Unexplained repeated pregnancy loss is associated with altered perceptual and brain responses to men’s body-odor
Abstract
Mammalian olfaction and reproduction are tightly linked, a link less explored in humans. Here, we asked whether human unexplained repeated pregnancy loss (uRPL) is associated with altered olfaction, and particularly altered olfactory responses to body-odor. We found that whereas most women with uRPL could identify the body-odor of their spouse, most control women could not. Moreover, women with uRPL rated the perceptual attributes of men's body-odor differently from controls. These pronounced differences were accompanied by an only modest albeit significant advantage in ordinary, non-body-odor-related olfaction in uRPL. Next, using structural and functional brain imaging, we found that in comparison to controls, most women with uRPL had smaller olfactory bulbs, yet increased hypothalamic response in association with men's body-odor. These findings combine to suggest altered olfactory perceptual and brain responses in women experiencing uRPL, particularly in relation to men's body-odor. Whether this link has any causal aspects to it remains to be explored.
Introduction
Mammalian reproduction is tightly linked to olfaction (Parkes and Bruce, 1961; Wysocki and Lepri, 1991; Boehm et al., 2005; Petrulis, 2013; deCatanzaro, 2015). This link has been extensively studied in rodents, where body-odors from adult males can lead to accelerated pubertal development in juvenile females (Vandenbergh, 1967), to synchronous estrous cycling in adult females (Whitten, 1956; Whitten, 1958), and to implantation failure (Bruce, 1959). Such social odors are typically transduced at the vomeronasal organ in the nose, and processed through the accessory olfactory system in the brain (Dulac, 2000; Luo and Katz, 2004), but there is also evidence for reproductive social odors acting through the main olfactory system as well (Kang et al., 2009; Baum and Cherry, 2015). Olfaction has also been linked to human reproduction. The best known, albeit controversial example of this is the phenomenon of menstrual synchrony. Here, women who cohabitate tend to have a synchronized menstrual cycle (McClintock, 1971), an effect mediated by women's body-odor (Stern and McClintock, 1998) (this effect has been challenged on statistical grounds (Strassmann, 1999), yet our view is that it holds). In a different set of related studies, the body-odor from breast-feeding women timed ovulation and menstruation in nulliparous women (Jacob et al., 2004), and increased their sexual motivation (Spencer et al., 2004). In turn, men’s body-odors were found to regulate menstrual cycle variability in women (Cutler et al., 1986; Preti et al., 2003). This is consistent with findings implying that women’s preference for men’s body-odors are menstrual cycle phase-dependent (Havlicek et al., 2005; Rantala et al., 2006). Moreover, maternal status was found to regulate neural responses to the body-odor of newborns (Lundström et al., 2013). Body-odors also impact reproduction relevant mechanisms in men. For example, attractiveness ratings applied by men to women's body-odors are menstrual-phase specific (Kuukasjarvi, 2004), with men rating women’s axillary and vaginal odors as most pleasant around ovulation (Doty et al., 1975; Singh and Bronstad, 2001; Gildersleeve et al., 2012). In addition, sniffing emotional tears collected from women donors reduced levels of testosterone (Gelstein et al., 2011; Oh et al., 2012) and ensuing arousal in men (Gelstein et al., 2011).
Given that the olfactory system is associated with human reproduction, we hypothesized that it may also be related to disorders of human reproduction. One such human reproductive disorder is unexplained repeated pregnancy loss (uRPL). A remarkable ~50% of all human conceptions, and ~15% of documented human pregnancies, end in spontaneous miscarriage (Rai and Regan, 2006). Only a limited portion of these miscarriages, however, can be explained (Clifford et al., 1994; Stephenson, 1996; Jaslow et al., 2010), suggesting the presence of major yet-unidentified human-miscarriage-related factors. Given the general link between olfaction and reproduction, we set out to test the hypothesis that olfaction is altered in uRPL. Moreover, because the role of olfaction in reproduction is typically related to body-odors, we used these as a primary target in our investigation.
Results
Whereas control women cannot identify their spouse by smell, women with uRPL can
A key olfactory ability investigated in the context of reproduction is the ability to identify mates by smell (Leinders-Zufall et al., 2004; Brennan and Zufall, 2006). To characterize olfactory spouse-recognition in humans, we tested 33 women with uRPL (mean age = 34.7 ± 6.4, mean number of miscarriages = 4.2) and 33 matched controls (mean age = 35.6 ± 4.1, never experienced miscarriage). We used a three-alternative forced- choice (3AFC) paradigm where on each trial the participant was asked to identify her spouse from three alternatives (delivered in specially designed body-odor sniff-jars, see Figure 1a); one containing the body-odor of her spouse, and two containing distractor odors; one from a non-spouse man and one Blank (carrier alone). Previous results on spouse odor identification in the general population are mixed, with two studies reporting ~33% group-level performance (Hold and Schleidt, 1977; Mahmut et al., 2019) but one study reporting ~74% performance (Lundström and Jones-Gotman, 2009). Here, consistent with the two former studies, as a group, control women were unable to identify their spouse above chance levels (chance = 33.3%, mean control = 28 ± 34%, Shapiro-Wilk test of normality (SW) = 0.77, p<0.001, difference from chance: Wilcoxon W = 206, p=0.18, effect size estimated by rank biserial correlation (RBC) = 1.17 (parametric comparison: t(32) = 0.88, p=0.39, Cohen’s d = 0.15)). In contrast, remarkably, uRPL women were able to identify their spouse by body-odor alone, at a level that was both significantly higher than expected by chance (chance = 33.3%, mean uRPL = 57 ± 41%, SW = 0.81, p<0.001), difference from chance: Wilcoxon W = 440, p=0.004, RBC = 1.71 (parametric comparison: t(32) = 3.32, p=0.002, Cohen’s d = 0.58), and two-fold that of controls (Mann-Whitney U = 332.5, p=0.005, RBC = 0.39, parametric comparison: t(64) = 3.12, p=0.003, Cohen’s d = 0.77) (Figure 1b). To further verify that this result was not a chance effects, we conducted a bootstrap test. We randomly shuffled the assignment of the data to uRPL or Control 10,000 times, and reconducted the analysis. We observe that the probability of obtaining our results coincidentally is 2 in 1000, further suggesting that this result was not a chance event (Figure 1c). We note that over time, we modified the task to use six rather than four 3AFC trials per participant (and three participants had only two 3AFC trials). However, this had no impact on group results, and when we reanalyze the entire dataset restricting to either four 3AFC trials per participant or two 3AFC trials per participant, the uRPL group remained significantly better than the control group in all cases (all Mann-Whitney U > 386.5, all p<0.026, all RBC > 0.29, parametric comparison: all t(64) > 2.4, all p<0.018, all Cohen’s d > 0.6). In other words, an increased tendency for miscarriage in women is associated with better identification of spouse body-odor at the group level.

Women with uRPL can identify their spouse by smell.
(a) A custom-designed Shirt Sniffing Device (SSD) to standardize body-odor sampling. The SSD consists of a glass jar containing the T-shirt, with an air intake port via soda lime filter, and air sampling port via one-way flap valve into individual-use airtight nose mask. This arrangement assured that environmental odors didn't contaminate the sample during the sampling process. The recognizable person in the figure is a co-author and not a participant. (b) Performance at 3AFC identification test (n = 66). uRPL women in purple, control women in green. Bar graphs depict group means, each dot represents a participant, dots are jittered to prevent overlay, error bars are s.e.m, **p<0.01. Black dashed line indicates chance level. (c) Bootstrap analysis. Gray lines represent the 10,000 repetitions; the purple line represents the actual uRPL-control difference value.
-
Figure 1—source data 1
All spouse identification values.
