1 Online Resource 2 3 Sea lamprey orient toward a source of a synthesized pheromone using odor- 4 conditioned rheotaxis 5 6 Nicholas S. Johnson,a,b Azizah Muhammad,c Henry Thompson,a Jongeun Choi,c* 7 and Weiming Lia* 8 9 a Department of Fisheries and Wildlife, Michigan State University, Room 13 Natural 10 Resources Building, East Lansing, MI 48824, USA 11 b 12 Road, Millersburg, MI 49759,USA 13 c 14 Building, East Lansing, MI 48824, USA USGS, Great Lakes Science Center, Hammond Bay Biological Station, 11188 Ray Department of Mechanical Engineering, Michigan State University, 2459 Engineering 15 16 *Corresponding Authors: e-mail: liweim@anr.msu.edu, Phone: 517-432-6705, Fax: 517- 17 432-1699 and e-mail: jchoi@egr.msu.edu, Phone: 517-432-3164, Fax: 517-353-1750 1 18 Methods 19 Synthesized 3kPZS, permits, and animals 20 3kPZS was custom synthesized by Bridge Organics, Inc. (Vicksburg, MI, USA). 3kPZS 21 was analyzed for chemical identity and purity using mass spectrometry and nuclear 22 magnetic resonance at Michigan State University. Purity was greater than 97%. 23 Application of synthesized 3kPZS to the Ocqueoc River (Millersburg, MI, USA) was 24 approved by United States Environmental Protection Agency through experimental use 25 permit 75437-EUP-2. 26 Pre-ovulatory female sea lampreys (Petromyzon marinus) were captured by 27 agents of the United States Fish and Wildlife Service, Marquette Biological Station, MI, 28 USA, and Department of Fisheries and Ocean Canada, Sea Lamprey Control Centre, ON, 29 Canada, in mechanical traps from tributaries to Lakes Huron and Michigan. Sea 30 lampreys were transported to United States Geological Survey, Hammond Bay Biological 31 Station, Millersburg, MI, USA, and maintained in 1000 L flow through tanks supplied 32 with Lake Huron water at ambient temperatures (6 to 15 oC). Sea lampreys placed in 33 holding cages to mature were checked at minimum every two days. When ovulated, 34 females were returned to Hammond Bay Biological Station for tagging procedures. 35 36 Visual observation of females, 3kPZS plume structure, water velocity mapping, and 37 female movement track overlay 2 38 To visually observe movement tracks of females, stream transects were physically 39 marked with synthetic string every 5 m from 110 m downstream of the pheromone 40 sources to within 10 m of pheromone sources. Transects were located every 2.5 m within 41 10 m of the pheromone sources. Along each transect, electrical tape was attached to the 42 string at 1 m intervals (Fig. S2). Transects and transect meter marks were transposed to 43 stream maps drawn on graph paper where a square equaled 1 m2. During 44 experimentation, a researcher would visually observe and describe the movement of a 45 lamprey to another researcher that was recording the lamprey’s position on the stream 46 map. The cross-sectional location of the female and experimental time was recorded 47 whenever a female crossed a visually marked stream transect, where experimental time 48 was defined as the time since the female was released. Therefore, the bifurcated stream 49 channel contained 26 visual stream transects and at minimum, the location and time of 50 each female was recorded during 26 events. These events occurred at different times 51 depending on the upstream swimming speed of the female. Additional movement 52 locations and times were recorded when the upstream swimming speed of the female 53 permitted. Thus, sampling times were non-uniform among individuals. 54 To trace 3kPZS distribution in the bifurcated test system, three dye tests were 55 conducted: 1) dye was applied in both channels on 30May07, 2) dye was applied to the 56 left channel on 31May07, and 3) dye was applied to the right channel on 31May07. 