Effects of habitat and pulp and paper mill contamination on a population of brook stickleback (Culaea inconstans)

Downstream of pulp and paper effluent discharge, deleterious effects on fish populations such as reductions in sex steroid hormone levels, gonad size and fecundity, changes in secondary sex characteristics, and delayed maturity have been observed (Munkittrick et al. 1992; Sandström & Neuman 2003; McMaster et al. 2006; van den Heuvel 2010). A review of two decades of research on reproductive effects of pulp mill effluents on fishes attributes such effects to all types of mill processes, such as bleached kraft, bleached sulphite, mechanical, multiprocess, and thermomechanical (Hewitt et al. 2008).

Despite decades of research, and the accumulation of considerable knowledge on the mechanism of reproductive effect, understanding of the causative agents of those effects is not conclusive. It is known that compounds with androgenic and oestrogenic activity, likely derived from wood, have been detected in pulp mill effluents (van den Heuvel et al. 2010). Reproductive impacts in fish may also be attributed to neuroendocrine disruption (Basu et al. 2009) or occur indirectly via nutritional deficiencies (van den Heuvel et al. 2010). Several recent studies have also consistently demonstrated an association between effluent biochemical oxygen demand (BOD) and decreased fathead minnow (Pimephales promelas) spawning success (Martel et al. 2011; Kovacs et al. 2011, 2013).

In 1987, Jackfish Bay, Lake Superior was deemed an area of concern under the Canada–United States Great Lakes Water Quality Agreement due to the effluent it received from a bleached kraft pulp mill in nearby Terrace Bay, Canada (JBRAP 1998). Monitoring efforts over two decades (1988–2007) revealed that effluent-exposed wild white sucker (Catostomus commersoni) consistently have smaller gonad sizes, reduced circulating sex steroid levels, increased liver size and condition factor (CF), increased age, delayed maturity, and increased CYP1A induction, when compared with unexposed wild white sucker (Munkittrick et al. 1991, 1992, 1994; Bowron et al. 2009). This pattern of metabolic disruption in Jackfish Bay is consistent with a national pattern of response from over 200 fish surveys conducted in three cycles of the Canadian Environmental Effects Monitoring (EEM) program between 1992 and 2004 (Barrett et al. 2010). As the Terrace Bay mill added secondary treatment facilities and converted to elemental chlorine-free (ECF) bleaching, some of the observed effects in white sucker became less pronounced (Bowron et al. 2009). For example, gonad size differences and liver size differences between wild-exposed and wild-reference white sucker were reduced although differences in the condition remained consistent. CYP1A induction, although reduced, remained significantly higher in wild-exposed than in reference white sucker.

Despite the extensive studies on effluent impacts in the Lake Superior-receiving environment, no examination of fish populations has been conducted in the receiving environment upstream of where effluent enters Lake Superior. This environment includes 14 km of stream, the origin of which is the effluent discharge, including a number of small lakes. The settlement of solids in this region has likely contributed to high levels of legacy contaminants such as chlorinated dioxins (Sherman et al. 1990). In such a depositional environment, the microbial production of the polycyclic aromatic hydrocarbon (PAH) retene – a known inducer of CYP1A enzymes in fish (Fragoso et al. 1998; Billiard et al. 1999; Brinkworth et al. 2003) – via anaerobic degradation in lake sediments is expected (Rämänen et al. 2010).

The main objective of this study was to examine for physiological and biochemical responses in a small forage fish, the brook stickleback (Culaea inconstans), exposed to effluent from the Terrace Bay pulp and paper mill. It was hypothesized that in small depositional lakes receiving high concentrations of pulp mill effluent, sediment contamination might result, including the production of the PAH retene. It was further hypothesized that given much higher effluent load when compared with the Lake Superior-receiving environment, biochemical, physiological and reproductive impacts may be detectable in this population of brook stickleback. These hypotheses were examined by measuring sediment chemistry and a number of standard environmental variables in three reference locations and in the receiving environment upstream of Lake Superior. Brook stickleback population, physiological, and biochemical parameters were measured in the receiving environment and the three reference locations between 2007 and 2013.