- https://cdn.elifesciences.org/articles/55305/elife-55305-fig1-data1-v1.xlsx
Women with uRPL have otherwise only marginally better olfaction than controls
In order to determine whether this advantage at identifying spouse body-odor was a reflection of a generally better sense of smell, we tested olfactory performance at several additional tasks. To test ordinary odor identification, 38 women with uRPL and 38 matched controls conducted a four-alternative forced-choice identification test using 20 every-day odorants such as soap and peanuts. We observed only a trend toward better identification in uRPL women than in controls (mean uRPL = 87.6% ± 7.2, Controls = 84.6% ± 7.3, uRPL SW = 0.91, p=0.005, control SW = 0.92, p=0.009, Mann-Whitney U = 560.5, p=0.087, RBC = 0.22 (parametric comparison: t(74) = 1.82, p=0.073, Cohen’s d = 0.42)) (Figure 2a). Subsequently, 38 women with uRPL and 38 controls were tested for detection of three odorant monomolecules that have often been studied in the context of human social chemosignaling (Jacob and McClintock, 2000). The monomolecules were Androstenone (ANN), Androstadienone (AND), and Estratetraenol (EST). We used each odorant at a fixed concentration, and tested its discrimination from blank using a 3AFC test. To account for the non-normal distribution of the data, a linear mixed-model with fixed effects of Group and Odor, with random intercept per participant and a random slope for odor within participant, revealed a main effect of Odorant (F(2,143) = 17.1, p<0. 0001), a modest effect of Group (F(1,74) = 3.97, p=0.05) and no interaction (F(2,143) = 0.27, p=0.77) (for comparison, a parametric repeated-measures analysis of variance (ANOVA) with factors of Group (RPL/Control) and Odorant (ANN/AND/EST) revealed a main effect of Odorant (F(2,148) = 17.5, p=1.5e−7), but only a trend toward a main effect of Group (F(1,74) = 3.28, p=0.074), and no interaction (F(2,148)=0.31, p=0.74) (Figure 2b)). The main effect of Odorant reflected that at these particular concentrations, EST was harder to detect than both AND and ANN (mean EST = 41.8 ± 29.4, mean AND = 58.1 ± 34.6, mean ANN = 66.2 ± 31: AND vs. ANN Wilcoxon W = 858, p=0.068, RBC = 0.29 (parametric comparison: t(70) = 1.7, p=0.093, Cohen’s d = 0.2); AND vs EST Wilcoxon W = 451, p=6.4e−4, RBC = 0.51 (parametric comparison: t(70) = 3.5, p=8.2e-4, Cohen’s d = 0.42); ANN vs. EST: Wilcoxon W = 237, p=6e−7, RBC = 0.74 (parametric comparison: t(75) = 5.86, p=1.2e−7, Cohen’s d = 0.67) (Figure 2b). The trend toward a main effect of Group reflected that uRPL participants again performed slightly better than controls, but this effect was near-significant for the odorant EST alone (EST: uRPL mean = 48.1 ± 30.3%, Control mean = 35.4 ± 27.4%, Mann-Whitney U = 544.5, p=0.065, RBC = 0.25 (parametric comparison: t(74) = 1.9, p=0.06, Cohen’s d = 0.44); AND: uRPL mean = 61.1 ± 32.5%, Control mean = 55.1 ± 36.8%, Mann-Whitney U = 585, p=0.6, RBC = 0.07 (parametric comparison: t(69) = 0.73, p=0.47, Cohen’s d = 0.17); ANN: uRPL mean = 71.1% ±27.1%, Control mean = 61.3 ± 34%, Mann-Whitney U = 587.5, p=0.15, RBC = 0.19 (parametric comparison: t(74) = 1.38, p=0.17, Cohen’s d = 0.32) (Figure 2b). Finally, a group of 18 women with uRPL and 18 controls were also available for a lengthy seven-reversal detection-threshold staircase paradigm using the alliaceous odorant dimethyl trisulfide (DMTS). We observed no significant difference between controls and uRPL (uRPL threshold, minus log concentration: 4.98 ± 0.41, control threshold: 5.06 ± 0.72, uRPL SW = 0.92, p=0.13, control SW = 0.9, p=0.052, t(34) = 0.42, p=0.68, Cohen’s d = 0.14) (Figure 2c). In reviewing the above results at olfactory identification, discrimination, and detection, women with uRPL repeatedly scored slightly better than controls, but this difference was not significant. For added power and sensitivity, in a final comparison we combined the above results, generating a composite olfaction score for each participant (reflecting her available scores at identification, discrimination, and detection) ranging between 0 and 100 (see Methods). This enabled a comparison between 39 women with uRPL and 39 matched controls. We observed a small but significant difference between the two groups (mean composite uRPL = 57.13% ± 15.23, Controls = 49.46% ± 16.14, uRPL SW = 0.97, p=0.49, control SW = 0.98, p=0.67, independent t-test t(76) = 2.16, p=0.034, Cohen’s d = 0.49) (Figure 2d). To verify that this was not a chance event, we again conducted a 10,000-repetion bootstrap like before, and observe that the probability of obtaining this result by chance is less than 2 in 100 (Figure 2e). We further tested for any relationship between this composite score and the overwhelmingly better olfactory identification of spouse in uRPL versus controls. We observed no correlation across all participants, but a trend in uRPL alone (uRPL group (n = 33): Spearman Rho = −0.33, p=0.06; Control group (n = 29): R = 0.06, p=0.76; Both groups together: R = −0.02, p=0.88). In conclusion, at the group level, women with uRPL were two-fold better than control women at identifying their spouse by smell, and marginally but significantly better than control women at ordinary olfactory tasks.

Women with uRPL have slightly better olfaction than controls.
uRPL (purple) and control (green) women were tested for various olfactory facets. (a) Percent accuracy at every-day odorant identification (n = 76). (b) Percent accuracy at monomolecule discrimination (EST, ANN, AND) (n = 76). (c) DMTS threshold (n = 36). (d) A composite score of identification, discrimination and threshold (n = 78). (e) Bootstrap test of ‘d’, 10,000 repetitions, the purple line represents the actual uRPL-control difference value. Bar graphs depict means, each dot represents a participant, dots are jittered to prevent overlay, error bars are s.e.m. *p<0.05. Black dashed line indicates chance level.
-
Figure 2—source data 1
All odourant identification, discrimination, and detection values, including composite score.