57 Rhodamine was the 3kPZS surrogate and was applied to reach a final in-stream 58 concentration of 1 x 10-8 M when mixed with the discharge from both channels using the 59 same approaches and equipment as for 3kPZS application. After a 10 min dye 3 60 introduction period, water samples were collected in 5 ml glass vials in transects across 61 the stream at every 0.5 m. Stream transects were located in the same position as transects 62 used for behavioral observation. The florescence intensity of each sample measured at 63 556 nm was determined in a luminescence spectrometer (Perkin Elmer LSS55, Downers 64 Grove, IL, USA) and rhodamine concentration was estimated using a standard curve (R2 65 = 0.9998). 66 Dye concentration and water velocity throughout the experimental test systems 67 were modeled by a Monotonic Piecewise Cubic Hermite Interpolation Polynomial 68 (Fritsch and Carlson 1980). Stream maps, dye distribution maps, and water velocity 69 maps were produced in Python (Version 2.4, http://www.python.org/ Copyright © 1990- 70 2006, Python Software Foundation) and ovulated female movement tracks were overlaid 71 in Python Imaging Library (Version 1.1.6 http://www.pythonware.com/ products/pil/ 72 Copyright © 1997-2006 by Secret Labs AB & Copyright © 1995-2006 by Fredrik Lundh, 73 Publisher: Secret Labs AB) and exported to Photoshop (Version CS2) for final display. 74 Dye map color contours were exponentially scaled to match the back calculated 75 concentration of 3kPZS in the stream. 76 77 Algorithm pseudo-code 78 79 The following symbols and parameters were used in the control algorithms: 4 80 3kPZS. z0 10 81 82 83 84 z 0 was the minimum concentration value at which sea lampreys exhibit a response to 14 M (Johnson et al. 2009). In odor-conditioned rheotaxis, zth was the threshold concentration value for the sea lamprey to determine if it was within the odor plume. In klinotaxis, z max (k ) was a memory function based on the last measured maximum 85 concentration. It was used to determine if the sea lamprey was directed toward 86 increasing concentration. 87 88 89 zt arg et was the threshold concentration value for the sea lamprey to determine if the current position was the source of the pheromone. 0 1 was the threshold for the sea lamprey to determine if it was directed 90 toward increasing concentration due to klinotaxis. The value was related to the 91 resolution of the chemo-sensory input and was chosen as 1014 (Johnson et al. 92 2009). 93 In klinotaxis, the forgetting factor 0 0.5 1 was used for the sea lamprey to 94 discount z max (k ) such that the half of the maximum value in the previous iteration 95 was remembered when the sea lamprey did not find a good direction. 96 v was the forward velocity of the sea lamprey. This quantity was calibrated for each 97 algorithm to yield results most similar to observed movements using least squares 98 optimization. 99 vmag (k ) was the flow velocity magnitude when the sea lamprey is at ( x(k ), y(k )) . 5 100 flow(k ) was used in rheotaxis and odor-conditioned rheotaxis to determine the 101 direction of the odor source. It is parallel to the flow but in the upstream direction, i.e. 102 180o plus the flow direction. 103 d (k ) was the desired heading angle of control effort of the algorithms. 104 In odor-conditioned rheotaxis, 1 and 2 defined the degree by which the sea 105 lamprey was allowed to deviate from flow(k ) in two cases: before the first 106 detection of the odor plume ( 1 ) and when it was within the plume ( 2 ). These 107 parameters were chosen such that the sea lamprey was headed in an upstream 108 direction in both of these cases; during plume reacquisition, the sea lamprey is 109 allowed to orient itself downstream when need to re-enter the plume. 1 2 so 110 that the sea lamprey is able to explore a wider area before the first detection of the 111 plume than when it is within it. 112 1 and 2 were the gains for the obstacle avoidance (all algorithms) and odor- 113 conditioned rheotaxis strategy. (k ) was the number of degrees the sea lamprey 114 rotated to avoid an obstacle or reacquire the odor plume and these gains determined 115 by how much. 1 and 2 were calibrated for each algorithm for the best results. 