Moberly Lake contained a population of brook stickleback at the time the study was initiated, and received effluent from the bleached kraft mill located in Terrace Bay, Ontario, Canada (Figure 1). During mill operational periods, effluent flow in 2012 averaged 85,000 m³ per day, reduced from about 102,000 m³ per day in 2008 and, during operational years between 2007 and 2014, pulp output averaged 326,000 air-dried metric tonnes per annum 2008 (Traci Bryar, Environmental Superintendent at AV Terrace Bay, personal communication). The Terrace Bay Mill uses ECF bleaching and effluent treatment consists of a primary treatment system that removes fibres and suspended solids and a secondary treatment system of an aerated lagoon (Munkittrick et al. 1992). The aerated lagoon has an 8–10-day retention time and upon its implementation, it significantly reduced BOD (by 95%), TSS (by 29%), and AOX (by 29%) (Munkittrick et al. 1992).

The mill discharges its effluent into Lake Superior via Blackbird Creek (13.7 km). Prior to release into Lake Superior, the effluent passes through a number of small lakes – the largest and final lake in the system is Moberly Lake, about 1.9 km in length. In the early 1940s, Lake A was 6.1 m at its deepest point and covered an area of about 19 ha (JBRAP 1998). Over time, the deposition of wood fibres from the mill caused the lake to be partially filled in, redirecting Blackbird Creek around it and into Moberly Lake in the early 1980s (JBRAP 1998). Prior to redirecting Blackbird Creek, Moberly Lake was approximately 28 ha and 6.4 m at its deepest point (JBRAP 1998). Since 1994, maximum depth has decreased to 5 m and the area covered has also been reduced (JBRAP 1998). Since 2006, the mill has experienced three long-term shutdowns: for 7 months in 2006, for 19 months in 2009–2010, and again for 10 months in 2011–2012 (Figure 2).

Three sites were chosen in addition to Moberly Lake and used as reference sites for this study. The smallest site is an unnamed pond in a small tributary of Blackbird Creek not receiving the effluent and isolated from the main branch of Blackbird Creek by beaver impoundment; this site will henceforth be referred to as Highway Pond. Highway Pond was used as a reference site in spring 2011. Minnow Lake (0.09 km²) is a shallow lake, similar to Moberly Lake (0.27 km²), and is also on the Blackbird Creek system about 3 km upstream of the main branch of the creek. The brook stickleback population in Minnow Lake was isolated from the Blackbird Creek water through a series of beaver impoundments and unfavourably fast water. Minnow Lake was used as a reference site in spring 2011. Dead Horse Creek is a more stream-like environment with a beaver impoundment containing brook stickleback in a separate watershed approximately 20 km away from the other sites. It was used as a reference site in spring 2007 and 2011.

In May 2011, TidbiT v2 UTBI-001 temperature loggers (Onset, Cape Cod, MA, USA) were deployed at 1–1.5 m water depth (at the time of snowmelt) in Moberly Lake, Dead Horse Creek, and Minnow Lake (the population at Highway Pond was only discovered at this time and a temperature logger was not available to deploy). Loggers were programmed to record the temperature every 4 h. Loggers were cased in an ABS plastic plumbing tee fitted with two permanent adapters with openings smaller than the TidbiT unit to allow for water flow. This entire system was attached to a concrete block with aircraft cable, and the block was attached to a solid tree on shore via a chain.

Sediment cores for chemical analysis were collected in 2007 with a 10 cm plastic tube. Five cores were collected in the south part of Moberly Lake adjacent to the brook stickleback capture area. As the 2007 analysis was meant only as an exploratory evaluation of the concentration of sediment extractives at a known pulp mill-contaminated site, samples were collected from reference locations. Sediment cores were again collected in Moberly Lake in 2010 and Minnow Lake was added in order to examine levels of retene, the major extractive in the region, in an environment not receiving pulp and paper effluent. For both sampling periods, a sediment pool for analysis was made from the five cores using the top 10 cm of each core.