- https://cdn.elifesciences.org/articles/55305/elife-55305-fig2-data1-v1.xlsx
Women with uRPL have altered perception of men’s body-odor
Given that the advantage at identifying spouse by smell was unrelated to general olfactory abilities, we next asked whether it was reflected in any explicit perceptual olfactory attributes in body-odor. A cohort of 18 women with uRPL and 18 matched controls rated men's body-odors on four traits: Intensity; Pleasantness; Sexual attraction; and Fertility. Each woman blindly (but see later note) rated stimuli of three kinds: Blank (carrier alone), a non-spouse man, and their actual Spouse. To account for the non-normal distribution of the data, a linear mixed-model with fixed effects of Group, Odor and Descriptor, and random effects of participant, revealed a main effect of Descriptor (F(3,374) = 9.67, p=3.69e−6), and significant interactions of Odor x Descriptor (F(6,374) = 5.69, p=1.13e−5) and Group x Odor (F(2,374) = 7.77, p=4.95e−4) (parametric comparison: a multivariate repeated measures ANOVA with factors of Group (uRPL/Control), Odor (Blank/Non-Spouse/Spouse) and Descriptor (Intensity/Pleasantness/Sexual attraction/Fertility) revealed a main effect of Descriptor (F(3,102) = 12.4, p=5.6e−7), a significant interaction of Odor x Descriptor (F(6,204) = 8.27, p=5.1e−8) and a significant interaction of Odor x Group (F(2,68) = 3.43, p=0.038) (Figure 3a)). The main effect of Descriptor is uninteresting in that it merely reflected that descriptors were applied differently (mean Intensity: 0.49 ± 0.21, mean Pleasantness: 0.42 ± 0.14, mean Sexual attraction: 0.33 ± 0.15, mean Fertility: 0.38 ± 0.15; Wilcoxon pairwise comparisons except intensity-pleasantness: all W > 475, all p<0.026, intensity-pleasantness Wilcoxon W = 431, p=0.127 (parametric comparison, all t(35) > 2.12, all p<0.041, all Cohen’s d > 0.35) (Figure 3—figure supplement 1). The Odor x Descriptor interaction was carried by intensity ratings alone (F(2,68) = 16.2, p=1.72e−6; all other descriptors: all F(2,68) < 2.05, all p>0.13). This intensity difference was carried solely by differences from Blank, which was, unsurprisingly, less intense than all other otherwise (and importantly) equally-intense stimuli (mean Intensity ratings: Blank:=0.31 ± 0.26, Non-Spouse = 0.58 ± 0.23, Spouse:=0.59 ± 0.35. Blank vs. Non-Spouse and Spouse: both Wilcoxon W < 98, p<0.0001, RBC >0.7 (parametric comparison: Both t(35) > 4.7, both p<3.8e−5, both Cohen’s d > 0.77); Non-Spouse vs Spouse: Wilcoxon W < 319, p=0.83, RBC = 0.04 (parametric comparison: t(35) = 0.15, p=0.88, Cohen’s d = 0.03) (Figure 3—figure supplement 1). As to the Odor x Group interaction, whereas uRPL and Control women rated Blank and Spouse the same (uRPL Blank:=0.36 ± 0.2; Control Blank:=0.39 ± 0.18, Mann-Whitney U = 173, p=0.74, RBC = 0.07 (parametric comparison: t(34) = 0.49, p=0.63, Cohen’s d = 0.16); uRPL Spouse = 0.46 ± 0.23; Control Spouse = 0.37 ± 0.21, Mann- Whitney U = 121, p = 0.2, RBC = 0.25 (parametric comparison: t(34) = 1.25, p=0.22, Cohen’s d = 0.42), Non-Spouse was rated lower by uRPL than by Controls (uRPL = 0.36 ± 0.18; Controls = 0.49 ± 0.14, Mann-Whitney U = 244, p=0.009, RBC = 0.51 (parametric comparison: t(34) = 2.35, p=0.025, Cohen’s d = 0.78) (Figure 3a). This effect that materialized in the combined descriptors was mostly carried by a difference is rated fertility attributed to the body-odors (Figure 3—figure supplement 1). In the spouse condition, uRPL and control women smelled different stimuli (each woman smelled her own spouse), so this was not an optimal condition for estimating differences in olfactory perception across these groups (we acknowledge that this condition adds the tantalizing possibility that in addition to altered olfaction in women with uRPL, men in uRPL relationships may have a unique body-odor, yet we do not have statistical power for this separate question within the current study). However, in the non-spouse man condition, both groups are smelling common stimuli, and we observed a significant difference in perception. Again, to verify that this result with non-spouse odors was not a chance event, we conducted a 10,000-repetion bootstrap like before, and observe that the probability of obtaining this result by chance is less than 6 in 1000 (Figure 3b). In other words, we observed a difference in the overall explicit perceptual ratings applied to independent men's body-odor by women with uRPL versus controls.

Women with uRPL have altered perception of men's body-odor.
(a) uRPL (purple) and control (green) women ratings of non-spouse men body-odor, score combines ratings of pleasantness, sexual attraction, intensity and fertility attributed to the odor (see separate ratings in Figure 3—figure supplement 1). n = 36. Bar graphs depict means, each dot represents a participant, dots are jittered to prevent overlay, error bars are s.e.m. *p<0.05. (b) Bootstrap test of the non-spouse result, 10,000 repetitions, the purple line represents the actual uRPL-control difference value.
-
Figure 3—source data 1
All perceptual ratings of blank, non-spouse, and spouse.
- https://cdn.elifesciences.org/articles/55305/elife-55305-fig3-data1-v1.xlsx
Women with uRPL have smaller olfactory bulbs and shallower olfactory sulci
Results to this point imply an altered sense of smell in women with uRPL, specifically in relation to men's body-odor. To probe for brain correlates of this difference, we used magnetic resonance imaging (MRI) to scan 23 women with uRPL and 23 controls. To compare brain structure, we applied both a targeted and a hypothesis-free approach. In the targeted investigation, we compared brain volume in the two primary regions of interest where brain structure has previously been related to olfactory function, namely the olfactory bulbs (Buschhüter et al., 2008) and the olfactory sulci (Rombaux et al., 2010). We observed that women with uRPL have significantly smaller olfactory bulbs (all SW >0.94, all p>0.2; Right bulb volume: uRPL: 45.9 ± 12 mm3, control: 55.4 ± 9.7 mm3, t(44) = 2.96, p=0.005, Cohen’s d = 0.87; Left bulb volume: uRPL: 46.4 ± 8.9 mm3, control: 55.7 ± 9.4 mm3, t(44) = 3.45, p=0.001, Cohen’s d = 1.02; Average bulb volume: uRPL: 46.1 ± 9.9 mm3, control: 55.5 ± 9 mm3, t(44) = 3.37, p=0.0016, Cohen’s d = 0.99) (Figure 4a). To verify that this was not a chance event, we again conducted a 10,000-repetion bootstrap like before, and observe that the probability of obtaining this result by chance is less than 1 in 1000 (Figure 4b). Women with uRPL also had significantly shallower olfactory sulci (Right olfactory sulcus depth did not reach significance: uRPL: 6.4 ± 1.1 mm, control: 7.1 ± 1.4 mm, t(44) = 1.68, p=0.1, Cohen’s d = 0.5; Left olfactory sulcus depth: uRPL: 6.1 ± 1.4 mm, control: 7 ± 1.4 mm, t(44) = 2.22, p=0.03, Cohen’s d = 0.66; Average olfactory sulcus depth: uRPL: 6.3 ± 1.1 mm, control: 7 ± 1.3 mm, t(44) = 2.12, p=0.04, Cohen’s d = 0.62) (Figure 4c). To verify that this was not a chance event, we again conducted a 10,000-repetion bootstrap like before, and observe that the probability of obtaining this result by chance is less than 2 in 100 (Figure 4d). To verify that these anatomical differences were not merely a reflection of smaller heads/brains, we conducted the same analysis on Total Intracranial Volume (TIV), and observe no difference between uRPL women and controls (TIV: uRPL mean: 1366 ± 134 cm3, Control mean: 1390 ± 93 cm3, uRPL SW = 0.95, p=0.31, control SW = 0.94, p=0.2, t(43) = 0.69, p=0.5, Cohen’s d = 0.21). As olfactory bulb volume has been positively associated with olfactory performance (Buschhüter et al., 2008; Rombaux et al., 2010), we tested for such correlations in these data. The one olfactory test that was completed by all 46 scanned participants was the odorant identification test, which revealed no correlation with bulb volume (uRPL: n = 23, Spearman Rho = 0.25, p=0.25. Control: n = 23, Spearman Rho = 0.146, p=0.507. Combined: n = 46, Spearman Rho = 0.068, p=0.652). To further probe this issue, we assigned an olfaction composite score to each participant, reflecting all the olfactory tasks she completed (see Methods). Using this more comprehensive olfactory score, we observed a significant correlation in the uRPL cohort alone (uRPL: n = 22, Spearman Rho = 0.45, p=0.037; Control: n = 23, Spearman Rho = −0.055, p=0.8; Combined: n = 45, Spearman Rho = 0.065, p=0.67) (Figure 4e).

Women with uRPL have smaller olfactory bulbs and shallower olfactory sulci.
(a) Olfactory bulb volume. A 3D reconstruction of uRPL (purple) and control (green) participants’ left (upper row of the two) and right (bottom row) olfactory bulbs. Bulbs sorted by size (see Figure 4—figure supplement 1 for sort by participant (uRPL to Control) match). Note that the reconstructions do not relate to the values on the Y axis, the values are reflected in the data lines alone. (b) Bootstrap test of ‘a’, 10,000 repetitions, the purple line represents the actual uRPL-control difference value. (c) Olfactory sulci depth (right, left and average) of uRPL (purple bars) and control (green bars) participants. (d) Bootstrap test of the average in ‘c’, 10,000 repetitions, the purple line represents the actual uRPL-control difference value. Bar graphs depict means, each dot represents a participant, dots are jittered to prevent overlay. Error bars are s.e.m. *p<0.05. (e) The relation between olfactory composite scores and olfactory bulb volumes for uRPL (purple) and control (green) participants. n = 46. Each square (uRPL) and triangle (control) represent a participant.