116 117 118 119 T 0.5 sec was the sampling time determined by the period of the sea lamprey’s inhalation (Kleerekoper and Sibakin 1956). wx (k ) ~ N (0, x2 ) , w y (k ) ~ N (0, 2y ) , and w (k ) ~ N (0, 2 ) were random variables from the environment and control effort modeled by the Gaussian white 6 120 noise. These parameters were chosen as x 0.1 m, y 0.1 m, and 0.1 m 121 (Chang et al. 1998). 122 In klinotaxis, N 0 and N1 were the number of iterations the sea lamprey was 123 allowed to keep its current direction before exploring a new one. When the odor was not 124 present, the lamprey would use N 0 . When the odor was present, except when the 125 concentration was increasing, the lamprey would use N1 . This sequence was inspired by 126 the “run” and “tumble” observed in bacterial klinotaxis (Adler 1966; Block et al. 1982). 127 128 Control algorithm 1: Rheotaxis (Fig. S3) 129 INITIALIZE 130 I1. Let the initial position and heading angle be ( x(1), y(1)) and (1) respectively. 131 I2. Measure vmag (1) . 132 I3. if vmag (1) 0 then: sea lamprey selects a random direction for flow(1) 133 else: flow(1) is measured at ( x(1), y(1)) end if 134 I4. Let (0) 0 and d (0) flow(1) . 135 I5. Let k 1 . 136 137 138 139 FIND ODOR SOURCE S1. if obstacle is encountered then: d (k ) is chosen such that the obstacle is avoided end if S2. (State update) x(k 1) x(k ) Tv cos (k ) wx (k ) 7 140 y(k 1) y(k ) Tv sin (k ) wy (k ) 141 (k 1) d (k ) w (k ) 142 S3. Measure vmag (k 1) . 143 S4. if vmag (k 1) 0 then: flow(k 1) flow(k ) 144 else: flow(k 1) is measured at ( x(k 1), y(k 1)) end if 145 S5. Increment k by 1 146 S6. Perform steps S1-S5. 147 148 Control algorithm 2: Odor-conditioned rheotaxis (Fig. S4) 149 INITIALIZE 150 I1. Let the initial position and heading angle be ( x(1), y(1)) and (1) respectively. 151 I2. Measure z (1) and vmag (1) . 152 I3. if vmag (1) 0 then: flow(1) is random direction 153 else: flow(1) is measured at ( x(1), y(1)) end if 154 I4. Let (0) 0 . 155 I5. if z (1) zth then: d (0) is random direction in 156 [ flow(1) 1, flow(1) 1 ] 157 else: d (0) is random direction in [ flow(1) 2 , flow(1) 2 ] 158 I6. Let k 1 . 8 159 FIND ODOR PLUME 160 P1. while sea lamprey is not in the odor plume ( z (k ) zth ) do: 161 P2. if obstacle is encountered then: d (k ) is chosen such that the obstacle is avoided 162 163 164 end if P3. if (k ) flow(k ) 1 AND (k ) flow(k ) 1 then: d (k ) flow(k ) end if 165 P4. (State update) x(k 1) x(k ) Tv cos (k ) wx (k ) 166 y(k 1) y(k ) Tv sin (k ) w y (k ) 167 (k 1) d (k ) w (k ) 168 P5. Measure z (k 1) and vmag (k 1) . 169 P6. if vmag (k 1) 0 then: flow(k 1) flow(k ) 170 171 172 else: flow(k 1) is measured at ( x(k 1), y(k 1)) end if P7. Increment k by 1 end while TRACK PLUME 173 T1. while sea lamprey has not located the odor source ( z (k ) zt arg et ) do: 174 T2. if sea lamprey exits odor plume ( z (k ) zth ) then: go to step R1 175 176 177 else: go to step S3 end if T3. if obstacle is encountered then: d (k ) is chosen such that the obstacle is avoided else: d (k ) d (k 1) and (k ) 0 end if 9 178 179 T4. if d (k ) flow(k ) 2 OR d (k ) flow(k ) 2 then: d (k ) flow(k ) end if 180 T5. Perform steps P4-P6. 181 T6. Increment k by 1 end while 182 REACQUIRE PLUME 183 R1. while sea lamprey is not in odor plume ( z (k ) zth ) do: 184 R2. if obstacle is encountered then: d (k ) is chosen such that the obstacle is 185 avoided 186 else: go to step R3 end if 187 R3. if sea lamprey exited odor plume on right side then: 188 d (k ) d (k 1) (k 1) and (k ) 1013 z (k ) zth 189 else: d (k ) d (k 1) (k 1) and (k ) 1013 z (k ) zth 190 end if 191 R4. Perform steps P4-P6. 