For the 2007 analysis, the freeze-dried sediment was mixed with granular sodium sulphate (BDH, UK) at a 1:1 ratio (w:w) and then ground with a mortar and pestle. The mixture was spiked with 10 μL of recovery standard containing d10-anthracene as a neutrals surrogate, dihydrocholesterol as a sterol surrogate and 8(14)-abietenic acid as a resin acid surrogate. The sediment was extracted using an Agilent 7680T supercritical fluid extractor. Three successive extraction steps of 25 min using CO₂ at 227 bar and 70 °C and a flow rate of 3.0 mL/min were employed to extract the analytes. After each extraction step, the analytes were trapped on a C18 column and then eluted using dichloromethane that was dried over sodium sulphate. The residue was transferred to 250 μL glass inserts housed in 1.8 mL screw-cap autosampler vials (Alltech, New Zealand). After the addition of dibromoanthracene (TCI, Japan) in pyridine as an internal standard, the samples were derivatized by adding 50 μL of bis(trimethylsilyl)trifluoroacetamide (BSTFA) +1% trimethylchlorosilane silylation reagent (Alltech, New Zealand) and heating for 1 h at 70 °C.

Analysis of the extracts was performed on a 6890N gas chromatograph (GC) and 5973N electron impact ionization mass selective detector (Agilent Technologies, USA) acquiring data in the full scan (50–800 m/z) mode. An autosampler was used to introduce 1 μL of the sample into a purged splitless injector maintained at 280 °C. The analytes were separated on an Agilent Ultra-2 column (50 m × 0.2 mm ID, film thickness 0.33 μm) with helium as a carrier gas. The oven parameters used were 60 °C (1 min), 10 °C/min until 205 °C then 3 °C until 270 °C and 6 °C/min until 300 °C (45 min). Final concentrations were corrected for the recovery of surrogate standards.

In 2010, the detailed GC–MS analysis was no longer available to investigators and only the substance retene was analysed to provide a comparison. The sediment was dried by mixing with sodium sulphate and extracted for 24 h using Soxhlet extraction with dichloromethane. Samples were evaporated and made up to 1 mL in methanol. Retene was analysed by HPLC using a C-18 column with fluorescence detection (excitation 260 nm, emission 380 nm) and quantified against a retene standard (Sigma). Retene recovery using this method was found to be 95–98%.

Minnow traps (40 cm long, 20 cm in diameter, made of 0.5 cm galvanized mesh with 2.5 cm openings) were used as the primary method of capture. Minnow traps were baited with cat food that was free for fish to feed upon and set in the habitat that appeared desirable for nesting males (quantities of leaves and debris). Occasionally, electrofishing using a Smith-Root LR-24 Backpack Electrofisher was used to supplement minnow trapping in order to attain the required number of brook stickleback. Traps were left overnight and checked by 1 pm of the following day. In the spring seasons, capture occurred during the first week of May which was chosen as this is the earliest that snow and ice conditions would allow fish capture. In fall 2008 and 2011, the capture was attempted during the last week of September which was chosen as the end of the growth period to obtain a measure of gonadal development and prior to the period where ice would prevent fish capture.

Fish were transferred into 20 L buckets containing aerated water from the site of capture and transported a short distance to a sampling area. Fish were euthanized with a sharp blow to the head followed by spinal severance. Weight and fork length were recorded to the nearest tenth of g and mm, respectively. Livers were excised, weighed, and divided into two parts; approximately 20 mg of tissue was stored in 200 μL of RNA later and the remainder was flash frozen in liquid nitrogen. Gonads were also excised and weighed. Whole gonads were placed in histocassettes, fixed in 10% neutral-buffered formalin, and later transferred to 70% ethanol. Carcasses were placed in individual Whirl-Pak^® bags, flash frozen, and stored at −20 °C.