-
Figure 4—source data 1
All olfactory bulb volumes and sulci depth.
- https://cdn.elifesciences.org/articles/55305/elife-55305-fig4-data1-v1.xlsx
Next, in a hypothesis-free analysis, we conducted whole-brain voxel-based morphometry (VBM). This analysis, with correction for the multiple comparisons associated with the entire brain, uncovered no significant group differences. We note that if we repeat this analysis at strict threshold (p<0.0001) but without correction for multiple comparisons, it implied reduced gray matter in the right fusiform in uRPL (Figure 5—figure supplement 1). Hence, structural brain imaging uncovered a significant difference in the olfactory structures (olfactory bulb and olfactory sulci) of women who experience uRPL, and hinted at a possible difference in non-primary olfactory structures (right fusiform), yet that have been implicated in social chemosignaling (Zhou and Chen, 2008). We next turn to functional imaging.
Women with uRPL have an altered brain response to men's body-odor
We used functional magnetic resonance imaging (fMRI) in 23 women with uRPL and 23 controls who watched varying-arousal movie-clips during subliminal exposure to body-odor from a Non-Spouse men. We chose not to use spouse body-odors in this particular experiment because this may have presented a confound: As noted, women with uRPL can recognize their spouse's body-odor. This pronounced difference between cohorts was overwhelmingly apparent in the previous rating experiment, where we can mention anecdotally, that although blind to condition, upon presentation with spouse body-odor, 15 of 33 women from the uRPL cohort spontaneously said ‘oh, that is my spouse’.
This never happened once in the control cohort. Thus, if we used spouse odors in the MRI, we would risk the confound of only uRPL women knowing that any odors, let alone their spouse body-odors, were presented. Moreover, given that uRPL relationships may involve significant emotional strain (Kolte et al., 2015), this potential confound would be exacerbated. Thus, we chose to use male body-odor collected from donors unrelated to all study participants, and use the odor at subliminal levels that were not spontaneously detected. We further decided to test these body-odors as modulators of an emotional response rather than an activating stimulus alone because the impact of human social odors is typically more evident in this manner (Jacob et al., 2001).
The functional study was made of two consecutive scans, one with olfactometer-delivered undetected non-spouse men's body-odor, and one with olfactometer-delivered blank (counterbalanced for order across participants). In each such scan, each participant observed 40 12s-long movie clips, and after each, participants self-rated the level of emotional arousal the clip induced (using an 8-level response box). Like in the structural analysis, in the functional analysis we also applied both a targeted and hypothesis-free approach. In the targeted approach, we first concentrated on the hypothalamus, the primary brain structure implicated in linking the brain olfactory and reproductive systems (Rosser et al., 1989; Yoon et al., 2005; Brennan, 2009). The hypothalamus has also been indicated in women's responses to a molecule (AND) that may occur in men's body-odor (Savic et al., 2001). We used a hypothalamic region of interest (ROI) independently defined in Neurosynth (Yarkoni et al., 2011), and applied a contrast of Group (uRPL/Control) and Stimulus (Undetected body-odor/Blank) with parametric modulation for the self-rated level of emotional arousal. Following small-volume correction for multiple comparisons, we observed a remarkable effect in the hypothalamus (Z > 2.3, Cluster-corrected) (Figure 5a). Although the p-value associated with this result is the statistical parametric map statistic (p<0.01, corrected), we used a repeated-measures ANOVA applied to the parameter estimates only to understand the directional drivers of this effect. We observed that whereas body-odor increased hypothalamic activation in uRPL (t(22) = 2.81, p=0.01), it decreased hypothalamic activation in controls (t(22) = 2.41, p=0.025), and the interaction was powerful (F(1,44) = 13.4, p=0.0007) (Figure 5a). In turn, a hypothesis-free approach full-brain contrast ANOVA with correction for the multiple comparisons involved, did not uncover additional brain regions differently impacted by arousal as a function of exposure to body-odor in uRPL vs. controls.

Women with uRPL have an altered brain response to male body-odor.
(a) Hypothalamus blood-oxygen level-dependent (BOLD) activity in uRPL (non-spouse >blank), compared to control. (b) Whole brain PPI test, reflecting greater correlation with hypothalamus (seed ROI) time series for all emotionally-weighted-movie-clips>fixation. Both scatterplots reflect the % signal change of each participant (uRPL in purple, controls in green). The diagonal is the unit slope-line (x = y). Dots above the slop line represent higher % signal change for Non-Spouse men body odor, and dots below the line represent higher % signal change for Blank. n = 46. All coordinates in MNI space.
-
Figure 5—source data 1
Parameter estimates for body-odor and blank.
- https://cdn.elifesciences.org/articles/55305/elife-55305-fig5-data1-v1.xlsx
Given the hypothalamic group-difference, we next used psychophysiological interaction (PPI) analysis (Friston et al., 1997) to ask whether functional connectivity with this region differed across groups as a function of exposure to body-odor. We observed that with increased arousal, body-odor increased connectivity between the hypothalamus and right insula significantly more in controls than in uRPL (peak Montreal Neurological Institute coordinates (MNI): [33,21,-8], 2336 voxels, Cluster-corrected z > 3.1) (Figure 5b) We again note that the p-value associated with this finding is the above PPI mapping statistic (p<0.001, corrected), but to only verify the directionality of the effect we applied a repeated-measures ANOVA on the parameter estimates, revealing that the decrease was significant in uRPL (t(22) = 2.45, p=0.022), the increase was significant in Controls (t(22) = 4.41, p=0.0002), and the interaction was powerful (F(1,44) = 20.6, p=0.00004) (Figure 5b). Patterns of brain activity can be impacted by anxiety (Holzschneider and Mulert, 2011), which can be increased in uRPL (Mevorach-Zussman et al., 2012). To address this possible source of variance in our data, all participants completed the trait-anxiety questionnaire (State-Trait Anxiety Inventory; STAI Teichman and Melnick, 1979), and a personality questionnaire (The ‘Big Five’ Inventory Etzion and Laski, 1998). We observed that both uRPL and controls had moderate anxiety levels, and there was no significant difference between the groups (STAI: uRPL = 39.5 ± 8, Control = 39.87 ± 10.86, SW = 0.97, p=0.3, t(44) = 0.14, p=0.89, Cohen’s d = 0.041).
Moreover, we observed no significant differences in personality (Extroversion (E), Conscientiousness (C), Neuroticism (N) Agreeableness (A), Openness to experience (O)); uRPL: E = 3.67 ± 0.59, C = 3.71 ± 0.49, N = 3.05 ± 0.67, A = 3.87 ± 0.46, O = 3.58 ± 0.67. Control: E = 3.51 ± 0.64, C = 3.94 ± 0.41, N = 2.97 ± 0.65, A = 3.73 ± 0.64, O = 3.73 ± 0.44 (All SW >0.957, all p>0.092, all t(44) < 1.68, all p>0.1, all Cohen’s d < 0.5). Therefore, the unique pattern of brain activity in response to body-odor in uRPL was not a reflection of differences in anxiety or personality across cohorts.
Finally, given these perceptual and brain differences, one may postulate a generalized different impact of body-odor in women with uRPL. To investigate this, we replicated an experiment where we previously observed an impact of body-odors on behavior and psychophysiology (Endevelt-Shapira et al., 2018). We tested 18 uRPL women and 18 controls in a single experimental session that included sniffing body-odors, followed by a face recognition task, an empathy task, and an emotional Stroop task, concurrent with Galvanic Skin Response (GSR) recording, once under exposure to Non-Spouse-men body-odor and again under exposure to Blank. Despite some potentially meaningful group and odorant differences, we failed to observe any significant interaction of group and odorant in this composite set of experiments. In other words, body-odors had the same impact on these measures in both uRPL and control women. Although one cannot read too much into such a null effect, it is important to report it when conveying our results, as it confines the extent of difference between these cohorts in their responses to body-odors.
