cells possessing parallel microvilli, 2) have microvilli oriented approximately perpendicular to those in adjacent cells, and 3) do not undergo a twist anywhere along their entire length, thus have consistent microvillar orientation.

The proximal retinas are comprised of three morphologically-distinct cell types, with two orthogonal microvillar orientations. Photoreceptors PT1 and PT3 exhibit a similar, dorsoventral microvillar alignment. Each rhabdomere of PT1 and PT3 cells is enclosed by parts of the rhabdomeres of two adjacent PT2 cells, the microvilli of which are aligned in horizontal direction. Hence, there are two sets of almost orthogonal rhabdomeres in the proximal retina, with the preferred orientation of vertical (PT1+PT3) and horizontal (PT2), in reference to the eye tube. Since there are about twice as many PT2 cells as PT1 and PT3, there are approximately the same number of cells for each microvillar orientation.

Similarly, the peripheral region of the medial retina has at least two of the three anatomical traits needed for polarization sensitivity: 1) receptor cells with parallel microvilli, and

2) adjacent rhabdomeres oriented orthogonally to one another. In all of our samples, these two sets of photoreceptors had rhabdomeres exhibiting either vertical (PMT1), and horizontal

(PMT2) alignment. We currently do not have sufficient data to confirm rhabdomeric consistency in all regions of the medial retina.

Recent data on opsin expression in larval Thermonectus marmoratus are also consistent with the possibility of polarization sensitivity in these regions. (Maksimovic et al., 2009) found that both proximal retinas and the peripheral region of the medial retina express only UV opsins, while the central region of the medial retina, and the distal retina express a long-wavelength opsin. Generally, sets of orthogonally-oriented photoreceptors in a polarization-sensitive visual system should have the same spectral sensitivity to avoid confusion with chromatic stimuli

(Wehner and Labhart, 2006). Moreover, polarization sensitivity in other insects often is mediated by UV sensitive receptor cells (Horváth and Varjú, 2004), additionally supporting the notion that these regions in Thermonectus marmoratus could be specifically sensitive to the polarization of light.

As previously mentioned, and described in detail in Buschbeck et al. (2007),

Thermonectus marmoratus larvae perform scanning movements during prey capture. This could potentially alter the polarization information that each cell receives. However, scanning takes place dorsoventrally, and as the microvilli tend to be oriented horizontally or vertically, scanning movements would result in a translation of microvilli rather than a rotation. Hence, scanning

should minimally influence the polarization sensitivity of those cells. While our ultrastructural data suggests that the stimulation strength of individual cells within these retinas is likely e- vector dependent, it remains unclear to what degree polarized light might be altered prior to reaching these receptor cells. As described above, the proximal retina is overlain with the somewhat diffuse rhabdomeric tissue of the distal retina. Therefore, the proximal retina only receives light that passes the distal retina (Figure 1d). Light also tends to pass the diffuse central region of the medial retina before entering its peripheral region. The only exception could be the most distal portion of the peripheral medial retina, which potentially could receive some light from outside the lens (Figure 1c). Interestingly, the tree-like organization of the DRe results in some microvilli being oriented obliquely, roughly in-between the horizontal and vertical orientations of the proximal retina. In principle the orientation of microvilli can alter the nature of polarized light. For example it has been demonstrated in species of mantis shrimp

(Odontodactylus) that the precisely aligned microvilli of some R8 cells can change circularly polarized light to linearly polarized light (Chiou et al. 2008). However, the organization of the

DRe is all but perfect, and its microvilli appear fairly diffuse, tending towards a dorsoventral orientation (as in the central region of the MRe). This might diffuse the e-vector pattern, or potentially even lead to a bias in the prevailing e-vector orientation. Physiological measurements will be necessary to establish if these cells do in fact respond differentially to different e-vector orientation.

2) The distal retinas and the central region of the medial retina are unlikely to mediate polarization sensitivity

Despite a prevailing dorsoventral orientation of the rhabdomeres in the distal retina, the alignment of the microvilli here do not meet essential criteria, making them unlikely candidates for polarization sensitivity. Most notably, the microvilli of individual cells are not oriented parallel, but rather approximately orthogonal to one another (reflecting the tree-like organization of these cells). Additionally, the microvillar organization of individual cells appears to be highly variable, with neighboring cells organized similarly. Furthermore, the rhabdoms of the central region of the medial retina are composed of microvilli with varying though predominantly dorsoventral orientation, again with neighboring cells similarly organized. Such ultrastructural organization is highly unlikely to support polarization sensitivity.

3) The ecological importance of polarization sensitivity

A number of aquatic insects, including hemipterans (Schwind, 1984, 1991) and chironomids (Lerner et al., 2008; Schwind, 1991) have been demonstrated to use polarized sensitivity to locate the surface of water bodies. Largely undisturbed water surfaces reflect linearly polarized light, and the latter is especially strong in the UV range of the spectrum

(Horváth, 1995). Thermonectus marmoratus larvae also have to find water. They hatch from eggs laid above the water surface within moist substrate (Morgan, 1992, 1995). It is imperative that hatchlings quickly find, crawl to and enter the water before they desiccate. Although their principal eyes are pointing upward (Mandapaka et al., 2006), crawling through uneven terrain may allow them to look sufficiently downward to detect the predominantly horizontally polarized light (Horváth, 1995).

Several marine invertebrates are also known to exploit polarized stimuli within the water.

For example, polarization signals for intraspecific communication are apparently used by various cephalopods (Boal et al., 2004; Mäthger and Hanlon, 2006; Shashar and Hanlon, 1997) and stomatopod crustaceans (Marshall et al., 1999). However, it is unlikely that a similar application would benefit Thermonectus marmoratus larvae. Sky polarization patters that would be visible in

Snell’s window (Wehner, 2001) are also unlikely to be relevant to Thermonectus marmoratus larvae. Their principal eyes tend to point primarily horizontally during swimming, and the scanning angles of the head are rarely steep enough (Buschbeck et al., 2007) to allow a view into

Snell’s window. More relevant might be the light environment outside of Snell’s window where the scattering of light creates a background underwater that is predominantly horizontally polarized (Wehner, 2001). The degree of polarization depends on the position of the sun, but has been shown to reach up to 50% in fresh water ponds (Cronin, 2006), or even 67% in lakes

(during dawn or dusk, Novales-Flamarique and Hawryshyn (1997)). If Thermonectus marmoratus larvae are in fact polarization sensitive, the most likely benefit to them would be enhanced prey detection. Polarization sensitivity can be advantageous underwater since it can dramatically improve contrast between an object and its background (Lythgoe and Hemmings,

1967; Wehner and Labhart, 2006). Contrast-enhancing polarization signals as a tool for predation are utilized by cuttlefish (Shashar et al., 2000) and squid (Shashar et al., 1998). As described in

(Shashar et al., 1998), prey camouflage such as transparency can be broken through the availability of polarization features, which aids in prey detection.

Mosquito larvae are a common and preferred prey of Thermonectus marmoratus larvae

(Morgan, 1992, 1995). Although it remains unclear what the beetle larvae actually see, especially given their unorthodox eye organization, it is possible to visualize to what extent certain tissues

of mosquito larvae provide polarization characteristics. Figure 5 illustrates that the mosquito’s muscle tissue shows distinctive polarization features against a polarized background.

It previously has been demonstrated that muscle tissue is optically-active (Johnsen, 2005;

Tuchin, 2007; Wehner and Labhart, 2006). This is particularly important in relatively small mosquito larvae, since they are fairly transparent. In addition, it is conceivable that the cuticle of some of their prey reflects light with a different e-vector orientation than that of the background illumination. At least in the context of communication among conspecifics, polarized reflection from cuticle is well known in the mantis shrimp (Marshall et al., 1999).

Figure 5 Visualization of polarization contrast in the larva of mosquito Aedes aegypti (4 mm body length). The mosquito is illuminated from behind with linearly polarized light. Visualization with a polarization analyzer that is oriented orthogonally to the background polarization (crossed analyzer).

Taken together it becomes clear that polarization sensitivity could have substantial benefits for Thermonectus marmoratus and other species of aquatic, predatory larvae. We suggest that if Thermonectus marmoratus larvae were to be sensitive to the polarization of light,

one of the key advantages could be improved detection of prey. We find that both gross morphological and ultrastructural features are consistent with polarization sensitivity, at least within the proximal, and possibly the peripheral portion of the medial retina. However, at this point it is still unclear if the larvae are in fact sensitive to the polarization of light, or if they take advantage of available polarization features. Physiological and behavioral experiments need to follow to determine if this actually is the case.

Chapter 2 The influence of polarized light illumination on the hunting success in the first instar larvae of the Sunburst Diving Beetle, Thermonectus marmoratus (Insecta, Dytiscidae).

Abstract

The Sunburst Diving Beetle, Thermonectus marmoratus, is an aquatic, visually guided predator in all of its larval stages. The larvae prey on small soft-bodied invertebrates such as mosquito larvae, which may be difficult to detect. One way to improve prey detection is to enhance visual contrast by taking advantage of polarization features. Anatomical studies on the eyes of

Thermonectus marmoratus first-instar larvae have shown that polarization sensitivity could potentially be present in some of the many retinas of the principal eyes, which are involved in prey capture. To test if this indeed is the case, we measured hunting performance under polarized and unpolarized illumination. Using our experimental setup, we found that with comparable foraging activity in either state of illumination, hunting in a polarized environment appears to shorten the amount of time to prey detection and improve capture success significantly.

However, it is unclear if these behavioral differences can be directly attributed to the state of polarization of the illumination, as our post-hoc light measurements reveal other potentially influential factors, such as light intensity differences. The detection of polarization features as a means of enhancing the visibility of potential prey items has only been documented in cephalopods. Here we suggest that hunting-related contrast enhancement might also be important in an insect larva.

Introduction

Perhaps one of the greatest challenges of any aquatic predator is to find sufficient prey.

Camouflage, small size and the effect of the light in the water column can make the visual detection of prey difficult. For example, many plankton species are transparent and thus provide poor contrast against their background (Sabbah et al., 2006; Shashar et al., 1998). However, this camouflage can be broken. Specifically to a predator with polarization-sensitive eyes, a transparent exoskeleton reveals intrinsically polarization-active underlying tissue such as muscles (Johnsen, 2005; Sabbah and Shashar, 2006; Shashar et al., 1998). Rhabdomeric visual systems, as are found in many invertebrates, often are inherently sensitive to polarized light

(Goldsmith, 1977; Waterman, 1975). This potentially could be used to exploit polarization cues in order to improve the contrast of prey against its background (Wehner and Labhart, 2006). This mechanism has been revealed through behavioral experiments with squid (Shashar et al., 1998), but thus far has never been demonstrated for an insect. Based on anatomical studies of their visual system, the first instar larva of the sunburst diving beetle, Thermonectus marmoratus, is expected to be sensitive to the polarization of light (Stecher et al., 2010).

Thermonectus marmoratus is a species of the predaceous Dytiscidae (Insecta) that can be found in the temperate and tropical zones of North America, where it inhabits clear water pools or slow moving seasonal streams with rocky bottoms and sparse vegetation (Evans, 2006). Adult beetles are preferential scavengers (Velasco and Millan, 1998), but their three larval instars tend to prey on soft bodied invertebrates including the larvae of mosquitoes (Morgan, 1992, 1995).

The eyes of the beetle larvae are quite unusual and have been described in detail for the first

larval instar (Maksimovic et al., 2009; Mandapaka et al., 2006; Stecher et al., 2010). There are a total of six larval eyes (stemmata) on either side of the head, and an additional area of photosensitive tissue is situated medially to the other eyes (eye patch). The forward-pointing principal eyes Eye 1 (E1) and Eye 2 (E2) are utilized in prey approach (Buschbeck et al., 2007).

Prey is frequently detected by the smaller lateral eyes first. The larva then orients its head so that the principal eyes are directed towards the prey. The larva slowly swims closer to the target while performing up-and-down scanning movements. Once they reach a certain distance to the prey, the larvae strike with one rapid movement (Buschbeck et al., 2007). As is the case for all their eyes, E1 and E2 each have two anatomically distinct retinas, the distal retina (DRe) and the proximal retina (PRe), which are stacked atop of one another. In addition, E1 possesses a retinal band, the medial retina (MRe), alongside the tubular crystalline-cone like structure that lies beneath the lens (Maksimovic et al., 2009; Mandapaka et al., 2006).

Anatomical studies of the ultrastructure of these retinas have shown that the PRe of E1 and

E2, and possibly the MRe of E1, could be sensitive to the e-vector orientation (angle or plane) of polarized light (Stecher et al., 2010). In rhabdomeric eyes this sensitivity arises in the retina and typically involves pairs of receptor cells with orthogonal rhabdomeric regions (Nilsson et al.,

1987; Waterman, 1975; Wehner and Labhart, 2006). The PRe of first instar Thermonectus marmoratus (E1 and E2) is divided into a dorsal and a mirror-imaged ventral portion of retinula cells. Each of those consists of three types of anatomically identifiable photoreceptor cells that, based on the ultrastructure of their rhabdomeric region, should be stimulated maximally by linearly polarized light with either a horizontal or a vertical e-vector. A similar arrangement can

be found in the medial retina of E1, except that the MRe has two different photoreceptor cells

(Stecher et al., 2010).

Generally, a number of invertebrates and vertebrates exhibit altered, and potentially pronounced characteristics under polarized conditions. Polarization contrast can be created by light reflection off of body surfaces, e.g., the antennae and telson of stomatopods (Marshall et al.,

1999), and arms, forehead and the region surrounding the eyes in cuttlefish (Shashar et al.,

1996), or through refraction within certain body tissues (birefringence; Wehner and Labhart,

(2006)), as in transparent zooplankton (Shashar et al., 1998). In stomatopods (Marshall et al.,

1999) and cuttlefish (Shashar et al., 1996), the former mechanism enables species-specific communication (Cronin et al., 2003; Marshall et al., 1999; Shashar et al., 1996). The animals are thought to convey some message with their display of specific body parts but its exact nature is still unclear. It is probably involved in reproductive behavior and/or staking out an individual territory (Boal et al., 2004; Chiou et al., 2008; Marshall et al., 1999). Nevertheless, none of these factors is likely to be applicable to the larvae of Thermonectus marmoratus because they are neither in a reproductive stage nor are they territorial (personal observation). They are, however, predaceous.

Many organisms exhibit some sort of camouflage in order to avoid being detected by a predator. A common concealment strategy is transparency (McFall-Ngai, 1990; Sabbah and

Shashar, 2006). Nontheless, being transparent might become a disadvantage because certain tissue types are intrinsically birefringent (polarization-active), which benefits polarization- sensitive predators (Shashar et al., 1998). Similar to many zooplankton species, early stages of

immature mosquitoes are fairly transparent and thus potentially offer distinct contrast-enhancing features under polarized illumination (Figure 6). In the natural habitat of Thermonectus marmoratus larvae, immature mosquitoes are among their preferred prey (Morgan, 1992, 1995).