7-Ethoxyresorufin-O-deethylase (EROD) analysis was conducted using a modification of the fluorescence plate reader technique outlined by van den Heuvel et al. (1995) as a catalytic measure of CYP1A. The entire fish liver (∼50 mg) was homogenized in 500 μL of cryopreservative buffer (0.1 M phosphate, 1 mM EDTA, 1 mM dithiothreitol, and 20% glycerol, pH 7.4) using a sonic dismembrator (Branson SLPe, Connecticut, USA). The homogenized solution was then centrifuged at 9,000 × g to obtain the post mitochondrial supernatant (PMS) containing cellular proteins. The EROD reaction contained 0.1 M HEPES buffer (pH 7.8; Sigma), 5.0 mM Mg²⁺, 0.5 mM NADPH (Sigma), 1.5 M 7-ethoxyresorufin (Sigma), and 0.5 mg/mL of PMS protein. Reactions were allowed to occur for 10 min and then terminated with acetonitrile and read on a fluorescence plate reader (Bio-Tek FLx800; 530 nm excitation, 590 nm emission). Protein content was estimated from fluorescamine fluorescence (390 nm excitation, 460 nm emission filters) against a bovine serum albumin standard (Sigma) and activity corrected for protein concentrations.

Gonads were embedded in paraffin, sectioned along the coronal plane to about 4–6 μm thick, placed on slides, stained with haematoxylin and eosin, and then permanently mounted for viewing under a compound microscope. The stage of maturity was determined based on methods for fathead minnows published by USEPA (2006). All samples were coded to ensure blind analysis of histological data. All oocytes in the ovarian section were counted. Individual oocytes were categorized as being perinuclear, cortical alveolar, early vitellogenic, late vitellogenic, or mature.

All data for parametric analysis were evaluated for normality using normal probability plots. Weight, liver size, and gonad size were analysed using analysis of covariance (ANCOVA), with logarithmically transformed values to meet the assumptions for normality, using length (weight) or body weight (liver, gonad), plus the categorical treatment variables. Somatic data were expressed as indices for presentation purposes using the least square means and covariate means from the ANCOVA. Gonadosomatic index is defined as the percent of body weight that is made up of gonads, and liver somatic index is defined as the percent of body weight that is made up of liver. Condition factor is a ratio of fish body weight to length (k = 100 × weight/length³). ANOVAs were performed to detect differences among sites in EROD induction. Tukey's post hoc tests were conducted to make pairwise comparisons among sites when more than two sites were being compared. All ANOVAs and ANCOVAs were performed on SYSTAT version 13.0 (Systat Software, San Jose, CA, USA). To evaluate the overall pattern of the relative frequency of ovarian stages between sites, an ANOSIM based on the Bray–Curtis similarity of the relative proportions of each ovarian stage was performed using 999 permutations using PRIMER, v6 software (2006 PRIMER-E Ltd, Plymouth, UK). The critical level of statistical differences for all analyses was α = 0.05.

At Moberly Lake, DO was 60–95% saturated during the first week of May in 2011, 2012, and 2013 (Table 1). During the mill operational period in the last week of September 2011, DO decreased to 7%. DO was relatively consistent among references sites within seasons, usually falling between 70 and 115%, this is excepting Dead Horse Creek's super-saturation to 153% in May 2011 and Minnow Lake's low of 18.3% in May 2011 (Table 1). Conductivity in Moberly Lake was consistently higher (1.4–17 times) than reference sites during mill operational periods. Upon the mill becoming non-operational in December 2012 (Figure 2), Moberly Lake's conductivity fell to 152 μS/cm, similar to the mean diluent steam conductivity of 144 μS/cm (Table 1). Based on calculations, effluent concentrations in Moberly Lake were 28, 72, 57, and 25% in May 2011, September 2011, May 2012, and May 2013, respectively (Table 1).