Discussion
We found that, as a group, women who experience uRPL have an overwhelming advantage over controls at recognizing their spouse by smell, yet only a slight general olfactory advantage. Whereas control performance at olfactory spouse identification was on par with two previous reports (Hold and Schleidt, 1977; Mahmut et al., 2019), it was poorer than in a third previous study, where participants performed on par with the current uRPL group (Lundström and Jones-Gotman, 2009). We speculate that these differences across studies reflect methodological differences impacting task difficulty. More specifically, the latter study reapplied body-odor collection-pads to donors for seven straight nights. This likely made for a very concentrated body-odor. In contrast, our donors wore t-shirts for two nights, making for a less concentrated stimulus, and hence lower overall performance. The question here, however, was not of a comparison across studies, but rather across groups in the same (our) study, where we obviously applied common methods for both groups.
We note two considerations related to this primary behavioral finding. First, considering the only marginal uRPL advantage at other olfactory tasks, this advantage may reflect some downstream uRPL advantage, perhaps even abilities at memory beyond the olfactory system alone. Second, as mentioned in the results, this advantage may be impacted by a unique body-odor in uRPL men. Alas, our experimental design did not permit addressing this question in the current results, and this remains an open and tantalizing question for future research. That said, if there is a body-odor difference in uRPL men, we speculate that this is in addition and not instead of altered body-odor perception in uRPL women. We say this because in addition to better performance in spouse identification, uRPL women also reported different perceptual qualities associated with a group of independent men: Women with and without uRPL rated the same group of non-spouse men, and most women with uRPL perceived them differently (Figure 3). Thus, we stand by the notion that, at the group level, women with uRPL have altered perception of men's body-odor (Figures 1c and 3b).
uRPL was also associated with significantly increased body-odor-induced activation in the hypothalamus, followed by altered patterns of functional connectivity, primarily with the insula. The hypothalamus may be thought of as the primary brain structure associated with reproductive behavior and state (Schally et al., 1972; Maffucci and Gore, 2009), and its link with the insula may be tied to processing human body-odors related to emotional state (Prehn-Kristensen et al., 2009), and in body-odor-based human kin recognition (Lundström et al., 2009). This combines with evidence of hypothalamic activation by androgen-like odorants in women but not men (Savic et al., 2001) to imply that body-odors may activate hormonal cascades in humans similar to those at the heart of several social chemosignaling effects characterized in rodents. This is consistent with the growing evidence for social chemosignaling in general human behavior (McClintock et al., 2001; Prehn et al., 2006; Zhou and Chen, 2009; Mitro et al., 2012; Semin and Groot, 2013; Olsson et al., 2014; Lübke and Pause, 2015; de Groot et al., 2017).
Although the functional results dovetail with the behavioral results to depict heightened processing of body-odors in uRPL, the structural results are counter-intuitive. uRPL was associated with smaller olfactory bulb volume. On one side, this result provides yet additional and particularly convincing evidence for an altered olfactory brain profile in uRPL. We label this as particularly convincing because volumetric differences are less susceptible to analysis strategies than are functional differences. On the other hand, whereas smaller OBs are typically associated with poorer olfaction (Buschhüter et al., 2008; Rombaux et al., 2010), here we observed smaller OBs in the group with better olfactory spouse identification (uRPL), and slightly better olfaction overall. A correlation between OB volume and general olfactory performance materialized only within the uRPL group, but not in the control group. We note that the relation between OB volume and olfactory performance is not always straight forward (Weiss et al., 2019), and we indeed have no explanation for this relationship in the current data.
We would like to also highlight several limitations of this study. First, the study with all its experiments was conducted over a very long period of time: nearly 7 years. This may have one advantage to it in that the results replicated at the hands of different experimenters over time, but the disadvantage is that methodological nuances were more varied across participants than one would want. This was an unnecessary source of variance. Second, some of the behavioral experiments were not double-blind in the strictest sense of the term. More specifically, although participants received instructions and entered replies through computer, the experimenter that placed the stimuli before them knew stimulus identity. A third limitation is the lack of correlation within participants across olfactory tasks. One typically expects participants who are good at one olfactory task to be good at other olfactory tasks, yet here these correlations were weak, a weakness we have no explanation for. Finally, the fMRI results were evident in targeted ROI-based analyses, but did not emerge from hypothesis-free strict parametric mapping. Although this is not unusual, especially with relation to hypothalamic results (Savic et al., 2005; Burke et al., 2012; Burke et al., 2014), it is less powerful than one would want. Despite these limitations, we think the overall image of an altered olfactory brain, and altered olfactory function in uRPL, particularly in relation to body-odor, holds strong.
Finally, it is tempting to combine the current picture of altered perception and brain responses to men's body odor in uRPL, with additional findings such as a specific polymorphism in an olfactory receptor gene (OR4C16G > A) that has been associated with uRPL (Ryu et al., 2019), and the tradition of protective isolation of women during early pregnancy evident in some tribal societies (Van Gennep, 1960), to jointly hint at a form of Bruce effect in humans. In the Bruce effect, an exposure to the odor of a non-sire male frequently leads to implantation failure in females (Bruce, 1959). Such an analogy, however, is restricted by major limitations. First, correlation is not causation, and altered behavioral and brain responses to body-odors in uRPL, convincing however they may be, do not alone imply that this is causal in the condition. Behavior may differ because of increased motivation in participants with uRPL, and brain structure and function may differ because of an independent covarying factor that we failed to identify. Although we could think of several experiments that would investigate causation, these are prevented by ethical considerations. Moreover, any notion of a human Bruce-like effect is further challenged by the underlying neuroanatomy. In rodents, the Bruce effect relies on two neural structures that humans may not possess: The first is the VNO, namely the dedicated sensory epithelia in the rodent nasal passage that is primarily involved in social chemosignaling (Keverne, 1999). Humans may retain a VNO pit on the nasal septum (Trotier et al., 2000), but this structure is considered vestigial in the human nose (Meredith, 2001). Second, the rodent VNO projects to the rodent accessory olfactory bulb (AOB), a structure also reported missing in the human brain (Meisami et al., 1998). It is in the AOB where the odor of the fathering male may be imprinted during mating in rodents (Brennan, 2009). Although humans may lack a VNO and AOB, humans do functionally express vomeronasal receptors in their olfactory epithelium (Rodriguez et al., 2000). Although rodent social chemosignaling is primarily attributed to the VNO and AOB, the main epithelium and bulb can also support social chemosignaling (Spehr et al., 2006). Therefore, the missing neural substrates do not altogether preclude a human Bruce-like effect, but do restrict such an analogy. Finally, whereas in the Bruce effect the pregnancy is terminated at the stage of implantation, in humans this would be almost impossible to identify, and women who are diagnosed with uRPL are past the implantation phase (Zipple et al., 2019). With these limitations in mind, we restrict our claims to the observation of altered perceptual and brain responses to men's body-odor in uRPL. This potentially provides for a much-needed reorientation of thinking in uRPL. Rather than looking at the uterus and the hormonal environment alone, we looked at the brain, and more specifically at the olfactory system, and identify unique patterns associated with uRPL. Indeed, as far as we know, structural and functional patterns of brain organization have not been previously associated with uRPL, and this discovery was guided by our investigation of olfaction. Thus, this framework and set of findings may combine with additional results on the potential impact of social environment on pregnancy loss (Catalano et al., 2016; Catalano et al., 2018) to jointly redirect thinking on this condition, which is currently not well-understood, or managed. With all that said, we will end in a sort of disclaimer, which is important in light of the subject matter: The relationships we identified were group level relationships. As a group, women with uRPL differed from controls, but at every given measure, there were women from the uRPL group with control-like results, and women from the control group with uRPL-like results. Given this variability and overlap, this manuscript does not contain results that merit any behavioral adjustments by women trying to maintain a pregnancy.