Insects that are known to be polarization-sensitive oftentimes perceive polarized light in the UV spectrum (Horváth and Varjú, 2004). Interestingly, Maksimovic et al. (2009) found that the proximal retinas in the stemmata of the first instar larvae of Thermonectus marmoratus express two types of UV opsins, TmUVI and TmUVII. Maksimovic et al. (in press) proceeded to determine the spectral sensitivity of the Tm opsins in third instar Thermonectus marmoratus larvae. TmUVII photon absorbance peaks at 360nm and has a half-width of 75 nm. In the third larval stage of Thermonectus marmoratus, TmUVI is not expressed in the proximal retina. Based on this anatomical evidence of polarization sensitivity in first instar larvae of Thermonectus marmoratus, we now addressed the question as to what purpose this visual feature would be likely to serve. Stecher et al. (2010) suggested that the first instar larvae of Thermonectus marmoratus could use polarization cues to enhance the detection of prey. In this study we used a behavioral test to investigate if this proposed polarization sensitivity is indeed a factor in successful prey detection.

If the larvae are polarization-sensitive, and this ability is employed in feeding, they would be more successful predators when prey items with polarization contrast are available (polarized illumination) when compared to the same type of prey that is less conspicuous against its background (unpolarized illumination). Therefore, we predicted that the larvae would detect and catch prey in a shorter amount of time and with higher acuity under polarized illumination than under unpolarized illumination.

Figure 6 Aedes aegypti (4mm body length) imaged against a linearly polarized background. a viewed without a polarization analyzer. b viewed with a polarization analyzer. Scale bars 1mm.

Materials and Methods

Animal care

Adult Thermonectus marmoratus were initially provided by the Insectarium of the Cincinnati

Zoo and Botanical Garden, and also repeatedly collected near Tucson, AZ (USA) between 2004 and 2008. Beetles were kept in a fresh water aquarium provided with moist terrestrial oviposition sites, maintained at 28°C on a 16h:8h light:dark photoperiod, and fed freshly-killed or frozen crickets. Hatchling larvae were reared in isolation in 8oz opaque plastic cups under the same maintenance conditions and fed live mosquito larvae and thawed frozen blood worms. The study was carried out exclusively on first instar larvae because of their high abundance in our cultures.

To ensure behavioral tests with similar hunger states of the test specimen, larvae were well fed

48 hours before the test day, then starved for a period of 24 hours prior to the trials.

Analysis of polarization features in mosquito larvae

Mosquito larvae of Aedes aegypti were imaged against a background of horizontally, linearly polarized light. Images were taken with an Olympus BKH022083 camera on an Olympus BX51 microscope (a) without polarization analyzer (Figure 6a), and (b) with a polarization analyzer

(Figure 6b) that was rotated to obtain the highest possible contrast, which was achieved at about

90°. In the former case, light intensities were adjusted to match the decrease in light intensity resulting from introducing the polarization filter into the path of light.

These two setups were chosen to represent (a) a visual system that is not polarization sensitive (no polarization analyzer, Figure 6a) and (b) a visual system that is capable of detecting different e-vectors (with polarization analyzer, Figure 6b). However, this setup is highly artificial and exaggerates natural conditions, as 100% polarized light does not occur in natural habitats.

Therefore, Figure 6b is a representation abstracted in order to make it visible to the human eye, but it is unlikely to be the actual appearance of a mosquito larva to a Thermonectus marmoratus larva (or any other polarization-sensitive animal). Nevertheless, Figure 6b provides evidence that mosquito larvae exhibit pronounced features under polarized illumination conditions, a circumstance which potentially alters the contrast between the mosquito larva and its surroundings.

Test chamber

The test arena was modified from an opaque 8 oz. plastic cup which was taller but otherwise identical to those in which individuals were reared in the laboratory. The cup, which was off- white in color, was cut in half length-wise and the front half was replaced with a 1mm-thick

glass plate (Bio-Rad). The edges were sealed with silicone to prevent water leakage. In all trials, the water level in the arena was kept at a depth of 2.5 cm.

Behavioral tests have shown that Thermonectus marmoratus first instars perform rather poorly under UV-only illumination (personal observation). Therefore, a broad-spectrum xenon lamp (Hamamatsu Photonics K.K., Model C2577) served as the light source to translate sunlight- like characteristics to laboratory conditions. To diffuse the light from the xenon lamp, the light was reflected down onto the test area with a reflector built from a rectangular piece of cardboard wrapped in aluminum foil (shiny side outside). Aluminium foil was used because metal surfaces do not polarize the light they reflect, but they can be used to diffuse light. In addition, we chose to illuminate the behavioral arena from the top because previous behavioral experiments have shown that T. marmoratus larvae are more likely to feed under these illumination conditions.

Furthermore, the larvae tend to present their dorsum towards a strong light coming from the side, which results in an unnatural swimming position (personal observation).

We chose a polarizer that transmits UV light as well as light within the visible range

(BVO UV, Bolder Vision, USA). A double layer of ordinary wax paper was used as a depolarizer as it was shown to be a highly effective diffuser (Johnsen, 1994). The depolarizer was cut to the same dimensions as the polarizer (5 x 5 cm) and both filters were combined for simultaneous use. Their position could be switched to create polarized light (POL; polarizer in bottom position) and unpolarized light (unPOL; depolarizer in bottom position), which we expected would lead to light with equal intensity. The rationale for this method was to ensure

that the light in the polarized and unpolarized trials differed only in polarization and not in intensity or spectral distribution (Via and Forward, 1975).

The polarizer-depolarizer sandwich was placed directly above the test arena, with the polarizer always oriented horizontally (transmission axis parallel to the long axis of the test arena). With this position it was possible to limit the test illumination to light that was passing through the depolarizer/polarizer only. Furthermore, any possible inherent polarization in the output of the light source was eliminated by the depolarizer (when in top position).

The completed setup (Figure 7) was left undisturbed during the entire period of behavior testing, with the exception of the test arena. The position of the latter was lined out with marking tape to ensure correct re-positioning after placing the test specimen in the arena. In addition, the polarizer/diffuser could be inverted. All trials were performed during the same period (early afternoon) on all test days. In primary tests it has been shown that the larvae are active during that time (personal observation). The test room was darkened for the experimentation (the lamp of the behavioral setup was the only source of illumination) and was kept at a constant temperature of 24ºC.

Figure 7 Experimental setup used to test hunting performance in Thermonectus marmoratus. The scale in the image does not represent the actual dimensions of the components, except that the filter sandwich completely covered the top surface of the arena, as shown here. The position of the polarizer and diffuser (=depolarizer) was switched to create linearly polarized and unpolarized experimental illumination. For example, the order of the filters displayed in this image would create POL conditions.

Behavior

The illumination condition (POL or unPOL) was randomized using the online Research

Randomizer Form v.4.0 (Social Psychology Network; set of numbers: 1, numbers per set: 12, number range: 1-12). The number of individuals tested per day somewhat depended on their availability in our culture, which underwent frequent fluctuations. The total number did not exceed 10 individuals per day.

Individual larvae were placed into the test arena and acclimatized for five minutes, after which one individual mosquito larva (Aedes aegypti, 3-5mm body length) was introduced and the trial was started. A test trial was terminated after the mosquito larva was caught or after a total of 10 minutes had passed. All tests were recorded with a video camera (Sony HDV

1080i/MiniDV) set down in front of the setup in a way that it was facing the glass front of the

arena (Fig. 2). The position of the camera was outlined on the table with marking tape. The video footage was analyzed with Adobe Premiere Pro CS4.

To score the hunting efficiency of Thermonectus marmoratus first instars, the following criteria were used:

I. Detection latency. This is the time elapsed to the first initiation of the overall strike. It includes attempts that were unsuccessful (Thermonectus marmoratus larva delivered a strike off- target, or the mosquito larva escaped), as well as the successful strikes (the mosquito larva was caught and subsequently eaten). For the analysis, a ‘strike’ is initiated when the Thermonectus marmoratus larva contracts its body in preparation of executing an attack. This definition was chosen because it is most reliable to ensure that the prey was focused with the primary eyes.

Even though Thermonectus marmoratus larvae typically perform characteristic scanning motions when approaching prey (Buschbeck et al., 2007), scanning does not always actually result in a prey attack (personal observation), whereas whole body contractions usually do.

II. Capture latency. Time elapsed to successful strike (the mosquito larva was caught and subsequently eaten).

III. Total number of strikes per trial (includes all successful and unsuccessful strikes as well as “no strike”). The trial duration varied among the specimens since an individual trial was terminated after the prey item was caught, or after a total of 10 minutes of unsuccessful or no attempts. However, since only a single mosquito larva was available as prey, it was unnecessary to extend a trial over the moment of a successful strike.

IV. Probability of capture (the number of successful strikes divided by total number of attacks in an individual trial). The number of successful strikes was either equal to 0 (no success

or no attempts) or 1 (mosquito larva was caught on first attempt) thus probability of capture ranged between 0 and 1 (e.g. if the mosquito larva was caught in a second or third attack).

Data analysis

Descriptive statistics (means, standard deviation and standard error) were calculated and graphs were created in Excel (Windows Office 2003, 2010). All data were analyzed using Oneway

ANOVA models, with the illumination condition (POL, unPOL) as the factor, and times of strike

(overall or successful), number of total strikes, and probability of capture as the respective response. Statistical tests were performed in JMP 8 (SAS Institute Inc., Cary, NC, USA). The data for ‘probability of capture’ were arcsine-transformed for statistical tests, and the data for the criterion ‘total number of strikes’ were square root-transformed.

A total of 108 trials were conducted, with 54 trials for polarized illumination condition

(POL) and 54 for unpolarized illumination (unPOL). A number of larvae under both

illumination conditions did not show any reaction to the prey (nPOL=17, nunPOL=16) and thus were only integrated into the analysis of (III.) 'total number of strikes per trial'.

Analysis of light

Three different absolute irradiance measurements were performed post-hoc to quantify the illumination conditions of the experimental setup. (1) The light intensity of the down-dwelling light was measured under both types of illumination. This was done to gain insight into the actual light intensity of either condition. (2) The second set of measurements was executed to describe the state of polarization of the down-dwelling light with both polarized and unpolarized

illumination. (3) The illumination conditions were determined for the body of water in the experimental setup in which Thermonectus marmoratus larvae were located.

All measurements were done using a miniature fiber optics spectrometer (USB2000+,

Ocean Optics) with a cosine corrector (acceptance angle: 180°), attached via USB 2.0 to a laptop computer with the specific software (Ocean Optics Spectrasuite August 2007). The spectrophotometer was calibrated with a Dh-2000 calibration. The dark spectrum was recorded initially and subsequently automatically deducted from the measurements. The integration time was adjusted for each set of measurements.

The probe (QP600-025-SR EOS-335 69-1, length 21cm, diameter 600µm) was stabilized by inserting it into a large-diameter plastic drinking straw. For measurements (2) and (3), a polarizer sheet (BVO UV, Bolder Vision, USA) was attached to the opening of the straw in a way that the tip of the probe (with the cosine corrector) almost touched the polarizer. Much like the polarizer/diffuser sandwich of the experimental setup, this polarizer was framed by two plastic sheets that had a cut-out in the center to allow light to pass through. This way, the polarizer sheet could be handled without destroying its surface. The other end of the straw was fastened to the bottom end of the probe so that the construction as a whole could be rotated while keeping the position of the polarizer fixed relative to the probe. This was necessary because of possible inherent polarization bias of the device as a whole (T. Cronin, personal communication).

Because of the small size of the test arena, and the vulnerability of the probe to water immersion, it was impracticable to place the probe directly into the water column. Thus, the test arena was rebuilt with the same materials except that a rectangular hole was cut into the center of the

bottom of the cup onto which a glass coverslip was fitted (22x22mm, thickness 1, Daiggerbrand) and sealed with silicone. The modified arena was positioned in the behavior setup in place of the arena used in the behavioral tests.

Three sets of absolute irradiance measurements were collected to characterize the illumination conditions of the behavioral setup. Each reported value was obtained from three repeats of 100 automatically obtained averages.

(1) Overall light intensity. Both the diffuser and especially the polarizer possess a reflective surface that might deflect the initial, incoming light (from the xenon lamp) away from the test arena. It was necessary to ensure that the behavioral variances between different states of polarization in the trials cannot be attributed simply to dissimilarities in light intensity.

Consequently, intensity measurements (integration time: 700ms) were performed from below the arena (through the cut-out region) under polarized (POL) and unpolarized (unPOL) illumination, without a polarization analyzer.

(2) Intensity, e-vector and degree of polarization of the down-dwelling light. To determine the effectiveness of the polarizer versus diffuser in creating a polarized versus an unpolarized test environment, measurements were carried out from below the arena (integration time: 2150ms), with a polarization analyzer. The analyzer was rotated manually to three positions, so that its transmission axis was parallel (horizontal, or 0°), in an angle of 45 degree, and orthogonal (vertical, or 90°) to the polarizer in the behavioral setup. The sequence was repeated twice more. Under POL conditions, a parallel alignment of the polarization analyzer to

the initial polarizer would theoretically result in the highest light intensity value, whereas the least (theoretically no) light would be transmitted through an analyzer in crossed (orthogonal) orientation. Since unpolarized light does not have a preferential e-vector, all light measurements should result in essentially the same value under unPOL settings, regardless of the alignment of the transmission axis of the polarization analyzer to the polarizer.

(3) Intensity, e-vector and degree of polarization of the light near the water surface.

These measurements (integration time: 2150ms) were performed from the front of the test arena, through the anterior glass plate. This window most likely gives access to the visual field as it is presented to the principal eyes of the swimming Thermonectus marmoratus larvae, which typically keep relatively close to the water surface (personal observation). Again, the polarization analyzer was rotated manually in a sequence of a 0°, 45° and 90° position, and the sequence was repeated twice more. The analysis of the light measurements was executed in Excel (Microsoft

Office 2010) and MATLAB fminsearch (MathWorks, version 7.0.1). Using the absolute irradiance measurements, e-vector and degree of polarization were calculated in Excel (see equations below). Some of the measurements resulted in a negative value. In order to adjust for this potential calibration error, and to gain positive numbers only, the lowest value of a given data set was identified and then added to all averages (POL and unPOL) of that set of measurements, which resulted in zero as the lowest possible value.

To calculate the phase (e-vector) and degree of polarization, the following formulae were used

(after Sabbah et al. (2006)).

degree of polarization p

( ) + (2 ) = 100 2 + 2 � 퐼0 − 퐼90 퐼45 − 퐼90 − 퐼0 푝 0 90 퐼 퐼 phase Ɵ

1 + 2 = tan 2 −1 퐼0 퐼90 − 퐼45 Ɵ � 90 0 � then, if < (if < ) = + 90°, 퐼 otherwise− 퐼 = 90°

90 0 45 0 퐼 퐼 퐼 퐼 Ɵ Ɵ Ɵ Ɵ −

I0, I45, and I90 are the absolute irradiance measurements with the polarization analyzer parallel, in

45° orientation, and orthogonal, respectively, to the polarizer in the down-dwelling light.