At Moberly Lake, the water temperature was affected by whether or not the mill was operational. Generally, while the mill was operational, the temperature increased 3–5 °C above the temperature of reference sites. For example, during the non-operational period (2 December 2011 until 11 October 2012; Figure 3), Moberly Lake temperature was very similar to Minnow Lake during the summer months, although somewhat warmer than the more lotic Dead Horse Creek at this time. During the winter period of the shutdown from December 2011 to April 2012, Moberly Lake's temperature profile aligned more closely with that of Dead Horse Creek. During that winter period, Minnow Lake exhibited warmer ∼4 °C temperature typically found under ice (none of the systems studied freeze entirely). The presence of effluent during the operational period resulted in Moberly Lake temperatures being warmer in the winter of 2012/2013 compared with 2011/2012 by about 3 °C.

In 2007, Moberly Lake sediment samples contained 2,375.8 μg/g dry weight of resin acid neutrals, 270.5 μg/g dry weight of resin acids, and 406.8 μg/g dry weight of phytosterols (Table 2). The neutrals were the dominant extractives in this sediment, and the PAH retene comprised a majority of this at 1,625.1 μg/g dry weight. Abietic acid was the most dominant resin acid and sitosterol was the most dominant sterol. The neutrals are derived from the bacterial decomposition of resin acids and the sum total of the neutrals represents 91% of the total of both neutrals and resin acids, representing a substantial conversion of resin acids to neutrals. In 2010, the measured retene concentration in the Moberly Lake sediment was 833.0 μg/g. As a reference, Minnow Lake sediments contained 4.3 μg/g dry weight of retene.

Brook stickleback were generally more difficult to capture during the September capture efforts in 2008 and 2011 when compared with spring sampling periods in 2007 and 2011. While fish were still present at all three reference sites in September 2011, exhaustive minnow trapping and seine netting efforts at Moberly Lake in September 2011, May 2012, and May 2013 indicated that brook stickleback were no longer present in the effluent-exposed site. Sampling effort was carried out over 3 days with more than 90 overnight minnow trap sets along the full 2 km of Moberly Lake at each sampling period. In addition, at least 20 seine net hauls were made without a single stickleback captured. This absence corresponded with the low DO levels measured during September 2011 (7% saturation) and considering the temperature in September had already dropped 10 °C from the high of nearly 25 °C experienced that summer, extended periods of anoxia were likely. In May 2012, brook stickleback remained highly abundant in Minnow Lake. While stickleback were no longer present in Moberly Lake and unidentified species of cyprinid, thought to be a species of dace, was captured in May of 2012 and 2013, but not in September 2011.

EROD analysis showed inconsistent results between the two sampling years (Figure 4). Moberly Lake male brook stickleback had 1.9-fold higher EROD induction than the pooled reference fish in 2011; however, in 2007, Dead Horse Creek male brook stickleback had 2.5-fold higher EROD induction compared with fish from Moberly Lake. The trend in females was more consistent; Moberly Lake females had higher EROD induction in both sampling years (2.3-fold higher in 2011 and 1.4-fold higher in 2007). However, this elevation in females was only significantly different in 2011.

Gonad size as it covaries with fish weight in females from Moberly Lake, the effluent-receiving site, was significantly higher compared with fish from the reference sites in 2007 and 2011, with the exception of Minnow Lake in May 2011 (Figure 5(a)). Minnow Lake brook stickleback gonad size was significantly higher than Moberly Lake females in 2011. There were no significant differences in gonad size in Moberly Lake male brook stickleback when compared with any of the other populations during 2007 or 2011. However, liver size as it covaries with body weight in male brook stickleback was consistently higher in fish from Moberly Lake, the effluent-receiving area, in both May 2007 and May 2011 (Figure 5(d)). The same pattern was observed for female fish (Figure 5(c)), except when compared with Minnow Lake in May 2011 where there was no significant difference in liver size between the Minnow Lake and Moberly Lake populations. While females from Moberly Lake consistently had the highest body weight as it covaries with body length, this difference was only statistically significant in 2011 (Figure 5(e)). Males did not show any consistent differences for body condition (Figure 5(f)). In both males and females, the average fork length of both male (Figure 5(h)) and female (Figure 5(g)) brook stickleback in Moberly Lake was significantly greater than the fork length of brook stickleback in Dead Horse Creek in 2007. No significant differences in fork length were seen among any fish samples from 2011, except that Minnow Lake males were significantly shorter than males from all other sites in 2011. Stickleback typically only lives for 2 years; thus, fork length of these individuals represents their maximum growth.