Materials and methods
Participants
Conducting studies in uRPL is complicated by the issue of long-term continuous participant availability. This manuscript contains many different experiments that ideally would have all been conducted in the same cohort. This would have allowed optimal relating of one result to another. However, ethical considerations prevent us from conducting any experiments with women who are currently pregnant, or trying to conceive. Because this cohort is often doing exactly that, it is almost impossible to conduct experiments in a given participant over an extended period of time. Instead, participants were available sporadically. To deal with this, we recruited an overall cohort of 40 women with uRPL (mean age: 35.06 ± 6.22), and 57 controls (mean age: 34.66 ± 4.24), and then assigned matched pairs based on uRPL availability per experiment (Supplementary file 1). We set the sample sizes based on an initial power analysis based on a previously published study (Lundström and Jones-Gotman, 2009). This implied that at alpha = 0.05% and 80% power we need to test a sample of 29 participants per group. All participants provided written informed consent to procedures approved by the Tel- Hashomer (behavioral studies, protocol reference number 8067–10-SMC) or Wolfson hospital (imaging studies, protocol reference number 0121–10-WOMC) Helsinki Committees. All participants were screened for good general health and no history of psychiatric disease.
Inclusion/exclusion criteria
Request a detailed protocolInclusion criteria for uRPL was at least two consecutive miscarriages prior to 15 weeks, with an unknown etiology for the pregnancy loss, despite careful medical investigation. Inclusion criteria for controls was no history of (known) pregnancy loss or abortions.
Control matching
Request a detailed protocolBecause this study involves natural groups rather than randomization, we paid close attention to controlling for factors which may confound our findings. For each comparison performed, we matched control participants to uRPL participants based on their age and number of children (see Figure 1—source data 1, Figure 2—source data 1 and Figure 4—source data 1 for details). If multiple choices were available, our choice was based on matching the average of the control group as a whole. For the olfactory bulb analysis, following recent findings regarding the impact of pregnancy on maternal brain anatomy after birth (Kim et al., 2010; Hoekzema et al., 2017; Kim et al., 2018), we additionally matched for age of offspring, particularly age of the youngest child (see Figure 4—source data 1). Additional factors, like marital status and living arrangements, were similar between groups: the overwhelming majority of women in our study were married and living with their spouses. One factor that is by definition confounded with group differences is number of pregnancies, as RPL is defined as two or more consecutive miscarriages. To ask whether number of pregnancies may be associated with our results, we examined the most prominent behavioral result, namely spouse identification, and its association with number of pregnancies within each group. We observed no significant correlations (uRPL group: Spearman’s Rho = 0.19, p=0.3; Control group: Spearman’s Rho = −0.29, p=0.11), implying that number of pregnancies alone does not impact this result.
Behavioral studies
Body-odor collection
Request a detailed protocolEach spouse was provided with a new 100%-cotton white T-shirt for body-odor collection according to standard procedures (Endevelt-Shapira et al., 2018). Briefly, the donors were instructed to wear the shirt for two consecutive nights. The donors were further instructed to avoid eating ingredients that can alter body-odor (fenugreek, asparagus, curry, etc.) prior to body-odor sampling. In addition, during the sampling days, donors were asked not to use soap, shampoo, conditioner or deodorant. Between the two nights, the T-shirts were kept inside a closed glass jar at room temperature, and after the second night, they were further stored at −20 °C to prevent bacterial growth.
Shirt sniffing device (SSD)
Request a detailed protocolOn the morning of the experiment shirts were thawed at room temperature inside the jars to avoid humidity condensation. Using sterile scissors, shirts were cut into two longitudinal pieces, such that each half of the shirt contained one axillary area. Each half shirt was then placed inside an SSD – a glass jar covered by a cap with an air filter, inhalation mask and a one-way flap-valve (Figure 1a). As a blank control, new unworn shirts were also frozen inside a glass jar, thawed, and cut to two halves prior to the experiment.
SSD experimental design
Request a detailed protocolAt the onset of every trial, the experimenter arranged the three SSDs in front of the participant, and then moved away. One SSD contained a shirt that originated from the participant's spouse, one SSD contained a shirt that originated from another participant’s spouse (non-spouse), and the third contained a Blank. The participant then received on-screen instructions to sniff each jar only once, before entering her selection of ‘spouse’ using a computer track-pad. Participants did not receive feedback as to whether they were right or wrong. Each triangle combination was presented either four or six times (Supplementary file 2), and percent accuracy was calculated. Although this experimental arrangement is not genuinely double-blind (as the experimenter placing the jars knew their content), the lack of verbal interaction between the experimenter and participant during trials, the participant self-initiated sampling action and timing, and the use of different experimenters across participants, together minimize the risk of experimenter-generated cues.
Odorant identification
Request a detailed protocolWe assessed ordinary odorant identification using a subset of an established and validated standardized test (University of Pennsylvania Smell Identification Test, UPSIT Doty et al., 1984). Although the full UPSIT contains 40 stimuli, the results with a subset of 20 stimuli are highly correlated to the results with 40 (Doty et al., 1989). To also verify this in our type of cohort, a subset of 26 uRPL participants completed the full 40 stimuli version. We observed a strong fit between the score with 20 and 40 odorants (Spearman Rho = 0.71, p<0.001). With this in hand, we proceeded to use the 20 stimuli test. Out of 97 participants (40 uRPL), three were excluded (2 uRPL) (>2.7 SD), retaining 94 participants (38 uRPL).
Monomolecules
Request a detailed protocolEach trial contained sequential presentation of three jars (counterbalanced for order), two containing the carrier alone (Propylene glycol) and one containing the carrier with the monomolecule: either androstadienone (AND, androsta-4,16,-dien-3-one, at a final concentration of 0.000025M), androstenone (ANN; 16, (5α)-androsten-3-o, at a final concentration of 0.000025M) or estratetraenol (EST; 1, 3, 5 (10), 16-estratetraen-3-ol, at a final concentration of 0.045M). Participants were allowed to take one 2 s long sniff at each odorant presentation and were then asked to pick out the jar that contained the dissimilar odorant. A subset of participants (same number of uRPL and Controls) completed three repetitions per triangle, and the remainder completed 10 repetitions per triangle. The order of presentation of each triplet was randomized, and both the experimenter and participant were blind for condition.
DMTS threshold test
Request a detailed protocolOlfactory detection thresholds for the alliaceous odor dimethyl trisulfide (DMTS) were determined, using a maximum-likelihood adaptive staircase procedure (MLPEST) (Linschoten et al., 2001), in which a participant was presented with two jars, one blank (containing carrier alone; Propylene Glycol) and one containing a changing concentration of DMTS, and had to choose which contained the odorant (forced choice). Out of 39 participants (19 uRPL), two were excluded (1 uRPL, all >2.7 SD), retaining 37 participants(18 uRPL).
Olfactory composite score
Request a detailed protocolThe composite score of identification, detection and threshold was calculated using the averaged monomolecule discrimination score, odor identification scores and threshold scores. The two latter scores were normalized to a scale of 0–1 (using the formula (x-min)/(max-min)), and all three scores were then averaged per participant to generate the comprehensive composite score. Out of 97 participants (40 uRPL), 1 uRPL was excluded (>2.7 SD) retaining 96 participants (39 uRPL). Next, we matched controls to uRPL women as described in the ‘control matching’ section, first by matching according to the type of tests these women have a score for, and if multiple options were available, by age and number of children. Two uRPL women did not have matched controls by test and age, so we matched controls by the closest type of tests performed and age, and removed the non-matching test from the composite average score of this participant.
Body-odor ratings
Request a detailed protocolThe three body-odors, Spouse, Non-Spouse and Blank, were presented using SSDs to participants in randomized order. Participants were instructed to sniff for 2 s and then rate odor on a visual analog scale (VAS) for perceived intensity, pleasantness, sexual attraction and fertility associated with the body-odor. Inter-Stimulus Interval was 30 s.