Results

Behavior

There was a significant effect of illumination on the latency to detect a prey item (F1,73=5.2668, p=0.0246) as the time elapsed to the first strike was less under polarized light conditions (3.6±2.3 min, nPOL=37) than under unpolarized illumination (5.0±2.8 min, nunPOL=38) (Figure 8a). Even though there is a tendency for larvae under polarized illumination to catch prey in a shorter amount of time (3.9±2.2 min, nPOL=35) than with unpolarized light (5.1±3.0 min, nunPOL=32), the difference is statistically not significant (F1,65=3.7232, p=0.058) (Figure 8b). The state of polarization did not have an influence on how often Thermonectus marmoratus larvae would strike at a prey item during a test. On average, larvae delivered 0.8±0.6 strikes per trial under polarized illumination (nPOL=54), which is not different (F1,106=0.3684, p=0.5452) (Figure 8c) from unpolarized conditions where larvae struck 0.9±0.7 times per trial (nunPOL=54). This value also includes the trials in which Thermonectus marmoratus larvae did not attempt to catch the prey at all. However, the larvae most often delivered only a single strike towards the prey item under either illumination condition (Figure 8e), whether it was successful or not. In POL trials, the test specimens were able to catch (and eat) the prey item in 92±22 % of their attempts

(nPOL=37), whereas the larvae in unPOL tests were significantly less successful (73±38 %, nunPOL=38) (F1,73=6.5837, p=0.0123) (Figure 8d).

Figure 8 Experimental results of testing hunting performance of Thermonectus marmoratus under linearly polarized (blue bars) and unpolarized (yellow bars) illumination. a Mean detection latency: time elapsed first to first overall strike, including all failed attempts. b Mean capture latency: time elapsed to first sucessful strike, which may or may not have been preceded by a failed attempt. c Mean umber of strikes per trial, including successful and failed attacks, as well as `no attack`. d Mean probability of capture, as a fraction of unsuccessful attacks to successful strikes. e Distribution of total number of strikes per trial. Zero signifies that the Thermonectus marmoratus larva did not attempt to catch the prey item. Per trial, one mosquito larva was available as prey.

Figure 9 a absolute irradiance measurement of the down-dwelling light under POL (blue line) and unPOL (yellow line) illumination. b percent difference between the absolute irradiance measurements of the down-dwelling light of the POL and the unPOL illumination conditions. c Relative number of photons absorbed in 2150 ms by an idealized photoreceptor in the proximal retina in the POL setup (blue line) and the unPOL setup (yellow line) per nm wavelength. The values were calculated from the absolute irradiance measurements from a multiplied by the relative absorption curve for the TmUVII opsin (red dotted line) from (Maksimovic et al., in press). TmUVII is expressed in 51

the proximal retina, and its spectral sensitivity ranges from 300-450 nm, with the absorption maximum at 360nm. This graph shows the potential relevance of brightness differences to the respective opsin. Within the spectral range of the TmUVII opsin, an idealized photoreceptor cell in the proximal retina would perceive the light under POL conditions to be approximately 11% brighter than with unPOL illumination.

Properties of the illumination

(1) Overall light intensity. Under polarized light conditions, the light intensity is approximately 10% higher than under unpolarized illumination (Figure 9a, b), as measured as the amount of light that enters the behavioral arena from above. In order to determine if the intensity differences would potentially be detectable for the photoreceptors of the proximal retina (which is hypothesized to be polarization sensitive), the relative sensitivity curve of the spectral sensitivity of the Thermonectus marmoratus UVII opsin (Maksimovic et al., in press) was fitted to the post-hoc absolute irradiance measurements gained from our behavioral setup. Figure 9c shows that the 10% brightness difference between our POL conditions and the unPOL illumination lie well within the high absorbance range of the UVII opsin.

(2) Intensity, e-vector and degree of polarization of the down-dwelling light. The polarizer above the experimental setup is oriented with its transmission axis parallel to the long axis of the experimental arena, which would result in a horizontal e-vector measureable from below the setup when the polarizer is in bottom position. Under POL conditions, the light intensity is highest (in the upper UV and blue range) when the polarization analyzer component of the spectrophotometer is parallel to the experimental polarizer ( Figure 10a). However, some significant amount of light is also transmitted through the polarization analyzer when the latter is oriented orthogonally to the polarizer ( Figure 10a). Altogether, the down-dwelling illumination

Figure 10 absolute irradiance measurements taken from below the experimental setup (a+b) and through the front glass (c+d) under polarized (a+c, marked by a blue bar on the right) and unpolarized (b+d, marked by a yellow bar on the right) illumination. Measurements were taken through a polarization analyzer that was oriented parallel (blue dotted line), in 45° (red solid line), and orthogonal (green dashed line) to the polarizer above the experimental arena (see Figure 7 for reference).

of the POL setup is predominantly horizontally polarized (Figure 11b), with a degree of

(horizontal) polarization of 12-18% within the spectral sensitivity range of the proximal retina

(Figure 11a). In the unPOL setup, we would theoretically find equal amounts of light being transmitted through the polarization analyzer in either orientation. As shown in Figure 10b, the light intensity measured is slightly higher when the analyzer is parallel to the above polarizer, but largely overlapping with the analyzer rotated by 45° or 90°. This results in a somewhat dominant horizontal e-vector (Figure 11a,b).

(3) Intensity, e-vector and degree of polarization of the light near the water surface

These data are of special importance because they describe the visual field that is presented to the swimming Thermonectus marmoratus larvae. In the POL setup, the amount of light transmitted through a parallel oriented analyzer is slightly higher than with an orthogonally or

45° rotated analyzer, which translates to about 5 % of horizontally polarized light (Figure 11c,d).

Under unPOL conditions, there is no detectable difference between light intensities (Figure 5d), and the degree of polarization is only approximately 2% (Figure 11c).

Figure 11 Degree of polarization Ɵ (a+c, solid line in blue for POL, yellow for unPOL) and e-vector p (b+d, dashed line in blue for POL, yellow for unPOL) of the (un)polarized experimental conditions. For b and d, an angle of 0° (or 180°) represents a horizontal e-vector, whereas 90° refers to a vertical e-vector. The values for p and Ɵ were calculated based on the absolute irradiance measurements carried out under POL, and unPOL, illumination and a polarization analyzer (see Figure 10) and the formulas given in Sabbah & Shashar (2006). See Methods & Materials for details on the equations.

Discussion

Our behavioral data suggest that Thermonectus marmoratus might be polarization-sensitive, as proposed in our earlier work (Stecher et al., 2010). To our knowledge this is the first report on an aquatic insect with polarization sensitivity in which this ability results in a higher probability of the detection of prey with polarization features.

Our behavioral results show that whether the background light field is polarized or unpolarized does not have an effect on the foraging activity of Thermonectus marmoratus larvae

(Figure 8c,e), as measured by the amount of strikes the larvae typically deliver. However, there are clear performance differences in their hunting behavior with our polarized illumination. The larvae are able to locate prey faster (a) and strike with higher accuracy (Figure 8d) than with unpolarized background illumination. Furthermore, initial successful strikes have a tendency to occur faster under polarized conditions (Figure 8b), yet the difference is not quite significant.

This could relate to the fact that the prey is often detected with the eyes E3-6 first before the

Thermonectus marmoratus larva orients towards it with the principal eyes (Buschbeck et al.,

2007b). Nevertheless, polarization features are not necessary in order to be successful predators under experimental conditions, as Thermonectus marmoratus larvae do perform well in a lobaratory setting even under unpolarized illumination (Figure 8d). It might be that our light source was sufficiently bright to provide enough contrast for the larvae to be able to hunt successfully.

Polarization sensitivity is a phenomenon likely to be present in many holometabolous insect larvae due to the rhabdomeric nature of their visual systems. Nevertheless, no example species has been demonstrated that exploits its polarization sensitivity for any task other than simple polarotaxis (Gilbert, 1994). In Thermonectus marmoratus larvae, this visual capability could serve a more sophisticated purpose as it might improve underwater contrast. Early stages of mosquito larvae, such as those we used in our behavioral trials, are rather transparent, while later stages of mosquito larvae are more sclerotized and hence appear darker against a well-lit background. Certain camouflaging attempts of prey towards its predator can be overcome if the predator possesses a polarization-sensitive visual system. For example, the polarization-sensitive squid is able to locate translucent zooplankton to prey upon on the basis of polarization characteristics that are revealed in the prey as a side effect of transparency (Shashar et al., 1998).

Furthermore, mosquito larvae are likely to become less visible with decreasing background light levels, as a polarized light field depends on different variables, such as time of the day. The degree of polarization in fresh water lakes increases from below 40% during the day to 67% at sunrise and sunset (Novales Flamarique and Hawryshyn, 1997), and a high degree of polarization is typical for the region just underneath the water surface during crepuscular periods, especially in the UV range (Horváth and Varjú, 2004). It is possible that the natural habitat of

Thermonectus marmoratus larvae behaves similarly. Early mosquito larval stages and other low contrast but birefringent prey might still be well conspicuous (Figure 6b) to a predator if this predator is able to analyze polarization information. For example, polarization sensitivity would allow larvae of Thermonectus marmoratus to forage effectively during crepuscular periods.

The degree of polarization is also proportional to the amount of scattering media within the water, such as floating particles, algae and plants (Horváth and Varjú, 2004). Since the habitat of Thermonectus marmoratus is described as being largely without vegetation (Evans,

2006a), polarization features would be less prominent during the day, and Thermonectus marmoratus larvae might then rely on other cues to detect prey.

properties of the illumination

A series of light measurements was performed after all behavior experiments were concluded, in order to describe the illumination conditions under which the experiments were executed. This was done to certify that the behavioral differences can be credited to differences in polarization, and therefore, to exclude other factors, such as variations in overall light intensity. As shown in

Figure 9, there are indeed deviations (up to 10%) in light intensity between the two lighting conditions in our behavioral trials. We calculated the apparent brightness difference for an idealized photoreceptor cell in the proximal retina. Based on the relative absorbance of the

TmUVII opsin (Maksimovic et al., in press) and the number of available photons within the spectral range of the TmUVII opsin in each the POL and unPOL setups (Figure 9c), an idealized photoreceptor cell would absorb approximately 11% more photons in our behavioral setup under

POL conditions than with unpolarized illumination. This difference could be biologically significant. Considering that the surface of the polarizing filter is shinier than that of the wax paper used as the diffuser, it is likely that more light is reflected off the polarizer-depolarizer sandwich, and away from the behavioral setup, when the polarizer is in top position, that is, in the trials in which the incoming light is set to be unpolarized. This was not taken into account at the time when the behavioral tests were performed because the use of a combined polarizer-

diffuser appeared to be a popular routine in polarization-based behavior experiments (Novales

Flamarique and Browman, 2001; Tuthill and Johnsen, 2006; Via and Forward, 1975) and unfortunately we did not question this method. We estimated the apparent brightness of our experimental illumination for photoreceptors of the PRe only for the Thermonectus marmoratus opsin UVII. The absorption data for TmUVI is not currently available because the spectral sensitivity of Tm opsins was determined in third instar larvae, which lack TmUVI opsin expression in the principal eyes (Maksimovic et al., in press). Since TmUVI expression in this retina is very low even in first instar larvae (Maksimovic et al., 2009), the role of UVI in photon absorption is probably negligible .

Despite the overhead polarizer and diffuser in alternate position, the degree of polarization is low under POL illumination (Figure 11a,c). This could be attributed to the material of the arena which is made out of off-white opaque plastic. The sides and the bottom might act as a diffuser that reflects the polarized light and depolarizes it, and thus dilutes the degree of polarization in the water body. This is especially significant for the level which the

Thermonectus marmoratus larvae occupied during the behavioral trials. Although we expected the unPOL conditions to be completely unpolarized, we recorded a low degree of polarization for these settings. The reason could be that any unpolarized incident light, upon entering a water body, becomes partially linearly polarized as a result of refraction, and further scattering within the water body (Wehner, 2001). However, under natural conditions, the resulting degree of polarization (p) under a water surface is at a minimum when the sun stands at its zenith (Cronin and Shashar, 2001; Novales Flamarique and Hawryshyn, 1997). In our experimental setup, we

positioned the light source essentially directly above the test arena, theoretically resulting in the lowest p of the incoming light that was possible under our test conditions.

There are a number of arguments that favor the polarization conditions as the governing factor for the presence of our differential behavioral results. At the same time, there is evidence that the variance in light intensity might have had a stronger impact on the experimental outcome than desired.

To begin with, the hunting activity (strikes per trial) of Thermonectus marmoratus larvae was similar with both the polarized illumination (which also appears to be brighter, Figure 9) and the unpolarized (“darker”) illumination. If the improved hunting performance (e.g. probability of capture) under our POL conditions were in fact a matter of a higher light intensity (Figure 8d), larvae might also be expected to strike at prey more frequently. This does not happen, since the average number of strikes per trial (Figure 8c) and also the total number of strikes per trial

(Figure 8e) does not differ between the two illumination conditions.

Furthermore, it is possible that a brightness difference of 10% is in fact not detectable for the photoreceptors of the proximal retina of Thermonectus marmoratus first instar larvae. For example, it has been shown that bees are not able to (behaviorally) distinguish between two light sources with a difference in light intensity of less than 14.5% (Labhart, 1974).

In addition, our light measurements might not be entirely accurate at describing the actual nature of the illumination conditions during our behavioral trials, or, for that matter in nature.

There are several factors that need to be taken into consideration. Firstly, all measurements were carried out from the outside of the experimental arena, with a number of surfaces in between the probe and the water body. This was necessary because some of the components of the spectrophotometer do not withstand water penetration. For example, at the site where the measurements were taken, the arena itself was covered with a glass plate, which might interfere with the light intensity measurements (by reflection) and the state of polarization (by differential reflection). Moreover, the polarization analyzer that was inserted between the probe and any of the sites of measurements does reflect light off its surface the same way to polarizer of the behavioral setup does, because it is the same material. Thus, our measured irradiance values might not represent the actual illumination conditions. Secondly, the cosine corrector samples an area over an angle of 180°, which may also include portions from outside the behavioral arena, so that our collected data are possibly ‘polluted’ by external illumination conditions.

On the other hand, based on the absolute irradiance measurements we determined that the degree of polarization is lower than expected. For the down-dwelling light we calculated a degree of polarization of 10-15% for the POL conditions, and about 6% for the unPOL illumination (Figure 11a). Interestingly, it has been demonstrated that interneurons in the polarization-sensitive dorsal rim area of crickets are active at a degree of polarization as low as

5% (Labhart, 1996). However, the light that enters the water from above is not as relevant to the primary eyes of Thermonectus marmoratus larvae but the light that is reflected off the sides of the test arena. Here, the degree of polarization is only about 3-5% under POL conditions, which is barely different from the unPOL illumination (Figure 11c). Since there is little difference in polarization between the light conditions and the degree of polarization is below a possible

threshold, this raises the question whether the behavioral differences can be conclusively attributed to the polarization.

Overall, while the experimental setup that is described here appears less conclusive in answering our question, the behavioral data suggest that first instar larvae of Thermonectus marmoratus very well might be polarization-sensitive. In order to improve the setup, one could replace the top illumination with a light that surrounds the experimental chamber homogeneously, and employ polarizers that can be submersed in water and thus be attached to the inside of the chamber walls. This will allows for a controlled e-vector, e.g. horizontal polarization. Another opportunity could be a behavioral experiment in which the illumination conditions are kept constant while different types of prey is provided, with and without polarization features. Alternatively, one could use a different, physiological approach to test for polarization sensitivity in these animals, such as electrophysiology.