An ANOSIM on the relative frequency of ovarian development stages of female fish showed that all sites were significantly different from one another with the exception of Dead Horse Creek and Highway Pond (Figure 6). Of the four populations, ovaries from Minnow Lake females were the most developed and different from the other site with global R-values (1 being no similarity between groups and 0 indicating that groups are identical) of 0.852, 0.761, and 0.404 when compared with the Dead Horse Creek, Highway Pond, and Moberly Lake populations, respectively. Ovaries from the Moberly Lake females also had significantly advanced maturation when compared with Dead Horse Creek and Highway Pond with global R-values of 0.456 and 0.169 for these comparisons, respectively. Only Dead Horse Creek and Highway Pond, the least developed of the sites, were not significantly different, with a global R-value of 0.062.

The discharge of the bleach kraft pulp mill in Terrace Bay resulted in high levels of the PAH retene in a depositional zone although brook stickleback liver CYP1A levels did not consistently correspond with the elevated sediment retene. Brook stickleback from the two shallow lake environments were reproductively advanced, had greater condition factor, and larger livers than fish from the more riverine environment, suggesting that responses in fish are related to warmer water temperatures due to effluent load in a low flow system rather than the chemical makeup of the effluent. The only consistent response in fish from the effluent-receiving environment was the greater liver size. The latter part of the study documented a complete collapse of the Moberly Lake brook stickleback population that followed a summer hypoxic period.

Retene is present in the sediment of Moberly Lake at levels approaching the highest reported in any receiving environment and CYP1A was inconsistent with this contamination. At over 1,600 μ/g dw, retene levels in Moberly Lake in 2007 are within the range of values (18.4–3,300 μg/g dw) previously reported at comparably contaminated sites (Leppänen & Oikari 1999a, 1999b, 2001; Lahdelma & Oikari 2005). Under anaerobic conditions, abietene resin acids are almost completely converted to retene (Tavendale et al. 1997a, 1997b), and in Moberly Lake sediment, they are 91% converted to neutrals. Sedimentary retene has been demonstrated as bioavailable in other circumstances (Leppänen & Oikari 1999a, 1999b; Oikari et al. 2003). The CYP1A-inducing ability of retene has been very well established in laboratory studies (Fragoso et al. 1998; Billiard et al. 1999; Brinkworth et al. 2003). From 1990 to 2000, Jackfish Bay white sucker showed significantly higher EROD levels compared with Mountain Bay (reference) white sucker (Bowron et al. 2009). Throughout 1990–2000, these differences declined steadily but remained significant (Bowron et al. 2009). Male white sucker captured in September 2000 still showed eight-fold EROD induction when compared with the reference location (Bowron et al. 2009). The directionally opposite response in male stickleback sampled in 2007 could be due to a refractory CYP1A phenotype in the exposed brook stickleback population. In a study by Meyer et al. (2002), killifish in a contaminated site showed a refractory CYP1A phenotype (decreased EROD activity). However, CYP1A response can also be altered during periods of reproductive development, which were variable across sampling years.

The patterns in somatic indices seen in Moberly Lake brook stickleback were inconsistent with the national response pattern of fish populations exposed to pulp and paper mill effluent. This pattern, observed in meta-analysis of data collected under the auspices of the EEM program, has been termed ‘metabolic disruption’ (Munkittrick et al. 1991). The pattern is characterized by higher condition factor and liver size, reflecting a more nutrient-enriched environment, but lower relative gonad size, reflecting an inability to metabolize stored energy into reproductive tissue. This response profile has been consistently observed in white sucker from Jackfish Bay between 1988 and 2006 with trends towards recovery.