192 R5. Increment k by 1 end while 193 As can be seen in the description of odor-conditioned rheotaxis, the sampling time 194 was chosen as a period of inhalation, which is simplistic and might limit the scope of the 195 model set. Particularly, with evidence that sea lamprey make decisions based on averaged 196 measurements over multiple periods of inhalations, the model needs to be extended 197 appropriately. The scope of the model set can be extended by selecting a sampling time 198 as a multiple periods of inhalations and replacing the single measurement with the 10 199 average of measurements over the multiple periods. This will produce another discrete- 200 time system with a larger sampling time and new measurement noise with a smaller 201 variance due to averaging. With this new model, the exact same performance evaluation 202 procedure as described in this paper can be performed. 203 204 Control algorithm 3: Klinotaxis (Fig. S5) 205 INITIALIZE 206 I1. Let the initial position and heading angle be ( x(1), y(1)) and (1) respectively. 207 I2. Measure z (1) . 208 I3. Let z max (1) 0 , (0) 0 and d (0) is random direction. 209 I4. Let k 1 . 210 DETECT ODOR 211 O1. while the sea lamprey does not detect any odor ( z (k ) z0 ) do: 212 O2. if obstacle is encountered then: d (k ) is chosen such that the obstacle is avoided 213 214 end if O3. if d has been the same direction for the last N 0 iterations then: 215 d (k ) is random direction and (k ) 0 216 else: d (k ) d (k 1) and (k ) 0 end if 217 O4. (State update) x(k 1) x(k ) Tv cos (k ) wx (k ) 218 y(k 1) y(k ) Tv sin (k ) wy (k ) 11 (k 1) d (k ) w (k ) 219 220 O5. z max (k 1) max{ z (k ), z max (k )} 221 O6. Measure z(k 1) . 222 O7. Increment k by 1 end while 223 FIND ODOR SOURCE 224 S1. while sea lamprey has not located the odor source ( z (k ) zt arg et ) do: 225 S2. if obstacle is encountered then: d (k ) is chosen such that the obstacle is avoided 226 end if 227 else: go to step S2 end if 228 229 S3. if 3kPZS concentration is not increasing ( z (k ) z max (k ) ) then: if sea lamprey does not detect odor ( z (k ) z0 ) then: 230 if d has been the same direction for the last N 0 iterations then: 231 d (k ) is random direction and (k ) 0 232 else: d (k ) d (k 1) and (k ) 0 end if 233 else: 234 if d has been the same direction for the last N1 iterations then: 235 d (k ) is random direction and (k ) 0 236 else: d (k ) d (k 1) and (k ) 0 end if 237 238 end if else: d (k ) (k 1) and (k ) 0 end if 12 239 S4. Perform steps O3-O5. 240 S5. if z (k ) z max then: z max (k 1) z max (k ) 241 else: z max (k 1) z (k ) end if 242 S6. Measure z(k 1) . 243 S7. Increment k by 1 end while 244 245 In the control algorithms the obstacle avoidance algorithm was written as 246 if obstacle is encountered then: d (k ) is chosen such that the obstacle is avoided end if. 247 The details of the obstacle avoidance algorithm is given as follows. 248 249 250 251 1) while obstacle is encountered do: 2) if obstacle was encountered at (k-1)th iteration then: go to step 3 else: go to step 4 end if 3) (k 1) (k 2) 252 4) if (k ) flow (k ) then: (k 1) 1 253 else: (k 1) 1 end if 254 5) d (k ) d (k 1) (k 1) 255 6) Increment k by 1 end while 256 When the obstacle avoidance algorithm is activated, the first step is to decide which 257 direction the lamprey must turn to swim away from the obstacle. To do this, we first 258 determine whether or not obstacle avoidance was active in the previous iteration. We do 259 this because the direction in which the lamprey must turn to avoid the obstacle has 260 already been decided, and we want to avoid the possibility of turning back into the 261 obstacle. In this case, we use the same turning direction until the lamprey is clear of the 262 obstacle. In the case that the obstacle avoidance was not previously active, we use the 13 263 lamprey’s heading direction relative to the flow direction to determine the turning 264 direction. Obstacle avoidance overrides all other algorithms, because a higher priority is 265 placed on clearing obstacles. The obstacle avoidance algorithm is deactivated when the 266 sea lamprey has cleared the obstacle. 