Chapter 3 Development of the larval eyes of the Sunburst Diving Beetle, Thermonectus marmoratus (Insecta: Dytiscidae)

Abstract

The morphology of adult compound eyes and larval stemmata greatly differ from each other, yet they are thought to have been derived from the same ommatidium-like structure. While it is difficult to assess potential homologies in the fully-developed eyes, evidence of their common ancestry might be better preserved in their development. The latter is especially well understood in adult Drosophila sp. We used basic histological methods to investigate the embryonic development of the principal stemmata E1 and E2 of the larval visual system of the sunburst diving beetle, Thermonectus marmoratus, and compare our findings against known features of adult compound eye development. The beetle larvae have six stemmata and a potentially light- sensitive area, the eye patch, on either side of the head. The eyes are characterized by a large biconvex lens, a cellular crystalline cone-like structure, and a stack of two retinas that are anatomically different from one another. The total number of participating cells in each stemma by far exceeds the traditional ommatidial number of eight cells. We find that, ss in the compound eye, the photoreceptors in Thermonectus marmoratus principal eyes morphologically differentiate before the accessory cells (corneagenous cell). Furthermore, the photoreceptor cells that are located in the proximal layer in the fully-developed stemma assume their position earlier than the prospective distal photoreceptors. In contrast, based on our samples, cell proliferation appears to be a continuous but low-level process during the stemmatal development. This is unlike Drosophila compound eye development, which is characterized by distinct periods of high mitotic activity. Moreover, in our samples, we did not observe the directional cell differentiation that is typical for Drosophila compound eye development. We anticipate that the

present study serves as groundwork for ensuing research that investigates developmental homologies between the adult compound eye and the larval stemma on a molecular basis.

Introduction

Thus far, little is known about larval eye development in holometabolous insects. Stemmata, the single-lens eyes of holometabolous insect larvae, are thought to be derivatives of adult compound eye ommatidia (Paulus, 2000; Paulus and Schmidt, 1978). However, the stemmatal structure is strongly modified and distinct from the common ommatidium, which raises the question of how stemmata develop.

The common cellular components of an adult insect ommatidium include eight photoreceptor cells (two central and six peripheral), four cone cells, and two primary pigment cells. The number of pigment cells can vary, as groups with larger and optically advanced eyes types often have a higher number of primary pigment cells (Nilsson and Kelber, 2007). In addition, an ommatidium generally possesses a corneal lens and a crystalline cone (Nilsson and

Kelber, 2007; Paulus and Schmidt, 1978).

In Drosophila sp., the adult compound eye develops from a single layer of epithelium.

The components of an ommatidium develop in a particular order, beginning with the cell that becomes the proximal central photoreceptor cell R8. The peripheral photoreceptor cells R1-6 differentiate next. The last photoreceptor cell to be recruited is the distal central R7. Afterwards, the four cone cells are added to the cluster of photoreceptor cells, followed by the primary pigment cells. Eventually, the cone cells and primary pigment cells secrete the ommatidial lens, and then the pseudocone (Cagan and Ready, 1989; Ready et al., 1976; Tomlinson, 1985).

There are two periods of high mitotic activity during the ommatidial development in

Drosophila. The first mitotic phase occurs at the monolayer stage and continues throughout all larval periods (Ready et al., 1976). In the late third instar, this initial cell proliferation is followed by the onset of cell differentiation, which begins posteriorly in the eye disc. It is recognizable by the morphogenetic furrow, a dorsolateral constriction of those cells that will undergo differentiation. The second mitotic wave produces the cells that will later develop into the last photoreceptor cells to be recruited (R1 and 6, R7) as well as the cone cells and all pigment cells

(Tomlinson, 1985; Wolff and Ready, 1991).

The ommatidia in the compound eye of the adult flour beetle Tribolium castaneum are organized similarly to adult Drosophila sp. ommatidia. Early ommatidial development in T. castaneum largely parallels that of Drosophila, except that the cone cells in the T. castaneum differentiate relatively earlier than in Drosophila (Friedrich et al., 1996).

The larval eyes of holometabolous insects are generally very diverse. However, larvae of adephagous Coleoptera (such as Dytiscidae, Carabidae and Gyrinidae) typically have a total of six pairs of stemmata (Paulus, 1986, 2000), two of which are often enlarged (Gilbert, 1994). In larvae of Thermonectus marmoratus, the forward-pointing principal stemmata Eye 1 (E1) and

Eye 2 (E2) are elongated, which results in their tubular shape. The remaining four stemmata (E3-

6) are more or less spherical (Mandapaka et al., 2006). The thick, circular lenses of all eyes are biconvex (Mandapaka et al., 2006), and it has been shown that at least the lenses of E2 are bifocal, producing a focused image simultaneously on two separate planes (Stowasser et al.,

2010).

In addition to the six stemmata, the larvae of some coleopteran taxa, e.g. Dytiscus, have a pigmented, potentially light sensitive spot adjacent the stemmata (Gilbert, 1994; Günther, 1912).

In Thermonectus marmoratus larvae, this region is termed the eye patch (EP). Located medial to

E1 and below the dorsal cuticle, the eye patch lacks a lens but possesses a retina (Mandapaka et al., 2006a) .

Generally, a crystalline cone is mostly absent in larvae of Coleoptera (Paulus, 1986,

2000; Paulus and Schmidt, 1978) although the adults in adephagous species have superposition eyes with highly refractive crystalline cones (Crowson, 1981; Kirchhoffer, 1908; Meyer-

Rochow, 1975; Paulus, 1975). However, the stemmata of Dytiscidae larvae have a hyaline structure above the rhabdom (Paulus 1986), which was described as a “vitreous body” (Günther,

1912) that is formed by hyaline elongated corneagenous cells (Paulus, 1986). In fact, in

Thermonectus marmoratus, the cone (labeled “crystalline cone” in Mandapaka et al. (2006a), and “crystalline cone-like structure” in Maksimovic et al. (2009) and Stecher et al. (2010)) is composed of numerous long, narrow (corneagenous) cells and extracellular material (Mandapaka et al., 2006).

A layered retina with a high number of photoreceptor cells is a frequently occurring feature among larvae of Adephaga (Paulus, 1986). For example, the retina in the stemmata of some tiger beetle larvae (Cicindelinae) consists of more than 1000 photoreceptor cells, albeit these cells are not arranged in layers (Paulus and Schmidt, 1978). In contrast, Thermonectus marmoratus stemmata have two anatomically distinct retinas, the distal and the proximal retina, which are stacked atop one another and contain at least 24 photoreceptor cells in the distal retina alone, plus more than 300 photoreceptor cells in the proximal retina (Mandapaka et al., 2006;

Stecher et al., 2010). Additionally, E1 possesses a medial retina which spans the long axis of the

crystalline cone-like structure medially (Mandapaka et al., 2006). Because the plesiomorphic number of larval stemmata is six (in Adephaga), the large number of photoreceptor cells in a single stemma is likely due to an increase of total cell numbers, and not a result of a fusion of several stemmata. A layered arrangement of the retina may have resulted from the necessity to rearrange the numerous photoreceptor cells (Paulus, 1986, 2000). Paulus (1979) and Paulus and

Schmidt (1978) suggested that an increased number of photoreceptor cells per stemma might have been an adaptation to the diurnal predatory mode of life of the larval stages of adephagous beetles.

Although highly variable (Gilbert, 1994), stemmata are generally considered to have been derived from the most posterior portion of compound eyes (Liu and Friedrich, 2004; Paulus,

2000; Paulus and Schmidt, 1978). The stemmata of adephagous Coleoptera, such as the larval visual system of the Sunburst Diving Beetle, Thermonectus marmoratus (original study by

Mandapaka et al. (2006)), are outstanding in their organization and functionality. Considering that there often is conservation in the sequence of developmental events, comparing the embryonic development of Thermonectus marmoratus to that of other species can gain insights in regards to homologies between structures such as eyes. For example, how do the stemmata of

Thermonectus marmoratus compare structurally to ommatidia? One way to answer this question is to examine their development for structural similarities. To what extent is the development of ommatidia comparable to the development of the stemmata of a holometabolous insect larva such as Thermonectus marmoratus? In order to structure our analysis, we are using such key events as the development of photoreceptor cells and corneagenous cells from the imaginal eye anlagen, the progress of retinal layering, and the formation of the lens and the crystalline cone-

like structure. Moreover, the basic bauplan of the visual system of larval Holometabola incorporates six eyes. In larvae of Thermonectus marmoratus (and other Dytiscidae) there is a seventh visual organ, called the ‘eye patch’ (EP, Mandapaka et al. (2006)). How does the EP develop? We are using basic histology and antibody labeling to address these questions.

Materials and Methods

Beetle husbandry

Adult Thermonectus marmoratus were initially provided by the Insectarium of the Cincinnati

Zoo and Botanical Garden, or collected near Tucson, AZ (USA) between 2004 and 2008, or ordered from Hatari Invertebrates (Portal, AZ, USA). Adult beetles were kept in a fresh water aquarium which also provided moist terrestrial oviposition sites (25°C) covered with moss and bark. The beetles were maintained on a 16 h:8 h light:dark photoperiod, and fed freshly killed or frozen-thawed crickets.

Embryonic stages

Adult female beetles lay eggs onto the designated oviposition sites underneath vegetation in clusters of about 7 to 19 eggs. The females do not seem to have a time preference for depositing their eggs but new eggs can often be found in the late morning hours (corresponding to 5-8 hours after the beginning of our artificial light period). New clusters were transferred to a small petri dish filled with tissue paper and plant material, labeled with the date and time of oviposition, and returned to the oviposition site. The eggs were transported with a thick paint brush to avoid deformations. For the analysis of the stemma development, the embryos were sampled in 6h intervals, starting with 12 hours post-oviposition to the emergence of the first instar larva (114-

118 hours). This served as a reference library in regards to developmental stages, which varied somewhat between different batches.

Ethyl gallate staining

Most embryos were removed from their chorion and decapitated in Insect Ringer (Adams and

O’Shea, 1983). The tissue of very young embryos is loose and difficult to dissect. Therefore, eggs 12, 18 and 24 hours post-oviposition were simply cut in half to avoid damaging the anterior region. The heads were prepared for Ethyl gallate staining as described in Mandapaka et al.

(2006). Specimens were sectioned either frontally, horizontally or sagittally at 8µm on a Leitz

(Wetzlar) microtome.

Phalloidin and DAPI staining

Phalloidin binds to actin molecules (Capani et al 2001) and thus can be used to visualize the site of microvilli, which in turn can point strongly to the existence of a (developing) rhabdom if actin is found in the eye region. DAPI is a marker that can be utilized to detect DNA molecules

(Kapuscinski 1995) and hence, the location of nuclei. Based on the histological series of ethyl gallate-stained samples, we estimated an appearance of rhabdomeric structures no earlier than 36 hours post-oviposition. Thus, we began sampling specimens for Phalloidin-staining at 30 hours post-oviposition and continued in six-hour-intervals to 84 hours post-oviposition. All Phalloidin preparations were double-stained with DAPI to aid in orientation and identification of tissue.

Embryos were dissected in Insect Ringer. The embryos were removed from their chorion and decapitated. The heads were fixed in a solution of 4% Formaldehyde (EMS) in Phospate Buffer

Saline PBS at pH 7.5 (EMS) for 3h at room temperature and washed thrice in PBS for 20 minutes each.

Cryosectioning: After fixation and rinsing in PBS, the heads of the embryos were cryoprotected in a solution of 20% sucrose (Acros Organics, NY, USA) in PBS overnight at 4°C. The tissue

was transferred to a frozen tissue embedding medium (Neg50, Thermo Scientific, or Tissue-Tek

O.C.T. compound, Sakura Finetek, Japan), placed onto a cryomount and subsequently sectioned at 10µm in a cryostat (Ames Lab-Tek). The sections were placed on adhesion slides (SuperFrost

Plus Gold Slide or Yellow ColorFrost Plus Slide, Electron Microscopy Sciences) and allowed to dry for at least 1h at room temperature and then washed 3x for 20 minutes in 10x Phosphate

Buffered Saline (PBS). All sequences of washing were performed with little or no agitation to ensure that the tissue was not lost. The sections were blocked in 1% bovine serum albumin

(BSA) in PBS for 1h at room temperature. All following procedures were performed at low ambient light levels to minimize the bleaching of fluorescent probes. The sections were incubated in 1:100 Alexa Fluor 568 Phalloidin (Invitrogen Molecular Probes) in 1% BSA for 30-

60 minutes at room temperature. Slides were washed 3x for 10 minutes in PBS, then covered with a solution of 1:500 DAPI (Sigma, St. Louis, MO) in PBS for 30 minutes at room temperature, and subsequently washed 3x in PBS for 10 minutes. Slides were briefly rinsed with distilled water and left to dry at room temperature. Finally, slides were mounted with Fluormount

G (EMS).

Mitosis marker

Tissue was sampled at two-hour-intervals beginning at two hours post-oviposition to 20 hours post-oviposition, then at 24, 30 and 48 hours. All mitosis marker preparations were double- stained with DAPI to aid in orientation and identification of tissue.

Specimens were dissected and prepared for cryosectioning as described for

Phalloidin/DAPI staining. The cryosections were transferred to adhesion slides (SuperFrost Plus

Gold Slide or Yellow ColorFrost Plus Slide, Electron Microscopy Sciences). The slides were

allowed to dry for at least 1h at room temperature and then washed thrice for 20 minutes in PBS.

All sequences of washing were done with little or no agitation to ensure that the tissue was not lost. Furthermore, all following procedures were done with low ambient light levels to minimize the bleaching of the fluorescent probes.

The sections were incubated in DAPI at a dilution of 1:500 in PBS for 30 minutes at room temperature. Slides were then washed once in PBS and twice in PBX (PBS & 0.03% Triton

X) for 10 minutes each. The tissue was blocked with 5-10% Normal Goat Serum (NGS) in PBX for 1-2 hours at room temperature or overnight at 4°C. Subsequently, the tissue was incubated in the primary antibody Anti-phospho-Histone H3 (Millipore, Billerica, MA) at a dilution of 1:500 in NGS for 4 hours at room temperature or overnight at 4°C. The slides were washed three times with PBX and then covered with a solution of the secondary antibody Alexa Fluor 568 goat anti- rabbit (Invitrogen Molecular Probes) for three hours at room temperature. The slides were washed three times in PBX, quickly rinsed with distilled water, and then prepared as described for the Phalloidin/DAPI staining.

Images

All photographs were taken with a camera (Olympus BKH022083 or Q-imaging 01-RGB-HM-

S-IR) that was mounted onto a microscope. Embryos were viewed and photographed with a dissection microscope (Olympus SZX16), and all sections (ethyl gallate and fluorescence) were viewed and photographed with a compound microscope (Olympus BX51). Fluorescence images were taken in black and white and subsequently pseudo-colored and merged using the Q-Capture

Pro 6.0 software (Media Cybernetics, Inc., and QImaging, Inc., version 6.0.0.605 for Windows).

Contrast and brightness of all images were adjusted and strong background staining in fluorescent images was decreased in Adobe Photoshop CS5.