Endpoints relative to reproduction were generally opposite that observed in both Jackfish Bay white sucker and in locally relevant fish at other pulp and paper mills in the past. The larger gonads in fish from the lake environments, coupled with much higher proportions of developed ovarian follicles, are due to differences in reproductive timing. Stickleback species undergo much of their gonadal development in the period just prior to spawning when compared with the large-bodied synchronous spawning white sucker that undergoes most of gonadal recrudescence in the fall and hold the follicles until final maturation in spring. The timing of reproductive development of monitoring species can substantially influence the conclusions. For example, rainbow darter (Etheostoma caeruleum) exposed to municipal sewage effluent demonstrated a different pattern of response in the spring than in the fall (Fuzzen et al. 2016). Brook stickleback exposed to municipal wastewater also demonstrated varying patterns of response between reference and exposed sites dependent on the month they were sampled (Tetreault et al. 2012). In that study, the May/June reproductive period was when the largest differences were exhibited between sampling locations although this also appeared to be a result of reproductive timing as was likely in the study described herein. The rapid gonad development of stickleback species makes it difficult to compare the relative energy put into reproductive tissue as vitellogenesis and egg maturation are occurring simultaneously within a short period. The nature of the reproductive behaviour of the two species may also impact the energetics of gonadal development. Sucker do not prepare or build and guard nests as stickleback species and as such may put more energy into gonadal tissue as eggs and fry may have a lower probability of survival when compared with species that have greater parental investment and care.

Temperature differences among sites throughout the year could explain the variable reproductive development between sites. Because of warmer temperatures during winter and into spring, fish living in a shallow lake environment matured earlier (Minnow Lake and Moberly Lake) than those in colder riverine environments, given that warmer waters play a major role in inducing maturation. In 2011, Moberly Lake was the warmest of the sites at spawning time. The presence of effluent appears to warm this lake by as much as 5° during the remainder of the year. These observations suggest that, in the winter prior to the 2011 spawning season, Moberly Lake's temperature was warm enough to encourage earlier reproductive development in brook stickleback. It could certainly be suggested that sampling brook stickleback in the spring is not ideal; however, unlike the white sucker, this species may not have a stable post-vitellogenic sampling window where fish gonad size is temporally stable, making brook stickleback less suitable for monitoring purposes (Barrett & Munkittrick 2010).

Changes in mill operation and treatment/processing of the effluent can improve effluent quality and reduce potency in its ability to induce reproductive changes in fishes. Brook stickleback were exposed to effluent for 6 and 8 months prior to the 2007 and 2011 sampling periods, respectively, in addition to being exposed to contamination in the sediment as described herein. White sucker captured in Lake Superior showed the strongest evidence of improved reproductive development during periods of mill closure (Bowron et al. 2009). In a laboratory study by Rickwood et al. (2006), exposure to 100% effluent from the Terrace Bay mill had no effect on gonad size of fathead minnow. While the number of spawning events was decreased, total egg production was not (Rickwood et al. 2006). In addition to the complexity of comparing different species between laboratory and dissimilar receiving environments, pulp mill effluent can vary temporally depending on current mill operations or treatment system upsets (Martel et al. 2011).

Condition factor in short-lived fishes may be more influenced by life history than food availability. There was no consistent effect of effluent exposure on condition factor as has been observed for in white sucker for the two decades from 1988 to 2007 (Bowron et al. 2009). Brook stickleback are thought to live for a single year or slightly longer (Acere & Lindsey 1986; King & Cone 2008). During this time, all available energy would go into growth in order to best maximize fecundity. Contrast this with a large-bodied, long-lived species such as the white sucker where the largest part of growth is completed in 3–4 years prior to maturation, after which growth gradually slows. Such species put proportionally less energy into growth; therefore, condition factor may change more dramatically in response to increased resource availability as may occur downstream of a pulp and paper mill effluent discharge (Environment Canada 2012, 2014). However, as the present study did not examine food availability, the influence of such cannot be ruled out here.