Behavioral statistics
Request a detailed protocolAll data analyses were performed using JASP (JASP Team (2019) Version 0.11.1). Each data-set was first tested for normality of distribution using Shapiro-Wilk test of normality. For normally distributed data, we used the following parametric measures: If only one variable was compared, we used two-tailed t-tests. To test for between-group differences across multiple related variables, we used multivariate analysis of variance (ANOVA) followed by post-hoc two-tailed t-tests. To test for correlations between variables, we used the Pearson correlation coefficient for continuous variables, or Spearman’s Rank Correlation Coefficient if one or both variables were categorical. For abnormally distributed data we used the following nonparametric measures: If only one variable was compared, we used Wilcoxon signed-rank test for one-sample and paired samples, and Mann-Whitney test for independent samples. To test for between-group differences across multiple related variables, we used linear mixed models (using R 3.6.1, packages readxl version 1.3.1, nlme version 3.1–140, car version 3.0–5) followed by the above-mentioned nonparametric tests, and tested for correlations using Spearman’s Rank Correlation Coefficient. Finally, to estimate power, we calculated Cohen's d for parametric measures and rank biserial correlation (RBC) for nonparametric measures. We note that in all cases where we relied on non-parametric tests, we also reported the parametric outcome in parenthesis for comparison. We observe that the non-parametric and parametric approaches yielded nearly identical outcomes in terms of assigning significance to the reported differences. Finally, we used bootstrap verification in order to verify the statistical validity of the more prominent results. In each case, we randomly shuffled the assignment of the data to uRPL or Control 10,000 times, and reconducted the analysis.
Brain imaging
Participants
A total of 48 participants (24 uRPL) were scanned in the fMRI study, and 55 participants (26 uRPL) in structural T2-weighted imaging for OB volume and OS depth analysis.
Data acquisition
Request a detailed protocolMRI Scanning was performed on a 3-Tesla MAGNETOM Trio Siemens scanner. For olfactory bulb (OB) volume and olfactory sulcus (OS) depth, a 32-channel head coil was used. Images were acquired using a T2-weighted turbo spin echo pulse sequence in the coronal plane, covering the anterior and middle segments. Sequence parameters: 35 slices, voxel size: 400 × 400 µm, slice thickness: 1.6 mm, no gap, TE = 85 ms, TR = 7000 ms, flip angle = 120°, two averages. Functional data were collected using a 12-channel head matrix coil and T2*-weighted gradient-echo planar imaging sequence, with the following parameters: 450 repetitions, TR = 2000 ms, TE = 25 ms, flip angle = 75°, FOV = 216 × 216 mm2, matrix = 72 × 72 mm2, voxel size = 3 × 3 mm, slice thickness = 3.7 mm, no gap, 34 transverse slices tilted to the AC-PC plane. Anatomical images for functional overlay were acquired using a 3D T1- weighted magnetization prepared rapid gradient echo (MP-RAGE) sequence at high resolution: 1 × 1×1 mm3 voxel size, TR = 2300 ms, TE = 2.98 ms, inversion time = 900 ms, and a flip angle = 9°.
All imaging data are publicly available at https://openneuro.org/datasets/ds002717.
Structural brain imaging
Request a detailed protocolOlfactory bulb and olfactory sulcus analysis: We excluded 3 of the 26 uRPL participants due to poor resolution (likely motion related) which did not allow a reliable measure of the sulcus or bulb. Next, 23 matching controls were selected based on parity and age of their children, specifically the age of their youngest child, following recent findings regarding the impact of pregnancy on maternal brain anatomy after birth (Kim et al., 2010; Hoekzema et al., 2017; Kim et al., 2018). We used independent sample t-tests to compare OB volume and OS depth between the two groups. OB volume and OS depth were measured according to standard methods (Rombaux et al., 2010). For delineating and measuring the OB we developed software that automatically identifies the region of the OBs, and then allows the user to manually delineate OB boundaries across interpolated images for automated volume estimation. This software was written by co-author LG, has been used before (Weiss et al., 2019), and is available for download at https://gitlab.com/liorg/OlfactoryBulbDelineation. To calculate estimated Total Intracranial Volume (eTIV) we used FreeSurfer version 6.0.0 (Buckner et al., 2004), next we normalize the OB by eTIV for 22 participants (MPRAGE for one uRPL participants was not obtained, so her matched control was also excluded).
OS depth was measured using ITK-SNAP version 3 (www.itksnap.org). The coronal slice was picked by the Plane of the Posterior Tangent through the eye-balls (PPTE), which in most individuals traverses the anterior-mid segment of the OB. In this slice, a virtual line that was tangent to the inferior border of the orbital and rectus gyri was drawn, and then perpendicular line connecting the above virtual line and the deepest part of the OS was marked. This line represents OS depth (Rombaux et al., 2010). Two independent raters, blind to participant identity and group-allocation, demarcated OB borders and marked OS depth. Next, all the OB volumes and OS depths in which between-rater difference was above 15%, were judged by a third blind review. Subsequent inter-rater correlation in OB volume was r = 0.92, p<0.001 and in OS depth r = 0.93,<0.001. Inter-rater agreement was further validated using intraclass correlation coefficient Tinsley and Weiss, 1975: OB volume: ICC r = 0.96, confidence intervals (CI) 0.925–0.981, OS depth: ICC r = 0.967, CI 0.939–0.985.
For voxel-based morphometry structural data from 23 uRPL and 23 matched-control was processed with FSL 6.0 software. First, the brain extracting tool (BET) was employed to cut the skull from all T1 MRI structural images Smith, 2002; Next, FSL Automated Segmentation Tool (FAST) was adopted to carry out tissue-type segmentation (Zhang et al., 2001). After careful check for segmentation quality, the segmented GM parietal volume images were aligned to the MNI standard space (MNI152) by applying linear image registration tool (FLIRT) and nonlinear registration (FNIRT) methods. A study-specific template was created by averaging the registered images, to which the native grey matter images were then non- linearly re-registered. The registered GM parietal volume images were modulated for the contraction/enlargement due to the nonlinear component of the transformation by dividing them by the Jacobian of the warp field. Next, the segmented and modulated images were smoothed with isotropic Gaussian kernels with a standard deviation of sigma = 3 mm. Finally, a voxelwise GLM was applied using permutation-based non-parametric testing, using p<0.0001, without additional correction for multiple comparisons.
Functional brain imaging
Odor stimuli
Request a detailed protocolTwenty-two heterosexual men (mean age 31.1 ± 6.4) donated their body-odor, following the standard procedures described above, with two differences: we used eight adhesive pads rather than t-shirts, for four nights rather than 2. After each night the pads were stored at -20°C. We used pads rather than t-shirts so as to have smaller and more concentrated odorant sources that would fit into the olfactometer. These pads were cut into small pieces and mixed such that the scanner stimulus contained a mix of 20 men.
Video clips
Request a detailed protocolAn independent group of 12 women rated the emotional arousal, on a scale of 1 to 10, of 30 ~ 1 min-long scenes. Scenes, all containing human characters, were chosen from 11 commercial films. From these we used the 20 scenes with the highest emotional scores. All scenes were edited into 12 s clips.
Procedure
Request a detailed protocolTo assure no differences related to prior exposure to the film-clips, and to introduce the narrative and emotional context of the stimuli, prior to the scan, participants watched the 20 1-min-long scenes and rated familiarity. Next, a one-hour long fMRI scan was conducted, which included two functional scans, separated by a T1 anatomical scan. After the second functional scan, a T2 anatomical scan was conducted. Each functional scan started with a 4 s emotional clip (to be excluded in the analysis) and next 12-sec-long video clips were presented: 20 clips of high emotional content and 20 landscape clips, in alternating order, with an ITI (fixation point) of 8–12 s. Following each video clip, participants were asked to rate their emotional arousal, on a scale of 1 to 8, with eight being the highest. Simultaneously with each video clip, body-odor or blank were delivered using a computer-controlled air-dilution olfactometer that embedded the odorant pulse within a constant stream of clean air at 1.5 L/min. The two functional scans were identical except for the odor content. The two conditions (blank vs. non-spouse-odor) were counterbalanced for order between participants. Participants were aware of the possibility of an odor being delivered but were not told when. After scanning, participants were asked to report whether there was an odor in each of the sessions, score of ‘1’=an odor (even if it was noticed only once) and ‘0’=no odor. First, we calculated the differences between body-odor sessions and blank sessions to receive three values: 0 (probability: 0.5),–1 (probability: 0.25), 1 (probability: 0.25). Next, we used a binomial test to calculate the probability of detection.