Three-dimensional reconstruction

Ethyl gallate-stained sections (8 µm thickness) were photographed in series and loaded into

Amira (Visage Imaging, Version 5.3.2 for Windows). Missing sections were filled in with duplicates of the adjacent section. The slices were initially aligned manually and then optimized with the automatic alignment tool. Specific structures were labeled in the image segmentation editor throughout all relevant slices. Using the label smoothing tool, the label markings were smoothed out. From optimized labels, surfaces were created in the surface generator.

Results

External morphology of the embryo of Thermonectus marmoratus

Under laboratory conditions, the embryonic period of Thermonectus marmoratus lasts approximately five days (108-114h). Several developmental stages of the embryo can be distinguished (Figure 12). The development of the stemmata and the eye patch can be traced externally, as well as stemmatal shape and position in reference to the head axis. A newly laid egg measures approximately 3.5 to 4mm in length. As the embryo differentiates and grows, the cephalic and caudal ends of the embryo develop ventrally so that the head and tail are tucked under the body portion of the embryo within the egg shell (Figure 12c,e&g).

New eggs (Figure 12a) look uniformly white since the embryo does not have any discernable body segmentation and completely lacks pigmentation. On the first day after oviposition (Figure 12b), the embryo, which lies ventrally, becomes morphologically distinguishable from the yolk (dorsal). On the second day after oviposition (Figure 12c&d), embryos have a well-developed head. The larval eyes E1-6 can be detected externally. The developing stemmata exhibit yellowish-brown pigmentation, which is darkest at the center of each stemma, but do not possess lenses. E1 and 2 are elongated approximately dorsoventrally (in reference to the head axis) and can be seen in both dorsal and ventral view of the head. E3 through E6 are approximately oval in shape. A yellowish-brown pigmented band projects from the dorsal end of E1 anteriorly. With the help of ethyl-gallate stained sections we were able to identify this projection as the developing medial retina (Figure 12c insert, mre). Furthermore, a

clear line runs through both E1 and E2 anteroposteriorly (Figure 12c insert, clear line at arrowhead).

Figure 12 External morphology of Thermonectus marmoratus embryos. a Embryo 8 hours post-oviposition. The egg appears uniformly white. b Embryo 27 hours post-oviposition. The embryo (ventral, dashed outline) is discernable from the egg yolk (dorsal). Arrow points to approximate location of developing head. c Embryo 56 hours post- oviposition, the stemmata 1 through 6 are clearly visible externally as separate units with yellow-light brown pigmentation. Insert: ellipse indicates the position of the medial retina (mre) of E1. Arrowhead points at the anteroposterior cleft. d Same stage as c, head in dorsal view. e Embryo 80 hours post-oviposition. The stemmata are wider than in earlier stages (c&d) and their pigmentation darkens. Insert highlights location of mre. f Same stage as e, head in dorsal view, box indicated position of eye patch (EP). g Embryo 104 hours oviposition. The stemmata possess lenses (arrow, L) and dark-brown pigmentation. Insert: EP is unpigmented but separated from E1. h Same stage as g, head in dorsal view. Orientation for specimens a-c, e and g (incl. inserts) given in a, unless otherwise noted. Orientation for specimens d, f, and h given in d. ant anterior, dor dorsal, lat lateral, post posterior. i Diagram of an embryo in lateral view, defining the line of reference used to describe positions of structures in relation to the body axis. k Diagram of an embryo in lateral view of the head, defining the line of reference used to describe positions of structures in relation to the head axis. Scale bars 1mm.

In three- day-old embryos (Figure 12e&f), the stemmata are externally visible as a separate region of dark/brown pigmentation surrounded by a clear zone. A lighter area also is visible towards the front of Eye 1 and E2 which are discernable as dorsoventrally oriented tubes. The medial retina is still largely separated from the pigmented body of E1 (Figure 12e insert, mre encircled). Although unpigmented, the eye patch (Figure 12f, EP marked with box) is visible externally medial and posterior to E1. E3-6 are oval to round in shape. The stemmata of embryos on the fourth day post-oviposition (Figure 12g) possess lenses (arrow). The tubular E1 and 2 are oriented dorsoventrally (in reference to the head axis). Externally, a dark/black pigmentation of the stemmata is visible. The eye patch is still unpigmented but separated from E1 (Figure 12g insert, EP). Under our laboratory conditions, Tm larvae hatch on day five (108-114h post- oviposition) and enter the water.

The origin of the stemmata

Imaginal disc primordia are recognizable as early as 12 hours after oviposition in our DAPI- and mitosis marker-stained samples (Figure 13a&b) as a continuous pseudostratified epithelium that runs ventrally (in relation to the body axis,

Figure 12i, Figure 13c) in the anterior part of the embryo (Figure 13a). Although the leg disc primordia can be traced, we are unable to determine specific cephalic imaginal disc anlagen (the initial clustering of embryonic cells) at this point. Mitotic activity in the cephalic disc primordia is low and spotty (Figure 13a&b, mitotic cells colored in red).

Figure 13 a&b Sagittal section of embryo 12 hours post-oviposition. Nuclei of imaginal disc epithelia are stained with DAPI (colored in blue). Mitosis-active cells are marked with anti-histone H3 (colored in red). a Approximate location of imaginal disc anlagen. b Higher magnification of cephalic imaginal disc as framed in a, showing low degree of mitosis activity. c Embryo at 8 hours, demonstrating the section levels of a&b. d Embryo at 27 hours, demonstrating the section level of e. e Ethyl gallate-stained sagittal section of the pseudostratified epithelium of the eye disc at 30 hours, highlighting one of the developing principal eyes. Scale bars 100µm.

During the next 12 hours, the pseudostratified epithelium elongates along the long axis of the embryo. Mitosis takes place on a small scale at every developmental stage that we sampled

(from 2-20 hours in two-hour intervals, as well as at 24 hours and 30 hours post-oviposition), suggesting the presence of prolonged, low level mitosis activity. By 30 hours (about 25% of th total duration of the embryonic development), the eye disc has distinctively developed into concrete clusters of particularly elongated cells that form the beginning of individual eyes

(Figure 13d&e).

Stemmata Eye 1 and Eye 2

By 36 hours, the developing stemmata are clearly visible within the head capsule of the embryo

(Figure 14a). The very long and narrow cells begin to differ in morphology. More specifically, some of the cells have started to project axons (Figure 14b, ax) into the optic lobe precursor

(Figure 14b, op lb), and can thus be identified as photoreceptor cells. By 42 hours (38% embryonic development), the stemmata can be identified individually in horizontal sections as cup-shaped clusters of morphologically differentiated cells (Figure 14d&e, Figure 15a-d). The cup shape appears to arise from photoreceptor cells that have shortened and have moved deeper into the tissue (Figure 14d&e, Figure 15a-d, dre). Moreover, some photoreceptor cells bend towards an apical indentation which runs along the dorsoventral axis (in reference to the head axis,

Figure 12k) of the stemmata (Figure 14d, Figure 15a-d, arrowhead). In the following, we are referring to that apical indentation as the embryonic fissure. The axons of the photoreceptor cells of one stemmatal cup bundle together before they reach their respective optic lobe (Figure 14d,

ax). The developing corneagenous cells can be recognized in ethyl gallate-stained horizontal sections on the periphery of the eye cup (Figure 14d&e, Figure 15a-d, cng). The proximal end of

Figure 14 a Ethyl gallate-stained sagittal section of eye anlagen 36 hours post-oviposition. The stemmata develop within the outer epidermis of the head. In sagittal sections, the stemmata appear as a continuous cell layer. Cell populations begin to morphologically differentiate. b horizontal section medial to a. Axonal projections (ax) identify the developing photoreceptor cells. c Embryo at 54 hours, in lateral view demonstrating the section levels of a&b, d&e. d Ethyl gallate-stained horizontal section of the cup-shaped principal stemmata E1 (right) and E2 (left) at 42 hours post-oviposition. Cell populations are distinctly differentiated morphologically and photoreceptor cells with their axonal projections into their respective optic lobes (op lb) are visible. Developing corneagenous cells (cng) are located on the periphery of each stemmatal cup. The corneagenous cells of one stemma form a continuum with the corneagenous cells of the adjacent stemma. The middle of each approximately symmetric stemma is established by the embryonic fissure (arrowhead) e higher magnification of the area outlined in d, highlighting the lack of 82

morphological separation between the adjacent populations of developing corneagenous cells of neighboring stemmata. Scale bars 50µm.

the cells narrows significantly as it forms the sheath of the stemmatal cup and reaches the base of the eye (Figure 14d). Furthermore, the corneagenous cells of one stemma appear to be continuous with the directly neighboring corneagenous cells of the adjacent stemma (Figure

14e). Serial horizontal sections of E2 (Figure 15a-d) show that the proportions of putative photoreceptor cells to corneagenous cells, as well as their relative position within the cup-shaped structure stay constant throughout the length of the developing eye. A three- dimensional reconstruction based on horizontal sections reveals that the photoreceptor cells and corneagenous cells are oriented in layers that run parallel in dorsoventral direction relative to the head axis

(Figure 15f).

Figure 15 Embryo 42 hours post-oviposition. a-d ethyl gallate-stained horizontal sections of E2 at different levels of depth. In horizontal sections, the location and appearance of the developing retina (re) and the corneagenous cells (cng) remain relatively unchanged throughout the long axis of the stemma, suggesting that the photoreceptor cells 83

and corneagenous cells are oriented in parallel layers, with the corneagenous cells at the periphery, and the photoreceptor cells centrally. In these sections the stemma appears approximately symmetric, with the apical embryonic fissure (arrowhead) marking its middle. Scale bars 50µm. e Embryo at 48 hours, indicating the approximate section levels of a-d. f three-dimensional reconstruction of E2 based on an ethyl gallate-stained horizontal series, illustrating the layered appearance of corneagenous cells (cng, colored in purple) and photoreceptor cells (re, colored in green). Approximate section levels of a-d are indicated.

Figure 16 Embryo 54 hours post-oviposition. a Ethyl gallate-stained sagittal section of E1. Distal (dre) and proximal retinas (pre) are morphologically separated. Distal photoreceptor cells have secreted a clear, refractive substance at their most dorsal region (ellipse). b Sagittal section of E1 with DAPI (blue, marked nuclei) and Phalloidin (red, marking actin) staining. Actin present in the developing rhabdoms of the distal retina (drh) and the proximal retina (prh). c Ethyl gallate-stained sagittal section of E1, apical-most region. In this region, corneagenous cells are rich in vesicles that contain a clear substance. d-g Ethyl gallate-stained horizontal sections of E2 at four different levels of depth; approximate section levels are indicated in i (left). d The vesicle-rich area of the corneagenous cells as well as the embryonic fissure are deepest most anteriorly. The anterior part of the stemma also is dominated by the corneagenous cells. e In a section posterior to d, the vesicle-rich region of the corneagenous cells is shallower, as is the depth of the embryonic fissure. Corneagenous cells overlie the distal retina, and the dre is situated above the proximal retina. f Section level posterior to e. The dre overlies the pre. g Section level posterior to f. The posterior- most region of the stemma is dominated by proximal retinal tissue. h Three-dimensional reconstruction of E2 based on an ethyl gallate-stained horizontal section series, showing the layered appearance of corneagenous cells (colored in purple), distal (colored in blue) and proximal (colored in yellow) photoreceptor cells. The narrow, proximal cell extensions of cng and distal photoreceptor cells were omitted in order not to obscure the proximal retina. i Embryo at 54 hours, indicating section levels of d-g (left) and k-n (right). k-n Ethyl gallate-stained series of sagittal sections; approximate section levels indicated in i (right). Confirming the rearrangement of the tissue layers compared to 42 hours post-oviposition (Figure 4). l Example of how narrow cell extensions of the cng (purple arrow) and dre (blue arrow) reach the base of the stemma. Scale bars 50µm.

By 54 hours post-oviposition (48% of embryonic development), the central photoreceptor cells have withdrawn from the epidermis in most of the eye but the ventral end. These cells give rise to the proximal retina and are now largely situated underneath the peripheral photoreceptor cells which give rise to the distal retina, as can be seen in ethyl gallate-stained horizontal sections

(Figure 16d-g) as well as in sagittal sections (Figure 16k-n). Similarly, the peripheral photoreceptors are overlain by the developing corneagenous cells (Figure 16e&f, k-n). A three- dimensional reconstruction of E2 (based on horizontal serial sections) demonstrate that the tissue layers that were parallel sheets by 42 hours now begin to morph into layers that are stacked dorsoventrally (Figure 16h). At this point the morphological differentiation of the distal retina

(formerly peripheral photoreceptors) versus proximal retina (central photoreceptors) is fairly distinct (Figure 16, dre and pre). Apically, both the distal and proximal photoreceptor cells stains positive for actin (Figure 16b), which indicates that the rhabdoms started to develop.

Furthermore, the dorsal-most cells of the distal retina have started to secrete a small portion of a refracting clear substance (Figure 16a, ellipse). The apical end of the distal photoreceptor cells now slants towards the center of the stemma, so that the cells of the distal retina start to be oriented obliquely to the direction of the incoming light (Figure 16e&f, k-n, dre). The long axis of the proximal photoreceptor cells is aligned mostly dorsoventrally (Figure 16e&f, k-n, pre).

The proximal end of the distal photoreceptor cells narrows substantially as cells extend past the base of the eye, thus enclosing the proximal photoreceptor cells (illustrated in Figure 16l, blue arrow). The distal photoreceptor cells are in turn surrounded by the narrow proximal region of the corneagenous cells (illustrated in Figure 16l, purple arrows), which likewise extend to the base of this structure. The embryonic fissure persists only in the dorsolateral portion of the eye, where the corneagenous cells are tightly anchored in the epidermis (Figure 16d&e,k&i, arrowhead, h). The apical region of the corneagenous cells is filled with vesicles that contain a clear substance which presumably mark the forming lens (Figure 16c, ves). The vesicle-rich cell proportion is largest at the dorsal end of the eye (Figure 16d) and flattens out more ventrally

(Figure 16e). The extent of the vesicle-rich region coincides with the depth of the embryonic fissure, which is deepest anteriorly (Figure 16d&e, arrowhead, h).

By 72 hours post-oviposition (~65% of embryonic development), the rearrangement of corneagenous cells, distal photoreceptor cells and proximal photoreceptor cells into a dorsoventral stack is largely completed (Figure 17), and E1 and 2 are beginning to acquire the tubular shape that is typical for the principal eyes in the fully developed larval stage. The eyes will further elongate during their subsequent development. The small portion of a clear substance above the distal retina that we observed in our samples at 54 hours (Figure 16c, ellipse) is still

Figure 17 Embryo 72 hours post-oviposition. a-d Ethyl gallate-stained sagittal sections of E2 at different section levels as depicted in e and f. a-c Corneagenous cells (cng), distal retina (dre) and proximal retina (pre) assumed their final, stacked, relative positions as they are found in the fully-developed larval stemma. b Distal photoreceptor cells have thin extensions (blue arrow) that reach the basal membrane and enclose the proximal retina. The distal retina is surrounded by narrow proximal extension of the corneagenous cells (purple arrow). d Posterior-most region of eye is dominated by the tissue of the pre. e Three-dimensional reconstruction of E2 based on a series of ethyl gallate- stained sagittal sections, in side view. In order not to occlude the dre and pre, narrow cell extensions of cng and dre were omitted from the reconstruction. Scale bars 50 µm.

present in our samples at 72 hours (Figure 17a, ellipse). The embryonic fissure is fully closed.