267 268 Orientation of female sea lampreys to a source of 3kPZS in a flow and reduced-flow 269 environment 270 271 Test system 272 A PIT antenna system with multiplexer recorded the number of females that moved 273 upstream during a trial. Cross channel PIT antennas were located 2.5 m upstream of the 274 confluence of the major and minor channel and at the furthest upstream point of the 275 minor channel (Fig. S6). A 0.5 m2 PIT antenna was placed around the 3kPZS application 276 location which was located 14 m upstream of the female release cage. 277 278 Procedures 279 Flow and reduced-flow experiments were not conducted at random because of the 280 difficulty of replicating fine-scale flow characteristics when switching the test stream. 281 Flow experiments were conducted from 12Jun08 to 17Jun08, reduced-flow experiments 282 were conducted from 18Jun08 to 26Jun08, and flow experiments were conducted again 283 on 27Jun08 and 28Jun08. A coin was flipped to determine whether 3kPZS or control 284 solvent was applied as the test odorant under each flow condition. Six trials were 14 285 conducted with flow and 3kPZS application, 4 trials were conducted with flow and 286 control solvent application, 8 trials were conducted with reduced-flow and 3kPZS 287 application, and 4 trials were conducted with reduced-flow and control solvent 288 application. Between 8 and 10 ovulated females were released per trial. Individual 289 females were not exposed to more than one treatment because once ovulated, female 290 lifespan is typically less than 7 days. 291 Under flow and reduced-flow conditions 3kPZS was applied at a rate to achieve a 292 fully mixed in-channel concentration of 5 x 10-13 M. It was assumed that under reduced- 293 flow conditions that 25 L/sec of water passed through the channel due to leakage of the 294 wing-dam and subsurface water flow. The amount of synthesized 3kPZS required for an 295 experiment was dissolved in 1 ml of methanol, mixed with 25 L of river water, and 296 applied to the stream at a rate of 167 ml/min via peristaltic pump through a 4.8 mm inside 297 diameter tube over a 2.5 h application period. Control solvent consisted of 1 ml of 298 methanol mixed with 25 L of river water applied to the stream at a rate of 167 ml/min for 299 2.5 h. 300 Ovulated females were pre-exposed to the test odorant 30 min prior to release and 301 were observed for 2 h after release. Movement tracks of individual sea lampreys were 302 manually observed and recorded on a stream map referencing a 0.5 m2 stream grid 303 system as described above (Johnson et al. 2009). The number of females leaving the 304 release cage, moving upstream through the downstream PIT antenna (move upstream), 305 entering inside the 0.5 m2 PIT antenna around the odorant source (visit odor), and 306 swimming through the upstream PIT antenna without visiting the odorant source (pass 15 307 odor) was confirmed by the PIT tag system. Significant differences among treatments in 308 the above listed metrics were determined with logistic regression models. Data were also 309 analyzed with mixed effect logistic regression with a random effect of trial date. All 310 statistical results from logistic regression models are robust to the inclusion of the 311 random effect of trial date, supporting the assumption that a single ovulated female can 312 be treated as an individual sample (Siefkes et al. 2005). Statistical results reported are 313 from two-tailed analyses conducted in R (R Development Core Team 2009). 