The liver size of brook stickleback from the effluent-receiving environment was clearly elevated compared with fish from reference sites. Minnow Lake and Moberly Lake females being further along in their maturity likely had engorged livers due to increased production of vitellogenin. However, the liver size in males is less confounded by reproductive development, and Moberly Lake males had consistently larger livers. Larger liver sizes in effluent-exposed fish when compared with unexposed fish is consistent with exposure to pulp mills in general (Munkittrick et al. 1994) and was observed in Jackfish Bay white sucker (1988–1993 females, 1988–2000 males), but inconsistent with the more recent trends in Jackfish Bay white sucker (1995–2006 females, 2002–2006 males; Bowron et al. 2009). In more recent years, there were no discernible differences in the relative liver sizes between Jackfish Bay white sucker and Mountain Bay white sucker. Unlike condition factor, liver size can respond to both increased energy intake (van den Heuvel et al. 2008) and the increased metabolic burden of contaminants including PAHs (Phalen et al. 2014).

The anoxia that occurred in Moberly Lake is the most probable reason for the eradication of brook stickleback. DO levels in Moberly Lake (0.7 mg/L and 7%) were lower than LC50 values for oxygen reported in Landman et al. (2005) for a range of both sensitive and resilient fish species. In that study, six of seven species had LC50 higher than 0.7 mg/L, and only the short-finned eel Anguilla australis had an LC50 lower than 0.7 (0.54 mg/L). The counterpoint to the hypothesis of hypoxia-related disappearance was the presence of an unidentified cyprinid species in 2012 and 2013 (although it was also absent in 2011). One possibility is that this species had a refuge from hypoxia not utilized by brook stickleback. There was a short ∼10 m fast-moving section of stream at the downstream end of Moberly Lake prior to a steep ∼15 m elevation drop that may have provided sufficient oxygenation to act as a refuge.

The combination of high temperature and large amounts of effluent passing through with a high BOD would have significantly reduced the amount of oxygen available. The furthest upstream portion of Moberly Lake is often observed to bubble, likely due to the significant amount of anaerobic bacterial activity. This is not the first time fish have disappeared from effluent-exposed areas in this basin. Before secondary treatment was installed in 1989, all populations of fish in Blackbird Creek were completely eliminated due to the effluent's acute toxicity (JBRAP 1991). Minnow trapping in 2007 by the investigators found brook stickleback at a number of locations throughout the creek. Since 1989, the addition of secondary treatment facilities monthly LC50 tests conducted as required by the regulatory regime indicate that the effluent from the outflow pipe is no longer acutely toxic (JBRAP 1991). However, a toxicity event from a spill or treatment failure cannot be ruled out.

Brook stickleback were able to thrive in a fairly harsh environment of both pulp mill effluent and severe historic sediment contamination. This species is well known to utilize beaver impoundments habitats, warm, organically enriched environments not dissimilar to the pulp mill-receiving environment that may be subject to periodic hypoxia. While there is no published experimental data for brook stickleback hypoxia tolerance, it might be assumed that they are relatively insensitive to hypoxia due to the nature of the environments they inhabit, and thus they can also inhabit shallow boreal forest lakes where few predators can survive (and, in fact, they are seldom found in any abundance where those fish predators exist). These characteristics, together with a short-lived life history, may make them less responsive to the types of physiological changes that manifest in larger longer-lived fishes in response to anthropogenic stress. While much work has been done on sublethal impacts of contaminants, particularly on reproduction, it is somewhat telling in this case that the local population may most likely have been eliminated by the acute effects of oxygen deprivation.

This study was financially supported by an NSERC Strategic Projects Grant to Dr. Kelly Munkittrick, and a Canada Research Chair held by Michael R. van den Heuvel. The authors wish to acknowledge Brad Scott, Travis James, Christina Pater, Colin Arens, Liane Leclair, Tim Arcizewski, Gerald Tetrault, and Mark McMaster for their assistance in the field.