The distal region of the corneagenous cells bends medially, which gives these cells an oblique orientation just above the distal retina but a nearly dorsoventral alignment close to the epidermis

(Figure 17a). In our ethyl gallate-stained samples, the vesicle-rich apical region in the corneagenous cells is somewhat restricted to those cells that lie in the middle of the eye when viewed in sagittal sections (Figure 17b, ves). Furthermore, the apical ends of some corneagenous cells (the cng region directly below the vesicle-rich area) cannot be easily resolved with ethyl gallate staining. This contrast-poor zone is especially conspicuous in the central corneagenous cells with immediately overlie the center of the distal retina (Figure 18a, arrow).

In our samples at 84 hours post-oviposition (~75% of embryonic development), the contrast-poor zone of the corneagenous still provides little cellular resolution with ethyl gallate staining (Figure 18b, arrow). However, the zone stains very dark compared to the proximal end of the cells. Above the vesicle-rich area, the corneagenous cells have secreted a thin lens with almost uniform thickness (Figure 18b&l).

By 90 hours (~81% of embryonic development), the thickness of the lens has increased more drastically in the central region than at the periphery. While the front surface of the lens is somewhat flat, its back surface has a stronger curvature. (Figure 18c&l). The corneagenous cells began to form the crystalline cone-like structure (Figure 18c, cc). More specifically, the crystalline cone-like structure is composed of the largely dorsoventrally oriented portion of the corneagenous cells which is positioned above the rhabdom of the distal retina (Figure 18c, drh).

The crystalline cone-area is separated from the more peripherally positioned part of the cell

Figure 18 a Embryo at 72 hours, sagittal section of E1. Centrally, the apical region of the corneagenous cells provides less contrast (arrow) with ethyl gallate staining than the apical region of peripherally located cng. As seen in earlier stages (Figure 3f), a clear, refracting substance fills the extracellular space above the central region of the distal retina (ellipse), and it also secreted by distal photoreceptor cells. b Embryo at 84 hours; ethyl gallate-stained horizontal section of E2. Corneagenous cells begin to secrete a uniformly thin lens (l) above the vesicle-rich area (ves). The region of the cng that is just proximal to the vesicle-rich region stains darkly with ethyl gallate, which causes poor resolution on a cellular level. c Embryo 90 hours post-oviposition; ethyl gallate-stained sagittal section of E2. The center of the lens is thicker than its periphery. The curvature of the lens is more flat on the dorsal surface than its back surface. The apical regions of the cnc form the crystalline cone-like structure (cc), which appears separated from the proximal end of the cnc by a small segment that stains darkly with ethyl gallate (arrow). d Embryo at 102 hours; ethyl gallate-stained sagittal section of E2. The back surface of the lens has an increased curvature compared to a lens at 90 hours (c). The drh appears approximately triangular in shape. The rhabdomeric region of distal photoreceptor cells is extended centrally, except in the distal-most area, in which the rhabdomeres are located distally (arrow). Extracellular material is located in small area between the cc and the drh (circle). Scale bars 50 µm.

bodies by a thin segment that stains darkly with ethyl gallate, and thus serves as the apparent border of the crystalline cone-like structure (Figure 18c, arrow).

At 102 hours (~92% of embryonic development), top-most cells of the distal retina are tilted away from the center, which extends the dorsal portion of the distal rhabdom extend towards the periphery (Figure 18d, arrow) and results in an approximately triangular shape of the rhabdom. The thickness of the lens has increased from earlier developmental stages (90 hours), especially in the central portion of the lens. Likewise, the curvature of the lens is more pronounced, particularly at its back surface. The small portion of extracellular material is still present above the distal rhabdom (Figure 18d, circle).

The medial retina and the eye patch

The developing eye patch is first noticeable in our samples at 42 hours post-oviposition in close proximity to E1 (Figure 19a-c, EP). More specifically, the dorsal surface of the eye patch, which contacts the epidermis, appears to be located just dorsal to the beginning of E1 (Figure 19a). In ethyl gallate-stained horizontal sections from a region slightly more ventral to the section level of

Figure 19a, the ventral part of the eye patch does not reach the epidermis but instead runs alongside the proximal and medial corneagenous cells and photoreceptor cells of the anterior- most portion of E1 (Figure 19b&c). The EP appears nearly round in ethyl gallate-stained horizontal sections. By 54 hours, both the dorsal and the central part of the eye patch contacts the epidermis (Figure 19d, EP). The upwards movement of its ventral part occurs simultaneously with the downwards migration of the photoreceptor cells which was described earlier for E2.

Figure 19 Development of the medial retina (mre) and the eye patch (EP). a-c Embryo at 42 hours. Three consecutive ethyl gallate-stained horizontal sections of the anterior end of E1 (a is most anterior). The EP develops in close physical proximity to E1. Embryonic fissure in E1 marked with arrowhead. d Embryo at 54 hours. Ethyl gallate-stained sagittal section of the medial portion of E1. Both the anteromedial photoreceptor cells of E1and the EP contact the epidermis. The upwards migration of the EP occurs in parallel with the downwards movements of the photoreceptor cells. In E1, the anteromedial-most photoreceptor cells of E1 do not fully migrate ventrally as in E2 (illustrated in Figure 5), but instead establish the medial retina. e&f Embryo at 72 hours. e Ethyl gallate-stained sagittal sections of medial portion of E1. The EP projects axons (ax). f Ethyl gallate-stained sagittal section of EP, section level medial to e. EP somewhat detaches from E1 to a position more medial and posterior to E1. g&h Embryo 84 hours post-oviposition. Medial retina positioned in the medioanterior region of E1. g Ethyl gallate- stained horizontal section of E1 and EP. Although they lie in close proximity, the EP and E1 appear completely detached. h Ethyl gallate-stained horizontal section of far-anterior region of E1. Unless otherwise noted, the orientation of all structures in each line is presented in the respective left-most image. Scale bars 50µm.

However, the anteriomedial-most photoreceptor cells of E1 do not fully move ventrally and thus establish the medial retina (Figure 19d, mre). Axons originating from the EP (illustrated Figure

19e, ax) extend posteriorly, and following subsequent sections it can be observed that the axon from the EP ultimately project into the same optic lobe as the photoreceptor cells of E1.

Although the eye patch generally remains in close proximity to E1, over the next 30 hours of embryonic development it migrates medially and separates from E1 (Figure 19e-h) in all but the most proximal region from which the axons converge. At the same time, the medial retina shifts from a mostly anterior orientation (Figure 19d&e, mre) to a more medial position (Figure

19g&h).

Discussion

Stemmata, the larval eyes of holometabolous insects, are highly diverse in terms of number, form and function (Gilbert, 1994). The stemmata of the Sunburst Diving Beetle, Thermonectus marmoratus, are probably among the most sophisticated eyes ever described for holometabolous insect larvae. The six pairs of eyes vary greatly from one another in structure and specific task

(Mandapaka et al., 2006). The principal stemmata E1 and E2 are characterized by their overall tubular shape, a large biconvex lens, a cellular crystalline cone-like structure and at least two different retinas that are anatomically distinct and arranged into stacks (Mandapaka et al., 2006).

In order to determine how these eyes develop, we used using basic histological methods. We investigated the morphological changes from the eye anlagen to the fully-developed stemma with its layered appearance.

The eyes originate from a pseudostratified epithelium (Figure 13). By ~32% of embryonic development, the stemmata take the form of a dorsoventrally oriented layer of parallel cells that appear to be folded around a central fissure (Figure 15f). Morphological cell differentiation has already started, and photoreceptor cells can be distinguished from corneagenous cells, with the latter being located at the periphery. By ~48% of embryonic development, retinal differentiation has progressed further, and a portion of the cell layers which originally was close to the surface has sunk deeper into the embryo, leading to more or less obliquely-oriented tissue (Figure 16h). The photoreceptor cells have withdrawn from the epidermis and are overlain by the corneagenous cells. Furthermore, the formerly centrally- located photoreceptor cells have moved to a position below the formerly more peripherally-

located photoreceptor cells, now establishing distinct retinal layers. Corneagenous cells, distal photoreceptor cells and proximal photoreceptor cells assume their final, stacked position by

~65% of embryonic development (Figure 17e). The corneagenous cells have begun to secrete the lens by ~75% of embryonic development, and ultimately give rise to the crystalline cone-like structure by ~81% of embryonic development.

Since the eyes of holometabolous insect larvae are considered to be homologous to the ommatidia in adult compound eyes (Paulus, 2000; Paulus and Schmidt, 1978), one might expect certain morphological similarities between the stemmatal development in Thermonectus marmoratus and the formation of a compound eye, such as that of Drosophila. On the other hand, the eyes of Thermonectus marmoratus are highly specialized, sophisticated lens eyes, which likely have changed substantially from the ancestral bauplan. In the following sections we discuss similarities and differences between known developmental events in ommatidia with those of Thermonectus marmoratus stemmata.

The prospective eye tissue

In Drosophila, the ommatidia develop from a simple epithelium. The eye imaginal disc primordium is undifferentiated and proliferative. Cell differentiation is induced at the posterior end of the eye disc and then spreads anteriorly. Those cells that undergo differentiation constrict dorsoventrally, which appears as an indentation in the eye disc, typically known as the morphogenetic furrow. A second phase of high cell proliferation, the second mitotic wave, is closely associated with the differentiation of the photoreceptor cells (Ready et al., 1976;

Tomlinson, 1985). Developmental events relating to the morphogenetic furrow have been

described in much detail for the compound eye development in Drosophila (Curtiss and

Mlodzik, 2000; Ma and Moses, 1995; Tomlinson, 1985; Wolff and Ready, 1991), but also for the adult Tribolium castaneum (Friedrich et al., 1996). Conversely, a morphogenetic furrow-like event of cell differentiation could not be observed during Thermonectus marmoratus ommatidial development (Sbita et al., 2007). Similar to the ommatidial development in Drosophila sp.

(Ready et al., 1976; Tomlinson, 1985; Wolff and Ready, 1993), the eye anlagen in Thermonectus marmoratus are composed of a single pseudostratified layer of epithelial cells (Figure 13).

However, we found no evidence for a morphogenetic furrow-like event during the stemmatal development, the apparent absence of which is also shared by larval eye development in Lytta viridana (Heming, 1982) and Tribolium castaneum (Liu and Friedrich, 2004). Moreover, since proliferation of the cells takes place in two mitotic periods during the monolayer stage in the ommatidial development in Drosophila, we anticipated that proliferation occurs early in the development of Thermonectus marmoratus stemmata also. There is no confirmation for a larger mitotic event at any one point during their development that we sampled. Instead, low level cell proliferation seems to be a continuous process, which is perhaps similar to the first mitotic phase in Drosophila ommatidial development. This also applies to the stemmatal development in the larvae of the blister beetle Lytta viridana (Heming, 1982) and the flour beetle Tribolium castaneum (Liu and Friedrich, 2004), in which mitosis activity is generally low throughout the entire embryonic phase.

Photoreceptor cells and accessory cells

In Drosophila ommatidial development, the photoreceptor cells differentiate sequentially. The first photoreceptor cell that is added to the developing ommatidium is the photoreceptor cell R8, which will later assume a proximal central position in the fully developed ommatidium. Then, the future outer (peripheral) photoreceptor cells R2 through R6 are recruited to the ommatidial cluster in pairs (Cagan and Ready, 1989; Tomlinson, 1985). Once all eight photoreceptor cells are formed, they begin to project axons (Ready et al., 1976; Tomlinson, 1985). Likewise, during the stemmatal development in Thermonectus marmoratus, the cell types differentiate from the pseudostratified epithelium and assume their position in the individual stemmatal cell cluster in a certain sequence. Those photoreceptor cells that will be positioned centrally in later developmental stages of the stemmata (~32-48% of embryonic development; Figure 14, Figure

16) are morphologically differentiated first, followed by the prospective peripheral photoreceptor cells. The central photoreceptor cells will subsequently form the proximal retina, while the peripheral photoreceptor cells give rise to the distal retina. The beginning of retinal differentiation on a morphological level is apparent by ~48% of embryonic development (Figure

16h).

The development of the axons of the photoreceptor cells during the ommatidial formation in Drosophila is accompanied by the addition of the prospective cone cells to the ommatidial clusters, followed by the future pigment cells. As the cells of each ommatidial cluster begin to lengthen, the cone cells extend over the apical tips of the photoreceptor cells, thus causing the photoreceptor cells to lose contact with the epidermis. Basally, the proximal end of the cone cells and photoreceptor cells rest on a membrane that forms the bottom of the ommatidium. In

Thermonectus marmoratus, the order in which the cell types differentiate is similar to that in

Drosophila ommatidial patterning. In the former, the corneagenous cells are morphologically recognizable at a later stage (~38% of embryonic development) than the photoreceptor cells. The developing photoreceptors can be morphologically identified at ~32% of embryonic development on the basis of thin axonal projections from these cells (Figure 14b), and by 38% of embryonic development, the axons have formed distinct bundles that project into their respective optic lobes (Figure 14d). The corneagenous cells are groups of cells that lie between the photoreceptor cells of adjacent stemmata, and which appear to be continuous at the levels of the epithelial surface. At the base of the eye however, the proximally narrowed corneagenous cells enclose the photoreceptor cells and thus can be ascribed to a specific stemma (Figure 14d&e).

Starting at ~38% of embryonic development, central photoreceptor cells withdraw from the epidermis, which perhaps causes the peripherally-located photoreceptor cells to bend toward the central fissure in their respective stemma (Figure 14d, Figure 15). In addion, as the long axis of the stemmata changes from an anteroposterior orientation to a dorsoventral stack, the corneagenous cells move over the peripheral photoreceptor cells and separate the latter from the epidermis. While the stemmatal cells elongate, the corneagenous cells continue to build a sheath around the photoreceptor cells (Figure 16l, Figure 17b).

The rhabdomeric portion of the photoreceptor cells in the ommatidium of Drosophila develops at the apical surface of the cells beginning at about ~38% of embryonic development.

However, because the cone cells displace the distal end of the photoreceptor cells about 90° as the former elongate and rise over the latter, the microvilli ultimately project into the center of the ommatidium (Cagan and Ready, 1989). Similarly, the distal ends of the peripheral (distal)

photoreceptor cells in the stemmata of Thermonectus marmoratus are forced first into an oblique and finally into a horizontal orientation perhaps by the combined effect of corneagenous cells extending over the peripheral photoreceptor cells and central photoreceptor cells sinking below the peripheral photoreceptor cells. The rhabdom of the prospective distal retina begins to develop at the apical surface of the peripheral photoreceptor cells at a time at which the cells are oriented obliquely (Figure 16b), and eventually, the microvilli of the distal rhabdom are aligned approximately horizontally (Figure 18d, drh) (Mandapaka et al., 2006; Stecher et al., 2010).