314 315 3kPZS plume structure, water velocity mapping, and female movement track overlay 316 317 In the minor channel the reduced-flow condition dye test was conducted on 26Jun08 and 318 the flow condition dye test was conducted on 30Jun08. Under flow and reduced-flow 319 conditions, rhodamine was applied to reach a final in-stream concentration of 5 x 10-8 M 320 (assuming a discharge of 25 L/sec under no-flow conditions) using the same approaches 321 and equipment as used for 3kPZS application. Rhodamine was applied to the 322 experimental channel 30 min prior to sampling. Water samples were collected from the 323 middle of the water column in 5 ml glass vials in transects across the channel at every 0.5 324 m. Channel transects were located every 2.5 m starting at the furthest downstream 325 portion of the channel extending to 5 m upstream of the odorant application location. 326 Dye concentration and water velocity throughout the experimental test systems was 327 measured, modeled, and displayed as described earlier. 328 329 16 330 Online Resource References 331 Adler J (1966) Chemotaxis in bacteria. Science 153:708-716 332 Block SM, Segall JE, Berg HC (1982) Impulse responses in bacterial chemotaxis. Cell 333 334 335 336 337 31:215-226 Chang H-G, Freeman WJ, Burke BC (1998) Biologically modelled noise stabilizing neurodynamics for pattern recognition. Int J Bifurcat Chaos 8:321-345 Fritsch FN, Carlson RE (1980) Monotone Piecewise Cubic Interpolation. SIAM J Numer Anal 17:238–246 338 Johnson NS, Yun S-S, Thompson HT, Brant CO, Li W (2009) A synthesized pheromone 339 induces upstream movement in female sea lamprey and summons them into traps. P 340 Natl Acad Sci USA 106:1021-1026. doi:10.1073/pnas.0808530106 341 Kleerekoper H, Sibakin K(1956) Spike potentials produced by the sea lamprey 342 (Petromyzon marinus) in the water surrounding the head region. Nature 178:490- 343 491. doi:10.1038/178490b0 344 R Development Core Team (2009) R: a language and environment for statistical 345 computing. R Foundation for Statistical Computing. Vienna, Austria. ISBN: 3- 346 900051-07-0. http://www.R-project.org. Accessed 25Aug09 17 347 Siefkes MJ, Winterstein JS, Li W (2005). 3-keto petromyzonol sulfate specifically 348 attracts ovulating female sea lamprey (Petromyzon marinus). Anim Behav 70:1037- 349 1045 18 350 Online Resource Figure Captions 351 Fig. S1 Water velocity in bifurcated stream channel where sea lamprey orientation 352 behaviors were observed. Color coding indicates the estimated water velocity through 353 the stream channel. Passive integrated transponder (PIT) antennas were placed around 354 the 3kPZS sources which were located 45 m upstream of the channel confluence and 355 cross channel PIT antennas were placed 1 m upstream of the channel confluence (white 356 lines). PIT antennas determined how many lampreys entered the left and right stream 357 channels and located the left and right source of 3kPZS 358 Fig. S2 Photograph of the downstream section of the bifurcated stream showing stream 359 transects used for behavioral observation and dye sampling 360 Fig. S3 Flowchart for control algorithm 1 (rheotaxis) 361 Fig. S4 Flowchart for control algorithm 2 (odor-conditioned rheotaxis) 362 Fig. S5 Flowchart for control algorithm 3 (klinotaxis) 363 Fig. S6 Water velocity of novel experimental system under flow (a) and reduced-flow 364 conditions (b). Passive integrated transponder antennas located at 5 m, 20 m, and around 365 the odor application location at 15 m (O). Females were released from the cage at the 1 366 m mark (RC). Color coding indicates the estimated water velocity through the stream 367 channel 19 368 369 370 Fig. S1 20 371 372 Fig.S2 21 373 START INITIALIZE FIND ODOR SOURCE 374 375 376 Fig. S3 22 START INITIALIZE z (k ) z th NO YES FIND PLUME NO z (k ) zth YES TRACK PLUME z (k ) zth YES NO NO z (k ) zt arg et YES REACQUIRE PLUME NO END 377 378 379 Fig. S4 23 z (k ) zth YES START INITIALIZE NO z (k ) z0 YES DETECT ODOR NO z (k ) z0 YES FIND ODOR SOURCE YES END 380 381 382 Fig. S5 24 z (k ) zt arg et NO 383 (a) (b) 384 385 Fig. S6 25