Lens and crystalline cone-like structure

Accessory cells are involved in the formation of the lens and the cone in both the ommatidium and the stemma. In Drosophila ommatidial development, the lens is produced by the cone and the (primary and secondary) pigment cells as fine threads of a clear material that is secreted apically. The lens material is formed in the primary pigment cells and the centrally-located cone cells in larger amounts than in the secondary pigment cells which lie at the periphery of the ommatidia, resulting in a biconvex lens (Cagan and Ready, 1989). The development of the lens in the stemmata of Thermonectus marmoratus is particularly interesting in the light of the recent discovery that at least the lenses of E2 are bifocal (Stowasser et al., 2010). The crystalline cone- like structure is necessary to provide adequate spacing between the lens and the retina to ensure that the image is projected onto the rhabdoms. In T. marmoratus stemmatal development, both the lens and the crystalline cone-like structure originate from the corneagenous cells. At about halfway through the embryonic development, the apical region of the corneagenous cells is enriched with vesicles containing a clear substance (Figure 16c). These vesicles are still visible at a later stage when a thin lens is present (Figure 18b) but are no longer visible once the lens

increases in volume (Figure 18c). It is likely that the vesicles contain lens material, which is then secreted topically after the tissues have rearranged into dorsoventral layers. Additionally, the vesicle-rich region is deeper, thus producing more ‘lens material’, in centrally-located corneagenous cells than it is at the periphery (Figure 16d&e, h, k-m). Indeed, the thickness of the lens increases most in its center (Figure 18b-d).

Unlike the cellular crystalline cone-like structure in stemmata of Thermonectus marmoratus, the crystalline cone in the ommatidia in Drosophila is structurally a pseudocone, as it is composed of extracellular material that is secreted by the cone cells (Cagan and Ready,

1989). In the stemmata of Thermonectus marmoratus, the crystalline cone-like structure derives from a cytoplasmic differentiation in the corneagenous cells, and therefore is the possible equivalent of a eucone crystalline cone. The portion of the corneagenous cells that is situated above the distal rhabdom gives rise to the crystalline cone-like structure. This region stains lighter with ethyl gallate than the corneagenous cells on the periphery of the eye tube (Figure

18c&d). In addition, the cone area is lined by a small zone in the corneagenous cells that stains very darkly (Figure 18c, arrow). This apparent border presumably is an area of a high density of organelles and pigment, which appear to have migrated out of the cone portion of the corneagenous cells.

The cone cells of adult ommatidia are thought to represent specialized corneagenous cells

(Snodgrass, 1993). Furthermore, cone cells can also develop into primary pigment cells (Paulus and Schmidt, 1978). It appears that the corneagenous cells in the stemmata of Thermonectus marmoratus indeed share characteristics with both cone cells and primary pigment cells of the

Drosophila ommatidium. Like cone cells and primary pigment cells, the corneagenous cells secrete the lens. The proximal portions of the cone cells as well as the corneagenous cells reach the bottom of the ommatidium or stemma, respectively, and the distal portion of both cell types contributes to the crystalline cone, although the actual process of cone formation differs. On the other hand, the corneagenous cells resemble pigment cells in that they contain pigment

(Mandapaka et al., 2006, 'pigmented cells').

Apart from the lens and the potentially refractive crystalline cone-like structure, the stemmata of Thermonectus marmoratus contain a third region of possibly refractive material. At

~48% of embryonic development of T. marmoratus, the distal retina start to secrete a clear substance that fills a narrow, elongated space above and between the top-most photoreceptor cells (Figure 16a, ellipse). This extracellular material is still evident at ~65% of embryonic development (Figure 18a, ellipse) and also after the crystalline cone-like structure is established

(Figure 18d, circle). In addition, Mandapaka et al. (2006a) described the crystalline cone-like structure of the first instar larva as being composed of elongated cells and extracellular material.

Thus, the function of this material could be to fill the space between the cone portion of the corneagenous cells and the distal rhabdom. On the other hand, in contrast to the proximally- located cells of the distal retina, the top-most cells do not meet in the center of the eye.

Consequently, the extracellular material could be the agent that enables the dorsal photoreceptor cells to separate. Interestingly, the extracellular space between photoreceptor cells in open rhabdom-type compound eyes have been shown to contain the protein ‘spacemaker’ (spam, also known as ‘eyes shut’). Spam plays a significant role in the physical separation of photoreceptor cells, and it is also conserved in species with closed-rhabdom ommatidia albeit not being

100

expressed in the eye itself (Zelhof et al., 2006). Perhaps the extracellular space between the

corneagenous cells and distal photoreceptor cells in Thermonectus marmoratus too contains the spacemaker protein.

Eye patch (EP)

The visual system of coleopteran larvae is typically composed of six pairs of stemmata (Paulus,

1986, 2000). However, some Dytiscidae have a seventh visual organ on either side of the head

(Gilbert, 1994; Günther, 1912; Mandapaka et al., 2006), which is termed the eye patch in the

larvae of Thermonectus marmoratus (Mandapaka et al., 2006). It has been suggested that this

additional (potentially) photosensitive region might represent a seventh stemma that was

arrested in its early development (Snodgrass, 1993).

In Thermonectus marmoratus, the eye patch is positioned medial to E1. It lacks a corneal

lens but possesses a non-tiered retina (Maksimovic et al., 2009; Mandapaka et al., 2006). The eye

patch also develops in close proximity to E1 (Figure 19a-c), and might in fact originate from this

stemma. More specifically, the eye patch potentially is formed from medially-located

corneagenous cells and photoreceptor cells of the dorsal portion of E1. Neighboring groups of

photoreceptor cells in the larval eye patch alternate in expressing either a long-wavelength opsin

(TmLW) or the UV opsin TmUV I. In contrast, in the larval principal eyes long wavelength-

expression is restricted to the distal retina while UV opsin expression (TmUV I and TmUV II)

only takes place in the proximal retina, and specifically TmUV I is only expressed in the dorsal

portion of the proximal retina (Maksimovic et al., 2009). It seems that the retina in the eye patch

combines the opsin expression patterns from distal retina and the dorsal portion of the proximal

101

retina. At the stage of stemmatal development in which the EP first appears (~38% of embryonic development), the distal and proximal retinas of E1 are not yet morphologically separated into dorsoventral layers but are next to each other (Figure 14d, Figure 15). It is possible that both the peripheral (later distal) photoreceptors and the central (later proximal) photoreceptors of the dorsal portion of E1 contribute to the formation of the eye patch. As the ventral portion of the eye patch migrates upwards to contact the epidermis (~48% of embryonic development, Figure

19d), the two photoreceptor cell populations might intertwine and become more or less evenly distributed, which could potentially account for the interspersed opsin expression patterns.

Furthermore, axons from the EP as well as E1 project into the same optic lobe. The innervation of the (morphologically still undifferentiated) photoreceptor cells is established before the EP develops (Figure 14b, ax). Consequently, it is likely that the photoreceptor cells which are part of the EP have branched off of E1. Lastly, in the pupa of Thermonectus marmoratus, the eye patch fuses with the degenerating E1 during early pupal development, which consequently suggests a developmental link between these two structures (Sbita et al., 2007).

Dytiscus (Günther, 1912) and Acilius (Patten, 1889) also possess a “pigmented, potentially light sensitive spot” adjacent to the stemmata which supposedly originates from one of the principal eyes (Günther, 1912). This “spot” likely corresponds to the eye patch in

Thermonectus marmoratus.

Ensuing research projects

During the past 25 years, vision research has largely focused on understanding the molecular basis of eye development. Among the invertebrates, compound eye development is especially well-studied in Drosophila, and many specification and differentiation events and contributing

102

molecular factors have been described in much detail. This vast body of research founded on the

Drosophila compound eye was made possible through early descriptive studies on morphological changes during its development (Cagan and Ready, 1989; Perry, 1968; Ready et al., 1976; Tomlinson, 1985; Waddington and Perry, 1960; Wolff and Ready, 1991).

Likewise, we anticipate that our present study on morphological changes during the stemmatal development in Thermonectus marmoratus will lay the groundwork for using molecular tools to deepen the understanding of stemmatal development, and to identify potential homologies to the compound eye. More specifically, morphological similarities between ommatidial development in Drosophila and the formation of the stemmata in Thermonectus marmoratus may help us to test the expression of eye-specific developmental genes that are known for Drosophila and which might be conserved in Thermonectus marmoratus also. For example, the signaling factor hedgehog (Hh) is involved in tissue patterning during the initiation of the morphogenetic furrow in Drosophila compound eye development, and it also regulates the expression of early eye-specific genes (Curtiss and Mlodzik, 2000). Based on our morphological findings we expect that if an Hh-like signaling molecule is required for tissue patterning in

Thermonectus marmoratus stemmata as well, its expression might be evident between 10% and

22% of embryonic development (Figure 13).

Moreover, the production of the lens material Crystalline (Janssens and Gehring, 1999) by cone cells and pigment cells in ommatidia of Drosophila is regulated by the transcription factor D-pax2 (Dziedzic et al., 2009). Finding homologies between lens development in

Drosophila and Thermonectus marmoratus would be especially valuable because it might enable

103

us to understand how the bifocal lens of E2 is formed. Apart from Thermonectus marmoratus, the possession of a bifocal lens thus far has not been shown for any other extant species

(Stowasser et al., 2010).

Another interesting experiment will be the application of antibodies to specimens in their late embryonic phase, with the goal of being able to describe the nature of the extracellular material that is located between the cone-portion of the corneagenous cells and the distal rhabdom (Figure 16a, Figure 18a, ellipse). A candidate element perhaps is the protein spacemaker. During eye development, spacemaker is secreted into the extracellular space between photoreceptor cells and forces these cells to separate by expanding the inter- rhabdomeral space. Spacemaker is essential for the formation of an open rhabdom, such as that of Drosophila ommatidia, and although it is not expressed in eyes of the closed rhabdom-type it is genomically conserved in such species (Zelhof et al., 2006). The rhabdom of Thermonectus marmoratus larvae is that of the closed type in all but the dosal-most portion of the distal retina

(Figure 7d, arrow). Thus, the extracellular material between the dorsal photoreceptor cells could potentially contain the spacemaker protein. Given the right tools, the stemmatal development in

Thermonectus marmoratus might provide valuable insight into homologies between the eyes of holometabolous insect larvae and adult compound eyes beyond morphological similarities.

104

Conclusion

The body of work that the Buschbeck lab has produced on the larval eyes of the sunburst diving beetle is substantial, but the research potential on the extraordinary eyes of the larvae is far from being exhausted. It is clear that this animal is outstanding, and with every study appear more interesting aspects.

My examination of the ultrastructure of the retinas in the principal E1 and E2 revealed that the larvae are possibly sensitive to the polarization of light (Stecher et al., 2010). Today we know that this phenomenon is a common visual feature among invertebrates, and according to

Gilbert (1994), polarization sensitivity should even be expected for many holometabolous insect larvae simply because their stemmata are of the rhabdomeric type. So what then is to be learned from Thermonectus marmoratus stemmata? We have learned that the different retinas in each of the stemmata potentially have their specific tasks. This was also suggested in the original study on the larval eyes (Mandapaka et al., 2006) , but considering the data on the ultrastructure I propose that, specifically, polarization sensitivity is the task of the proximal retina. Polarization sensitivity is thought to be an ancestral trait, and that polarization insensitive retinas must have been secondarily modified (Horváth and Varjú, 2004). For example, in honey bees, the rhabdomeres in all polarization insensitive ommatidia are rotated along the long axis of their photoreceptor cells (Wehner and Bernard, 1993; Wehner and Meyer, 1981). Although the rhabdom of the distal retina in Thermonectus marmoratus does not twist, the alignment of the microvilli is irregular (Stecher et al., 2010). It is possible that this is such a case of secondary reduction of polarization sensitivity, so that the distal retina would be available to fulfill a different task.

105

What could be the use of polarization sensitivity in the Thermonectus marmoratus larvae? Polarization sensitivity is widely utilized for navigation, intraspecific communication and object recognition. Polarization-guided orientation is perhaps the most common exploit of polarization sensitivity and is evident even in simple eyes (for a review, see Horváth and Varjú

(2004)), and thus is probably ancestral, but perhaps not relevant for Thermonectus marmoratus larvae. Intraspecific communication unlikely plays a role since the main goal of the larvae is to feed, and there is little to no intraspecific interaction. Thermonectus marmoratus larvae are efficient predators, and thus I tested their hunting performance under polarized and unpolarized light conditions, on prey that reveals distinct polarization features with polarized illumination. I have shown that with polarized light conditions, the larvae detect prey in a shorter amout of time and are more successful capturing prey than under unpolarized illumination. However, I was unable to unambiguously attribute these results to the state of polarization in my setup. That the larvae are indeed polarization sensitive has now been demonstrated in the Buschbeck lab through recent electrophysiological recordings from the proximal retina. In addition, the e-vector preference that I have predicted for the different cell types matches the results from these recordings. Nevertheless, the presence of polarization sensitivity in such highly derived eyes as the Thermonectus marmoratus stemmata potentially serves a more advanced purpose than navigation, which is considered a relative simple task (Gilbert, 1994). In order to navigate along a polarization pattern, a polarization sensitive individual would only have to align its polarization analyzer with the existing polarization pattern. For example, in the dorsal rim area of the honey bee, the axis of maximal stimulation of the polarization sensitive rhabdoms shifts across the dorsal rim area in a fashion that mirrors the polarization pattern of the sky. Thus, all polarization sensitive rhabdoms would be maximally stimulated when the bee faces the sun, and would be

106

minimally stimulated when the bee is oriented perpendicular to the sun. In order to navigate, the visual system would have to determine the strength of the stimulus (Labhart, 1980; Rossel, 1993;

Rossel and Wehner, 1982).

The visual system of Thermonectus marmoratus larvae is so sophisticated and so structurally different from compound eyes, and yet stemmata and compound eyes are considered to be homologous (Paulus, 1986, 2000; Paulus and Schmidt, 1978). It is in their development that similarities can be found. Comparing the structural changes in Thermonectus marmoratus stemmatal development with the well-described pattern formation of Drosophila ommatidia, I have shown that the origin of the eyes and individual cell differentiation events during either development are similar to one other. Using my morphological data as a reference, it will now be possible to test for homologies on a molecular basis.

107

Bibliography

Bernard GD, Wehner R (1977) Functional similarities between polarization vision and color vision. Vision Res 17:1019-1028

Blest AD, Hardie RC, McIntyre P, Williams DS (1981) The spectral sensitivities of identified receptors and the function of retinal tiering in the principal eyes of a Jumping Spider. Journal of Comparative Physiology 145:227-239

Boal JG, Shashar N, Grable MM, Vaughan KH, Loew ER, Hanlon RT (2004) Behavioral evidence for intraspecific signaling with achromatic and polarized light by cuttlefish (Mollusca : Cephalopoda). Behaviour 141:837-861

Brown PK, Brown PS (1958) Visual pigments of the Octopus and cuttlefish. Nature 182:1288- 1290

Buschbeck EK, Sbita SJ, Morgan RC (2007) Scanning behavior by larvae of the predacious diving beetle, Thermonectus marmoratus (Coleoptera : Dytiscidae) enlarges visual field prior to prey capture. Journal of Comparative Physiology a-Neuroethology Sensory Neural and Behavioral Physiology 193:973-982

Cagan RL, Ready DF (1989) The emergence of order in the Drosophila pupal retina. Developmental Biology 136:346-362

Chiou TH, Kleinlogel S, Cronin T, Caldwell R, Loeffler B, Siddiqi A, Goldizen A, Marshall J (2008) Circular polarization vision in a stomatopod crustacean. Current Biology 18:429-434

Cronin TW (2006) Invertebrate vision in water. In: Warrant E, Nilsson D-E (eds) Invertebrate Colour Vision. Cambridge University Press, Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, Sao Paulo, pp 211-249

Cronin TW, Shashar N (2001) The linearly polarized light field in clear, tropical marine waters: Spatial and temporal variation of light intensity, degree of polarization and e-vector angle. J Exp Biol 204:2461-2467

Cronin TW, Shashar N, Caldwell RL, Marshall J, Cheroske AG, Chiou TH (2003) Polarization vision and its role in biological signaling. Integr Comp Biol 43:549-558

Crowson RA (1981) The Biology of the Coleoptera. Academic Press, London, Orlando, New York, San Diego, Austin, Boston, Tokyo, Sydney, Toronto

Curtiss J, Mlodzik M (2000) Morphogenetic furrow initiation and progression during eye development in Drosophila: the roles of decapentaplegic, hedgehog and eyes absent. Development 127:1325-1336

108

Dziedzic K, Heaphy J, Prescott H, Kavaler J (2009) The Transcription Factor D-Pax2 Regulates Crystallin Production During Eye Development in Drosophila melanogaster. Developmental Dynamics 238:2530-2539

Eguchi E, Waterman TH (1967) Cellular basis for polarized light perception in the spider crab, Libinia. Cell Tissue Res 84:87-101

Evans AV (2006) Field guide to Beetles of California. University of California Press., Berkeley, Los Angeles, London

Friedrich M, Rambold I, Melzer RR (1996) The early stages of ommatidial development in the flour beetle Tribolium castaneum (Coleoptera; Tenebrionidae). Development Genes and Evolution 206:136-146

Gilbert C (1994) Form and function of stemmata in the larvae of holometabolous insects. Annual Review of Entomology 39:323-349

Glas HW (1977) Models for unambiguous e-vector navigation in the bee. J Comp Physiol A 113:129-159

Goldsmith TH (1975) The polarization sensitivity - dichroic absorption paradox in arthropod photoreceptors. In: Snyder AW, Menzel R (eds) Photoreceptor optics. Springer-Verlag, Berlin ; New York, pp 392-409

Goldsmith TH (1977) Membrane adaptations of visual photoreceptors for the analysis of plane- polarized light. In: Castellani A (ed) Research in Photobiology. Plenum Press, New York, London, pp 651-658 Goldsmith TH, Philpott DE (1957) The microstructure of the compound eye of insects. Journal of Biophysical and Biochemical Cytology 3:429-438

Günther K (1912) Die Sehorgane der Larve und Imago von Dytiscus marginalis. Zeitschrift für wissenschaftliche Zoologie 100:60-115

Hartline HK (1949) Inhibition of activity of visual receptors by illuminating nearby retinal areas in the Limulus eye. Fed Proc (abstract), p 69

Hartline HK, Wagner HG, Ratliff F (1956) Inhibition in the eye of Limulus. J Gen Physiol 39:651-673

Heming BS (1982) Structure and development of the larval visual system in embryos of Lytta viridana LeConte (Coleoptera, Meloidae). Journal of Morphology 172:23-43

Horváth G (1995) Reflection polarization patterns at flat water surfaces and their relevance for insect polarization vision. J Theor Biol 175:27-37

109

Horváth G, Varjú D (2004) Polarized Light in Animal Vision: Polarization Patterns in Nature. Springer Verlag, Berlin, Heidelberg, New York

Janssens H, Gehring WJ (1999) Isolation and characterization of drosocrystallin, a lens crystallin gene of Drosophila melanogaster. Developmental Biology 207:204-214

Johnsen S (1994) Extraocular sensitivity to polarized light in an echinoderm. Journal of Experimental Biology 195:281-291

Johnsen S (2005) Visual ecology on the high seas. Mar Ecol Prog Ser 287:281–285

Kirchhoffer O (1908) Untersuchungen über die Augen pentamerer Käfer. Arch Biontol 2:237- 287

Labhart T (1974) Behavioral analysis of light intensity discrimination and spectral sensitivity in honey bee, Apis mellifera. Journal of Comparative Physiology 95:203-216

Labhart T (1980) Specialized photoreceptors at the dorsal rim of the honeybee's compound eye: polarizational and angular sensitivity. J Comp Physiol A 141:19-30

Labhart T (1996) How polarization-sensitive interneurones of crickets perform at low degrees of polarization. Journal of Experimental Biology 199:1467-1475

Land MF (1969a) Movements of retinae of Jumping Spiders (Salticidae - Dendryphantinae) in response to visual stimuli. Journal of Experimental Biology 51:471-&

Land MF (1969b) Structure of retinae of principal eyes of Jumping Spiders (Salticidae - Dendryphantinae) in relation to visual optics. Journal of Experimental Biology 51:443-&

Lerner A, Meltser N, Sapir N, Erlick C, Shashar N, Broza M (2008) Reflected polarization guides chironomid females to oviposition sites. J Exp Biol 211:3536-3543

Liu ZY, Friedrich M (2004) The Tribolium homologue of glass and the evolution of insect larval eyes. Developmental Biology 269:36-54

Lythgoe JN, Hemmings CC (1967) Polarized light and underwater vision. Nature 213:893-894 Ma CY, Moses K (1995) Wingless and patched are negative regulators of the morphogenetic furrow and can affect tissue polarity in the developing Drosophila compound eye. Development 121:2279-2289

Maksimovic S, Cook TA, Buschbeck EK (2009) Spatial distribution of opsin-encoding mRNAs in the tiered larval retinas of the sunburst diving beetle Thermonectus marmoratus (Coleoptera: Dytiscidae). Journal of Experimental Biology 212:3781-3794

Maksimovic S, Layne JE, Buschbeck EK (in press) The spectral sensitivity of the principal eyes of the Sunburst Diving Beetle Thermonectus marmoratus (Coleoptera: Dytiscidae) larva. 110

Mandapaka K, Morgan RC, Buschbeck EK (2006) Twenty-eight retinas but only twelve eyes: An anatomical analysis of the larval visual system of the diving beetle Thermonectus marmoratus (Coleoptera : Dytiscidae). Journal of Comparative Neurology 497:166-181

Marshall J, Cronin TW, Shashar N, Land M (1999) Behavioural evidence for polarisation vision in stomatopods reveals a potential channel for communication. Curr Biol 9:755-758

Marshall NJ, Messenger JB (1996) Colour-blind camouflage. Nature 382:408-409

McFall-Ngai MJ (1990) Crypsos in the pelagic environment American Zoologist 30:175-188

Meyer-Rochow VB (1975) The dioptric system in the beetle compound eyes. In: Horridge GA (ed) The compound eye and vision of insects. Clarendon Press, Oxford, p 595

Morgan RC (1992) Natural history, captive management and the display of the sunburst diving beetle Thermonectus marmoratus. AAZPA/CAZPA Annual Conference, pp 457–464

Morgan RC (1995) Sunburst diving beetle Thermonectus marmoratus biology, husbandry and display. . Invertebrate Captivity pp 50–57

Mäthger LM, Hanlon RT (2006) Anatomical basis for camouflaged polarized light communication in squid. Biology Letters 2:494-496

Nilsson DE, Kelber A (2007) A functional analysis of compound eye evolution. Arthropod Structure & Development 36:373-385

Nilsson DE, Labhart T, Meyer E (1987) Photoreceptor design and optical properties affecting polarization sensitivity in ants and crickets. Journal of Comparative Physiology a-Sensory Neural and Behavioral Physiology 161:645-658

Novales Flamarique I, Browman HI (2001) Foraging and prey-search behaviour of small juvenile rainbow trout (Oncorhynchus mykiss) under polarized light. pp 2415-2422

Novales Flamarique I, Hawryshyn CW (1997) Is the use of underwater polarized light by fish restricted to crepuscular time periods? Vision Research 37:975-989

Patten W (1889) Studies on the eyes of Arthropods. II Eyes of Acilius. Journal of Morphology 2

Paulus HF (1975) The compound eyes of apterygote insects. In: Horridge GA (ed) The compound eye and vision of insects. Clarendon Press, Oxford, p 595

Paulus HF (1986) Evolutionswege zum Larvalauge der Insekten - ein Modell für die Entstehung und die Ableitung der ozellären Lateralaugen der Myriapoda von Fazettenaugen. Zool Jb Syst 113:353-371

111

Paulus HF (2000) Phylogeny of the Myriapoda-Crustacea-Insecta: a new attempt using photoreceptor structure. Journal of Zoological Systematics and Evolutionary Research 38:189- 208

Paulus HF, Schmidt M (1978) Evolutionswege zum Larvalauge der Insekten: Die Stemmata der Trichoptera and Lepidoptera. Zeitschrift Fur Zoologische Systematik Und Evolutionsforschung 16:188-216

Perry MM (1968) Further studies on development of eye of Drosophila melanogaster. I. Ommatidia. Journal of Morphology 124:227-&

Ready DF, Hanson TE, Benzer S (1976) Development of Drosophila retina, a neurocrystalline lattice. Developmental Biology 53:217-240

Rossel S (1993) Navigation by bees using polarized skylight. Comp Biochem Physiol 104:695- 708

Rossel S, Wehner R (1982) The bee's map of the e-vector pattern in the sky. PNAS 79:4451- 4455

Sabbah S, Barta A, Gal J, Horvath G, Shashar N (2006) Experimental and theoretical study of skylight polarization transmitted through Snell's window of a flat water surface. Journal of the Optical Society of America a-Optics Image Science and Vision 23:1978-1988

Sabbah S, Shashar N (2006) Polarization contrast of zooplankton: A model for polarization- based sighting distance. Vision Research 46:444-456

Sbita SJ, Morgan RC, Buschbeck EK (2007) Eye and optic lobe metamorphosis in the sunburst diving beetle, Thermonectus marmoratus (Coleoptera : Dytiscidae). Arthropod Structure & Development 36:449-462

Schwind R (1984) The plunge reaction of the backswimmer Notonecta glauca. J Comp Physiol A 155:319-321

Schwind R (1991) Polarization vision in water insects and insects living on a moist substrate. J Comp Physiol A 169:531-540

Shashar N, Hagan R, Boal JG, Hanlon RT (2000) Cuttlefish use polarization sensitivity in predation on silvery fish. Vision Res 40:71-75

Shashar N, Hanlon RT (1997) Squids (Loligo pealei and Euprymna scolopes) can exhibit polarized light patterns produced by their skin. Biol Bull 193:207-208

Shashar N, Hanlon RT, Petz AD (1998) Polarization vision helps detect transparent prey. Nature 393:222-223

112

Shashar N, Rutledge PS, Cronin TW (1996) Polarization vision in cuttlefish - A concealed communication channel? Journal of Experimental Biology 199:2077-2084

Smola U, Tscharntke H (1979) Twisted rhabdomeres in the dipteran eye. Journal of Comparative Physiology 133:291-297

Smola U, Wunderer H (1981) Fly rhabdomeres twist invivo. Journal of Comparative Physiology 142:43-49

Snodgrass RE (1993) Principles of Insect Morphology. Cornell University Press

Snyder AW, McIntyre P (1975) Polarisation sensitivity of twisted fused rhabdoms. In: Snyder AW, Menzel R (eds) Photoreceptor optics. Springer-Verlag, Berlin ; New York, pp 388-391

Stecher N, Morgan R, Buschbeck EK (2010) Retinal ultrastructure may mediate polarization sensitivity in larvae of the Sunburst diving beetle, Thermonectus marmoratus (Coleoptera: Dytiscidae). Zoomorphology (Berlin) 129:141-152

Stowasser A, Rapaport A, Layne JE, Morgan RC, Buschbeck EK (2010) Biological Bifocal Lenses with Image Separation. Current Biology 20:1482-1486

Strausfeld NJ, Seyan HS (1985) Convergence of visual, haltere, and prosternal inputs at neck motor neuron of Calliphora erythrocephala Cell Tissue Res 240:601-615

Tomlinson A (1985) The cellular dynamics of pattern formation in the eye of Drosophila. Journal of Embryology and Experimental Morphology 89:313-331

Tuchin VV (2007) Tissue optics: light scattering methods and instruments for medical diagnosis. Society of Photo-Optical Instrumentation Engineers

Tuthill JC, Johnsen S (2006) Polarization sensitivity in the red swamp crayfish Procambarus clarkii enhances the detection of moving transparent objects. Journal of Experimental Biology 209:1612-1616

Velasco J, Millan VH (1998) Feeding habits of two large insects from a desert stream: Abedus herberti (Hemiptera: Belostomatidae) and Thermonectus marmoratus (Coleoptera: Dytiscidae). Aquat Insects 20:85-96

Via SE, Forward RB, Jr. (1975) The Ontogeny and Spectral Sensitivity of Polarotaxis in Larvae of the Crab Rhithropanopeus harrisi (Gould). Biological Bulletin 149:251-266 von Frisch K (1949) Polarisation des Himmelslicht als orientierender Faktor bei den Tänzen der Bienen. Experientia 5:142-148 von Frisch K, Lindauer M (1956) The "Language" and Orientation of the Honey Bee. Annu Rev Entomol 1:45-58 113

Waddington CH, Perry MM (1960) The ultrastructure of the developing eye of Drosophila. Proc R Soc Lond B 153:155-178

Waterman TH (1975) The optics of polarization sensitivity. In: Snyder AW, Menzel R (eds) Photoreceptor optics. Springer-Verlag, Berlin ; New York, pp 339-371

Wehner R (2001) Polarization vision - a uniform sensory capacity? J Exp Biol 204:2589-2596

Wehner R, Bernard GD (1993) Photoreceptor twist - a solution to the false-color problem. PNAS 90:4132-4135

Wehner R, Bernard GD, Geiger E (1975) Twisted and non-twisted rhabdoms and their significance for polarization detection in the bee. J Comp Physiol 104:225-245

Wehner R, Labhart T (2006) Polarisation vision. In: Warrant E, Nilsson D-E (eds) Invertebrate Vision. Cambridge University Press, Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, Sao Paulo, pp 291-348

Wehner R, Meyer E (1981) Rhabdomeric twist in bees — Artefact or in vivo structure? Journal of Comparative Physiology A: Neuroethology, Sensory, Neural, and Behavioral Physiology 142:1-17

Wehner R, Strasser S (1985) The POL area of the honey bee's eye: behavioral evidence. Physiol Entomol 10:337-349

Williams DS, McIntyre P (1980) The principal eyes of a Jumping Spider have a telephoto component. Nature 288:578-580

Wolff T, Ready DF (1991) The beginning of pattern formation in the Drosophila compound eye - the morphogenetic furrow and the 2nd mitotic wave. Development 113:841-&

Wolff T, Ready DF (1993) Pattern formation in the Drosophila retina. Cold Spring Harbor Laboratory Press

Zelhof AC, Hardy RW, Becker A, Zuker CS (2006) Transforming the architecture of compound eyes. Nature 443:696-699

114