Following scanning, participants watched the emotional 12 s clips again, outside the scanner, and rated them for specific emotions: compassion, happiness, fearfulness, sadness, stress and emotional power, and familiarity prior to the experiment. There were no significant differences between the two groups in any of these ratings (all t(44) < 1.39, all p>0.17).
Functional MRI analysis
Request a detailed protocolPreprocessing and functional data analysis were conducted within FSL (FMRIB’s Software Library, www.fmrib.ox.ac.uk/fsl), FEAT (FMRI Expert Analysis Tool) Version 6.00 (Woolrich et al., 2001), and MATLAB R2018a (MathWorks, Inc).
Preprocessing pipeline
Request a detailed protocolTwo participants were excluded from the analysis: one uRPL woman due misunderstanding of the task (she did not rate the emotional arousal of the landscape clips), and one control participant due to miscarriage that was reported only after the scanning. Thus, a total of 46 participants, 23 in each group were submitted to further analysis. The first eight volumes (4 s clip) in each run were discarded to allow the MR signal to reach steady-state equilibrium. The structural brain images were skull-stripped using the FSL brain extraction tool (Smith, 2002). Functional images were corrected for slice-timing and head motion using six rigid-body transformations with FSL MCFLIRT (Jenkinson et al., 2002). Within each run, functional images were spatially normalized to the individual’s anatomy and co-registered to the MNI 152 T1 template, using a combination of affine and non-linear registrations. Images were spatially smoothed with an 8 mm Gaussian kernel, and a high-pass filter (cut off = 128 s) was incorporated into the GLM to correct for scanner drift.
Voxel-wise analyses
Request a detailed protocolOn the first level, all 3 EV regressors were modeled by a stick function convolved with a double gamma function. The regressors included: All video-clips with no weightings (and their temporal derivative), video-clips weighted with emotional arousal ratings (demean), reaction time (orthogonal to the other two regressors). This model also included a regressor of no interest for each volume, with >0.9 mm framewise displacement (using FSL motion outliers). At the second level, the parametric estimation of the weighted-clips was a contrast of non-spouse body-odor >blank using fixed-effects. Finally, for between group analysis we used small-volume correction of the hypothalamus (defined by Neurosynth), and FLAME1 (FMRIB's Local Analysis of Mixed Effects), with cluster-size correction z > 2.3 (p<0.01).
Functional connectivity
Request a detailed protocolWe used psychophysiological interaction (PPI) analysis (Friston et al., 1997) to measure changes in functional connectivity modulated by emotional arousal. We conducted whole-brain PPI tests, reflecting greater correlation with hypothalamic (seed ROI) time series (physiological regressor) for all emotionally-weighted-movie-clips>fixation (psychological regressor). At the second level, the parametric estimation of the weighted-clips was a contrast of body-odor >blank using fixed-effects. Finally, for between group analysis, we used FLAME1, with cluster-size correction z > 3.1 (p<0.001).
Data availability
Source data files have been uploaded for Figures 1-5. All imaging data is publicly available on OpenNeuro at https://openneuro.org/datasets/ds002717.
-
OpenNeuroID ds002717. Unexplained Repeated Pregnancy Loss is Associated with Altered Perceptual and Brain Responses to Men's Body-Odor.
References
-
Outstanding issues surrounding vomeronasal mechanisms of pregnancy block and individual recognition in miceBehavioural Brain Research 200:287–294.https://doi.org/10.1016/j.bbr.2008.10.045
-
A unified approach for morphometric and functional data analysis in young, old, and demented adults using automated atlas-based head size normalization: reliability and validation against manual measurement of total intracranial volumeNeuroImage 23:724–738.https://doi.org/10.1016/j.neuroimage.2004.06.018
-
Twinning in Norway following the Oslo massacre: evidence of a 'Bruce Effect' in HumansTwin Research and Human Genetics 19:485–491.https://doi.org/10.1017/thg.2016.58
-
Separating the Bruce and Trivers-Willard effects in theory and in human dataAmerican Journal of Human Biology 30:e23074.https://doi.org/10.1002/ajhb.23074
-
On the communicative function of body odors: a theoretical integration and reviewPerspectives on Psychological Science 12:306–324.https://doi.org/10.1177/1745691616676599
-
Sex steroids as pheromones in mammals: the exceptional role of estradiolHormones and Behavior 68:103–116.https://doi.org/10.1016/j.yhbeh.2014.08.003
-
Sensory coding of pheromone signals in mammalsCurrent Opinion in Neurobiology 10:511–518.https://doi.org/10.1016/S0959-4388(00)00121-5
-
Altered responses to social chemosignals in autism spectrum disorderNature Neuroscience 21:111–119.https://doi.org/10.1038/s41593-017-0024-x
-
WebsiteHebrew version of the big five index by permissionTel Aviv University, Faculty of Management, the Institute of Business Research, Tel-Aviv. Accessed May 5, 2016.
-
Pregnancy leads to long-lasting changes in human brain structureNature Neuroscience 20:287–296.https://doi.org/10.1038/nn.4458
-
The importance of human odour in Non-verbal communicationZeitschrift Für Tierpsychologie 43:225–238.https://doi.org/10.1111/j.1439-0310.1977.tb00072.x
-
Neuroimaging in anxiety disordersDialogues in Clinical Neuroscience 13:453–461.
-
Effects of breastfeeding chemosignals on the human menstrual cycleHuman Reproduction 19:422–429.https://doi.org/10.1093/humrep/deh057
-
Psychological state and mood effects of steroidal chemosignals in women and menHormones and Behavior 37:57–78.https://doi.org/10.1006/hbeh.1999.1559
-
Cortical thickness variation of the maternal brain in the first 6 months postpartum: associations with parental self-efficacyBrain Structure and Function 223:3267–3277.https://doi.org/10.1007/s00429-018-1688-z
-
Fast and accurate measurement of taste and smell thresholds using a maximum-likelihood adaptive staircase procedurePerception & Psychophysics 63:1330–1347.https://doi.org/10.3758/BF03194546
-
The neuronal substrates of human olfactory based kin recognitionHuman Brain Mapping 30:2571–2580.https://doi.org/10.1002/hbm.20686
-
Maternal status regulates cortical responses to the body odor of newbornsFrontiers in Psychology 4:597.https://doi.org/10.3389/fpsyg.2013.00597
-
Romantic love modulates women's identification of men's body odorsHormones and Behavior 55:280–284.https://doi.org/10.1016/j.yhbeh.2008.11.009
-
Encoding pheromonal signals in the mammalian vomeronasal systemCurrent Opinion in Neurobiology 14:428–434.https://doi.org/10.1016/j.conb.2004.07.001
-
Chapter 2: hypothalamic neural systems controlling the female reproductive life cycle gonadotropin-releasing hormone, glutamate, and GABAInternational Review of Cell and Molecular Biology 274:69–127.https://doi.org/10.1016/S1937-6448(08)02002-9
-
ConferencePheromones and vasanas: the functions of social chemosignalsNebraska Symposium on Motivation. Nebraska Symposium on Motivation.
-
Human olfactory bulb: aging of glomeruli and mitral cells and a search for the accessory olfactory bulbAnnals of the New York Academy of Sciences 855:708–715.https://doi.org/10.1111/j.1749-6632.1998.tb10649.x
-
Anxiety and deterioration of quality of life factors associated with recurrent miscarriage in an observational studyJournal of Perinatal Medicine 40:495–501.https://doi.org/10.1515/jpm-2011-0313
-
The scent of disease: human body odor contains an early chemosensory cue of sicknessPsychological Science 25:817–823.https://doi.org/10.1177/0956797613515681
-
Chemosignals, hormones and mammalian reproductionHormones and Behavior 63:723–741.https://doi.org/10.1016/j.yhbeh.2013.03.011
-
Chemosensory anxiety signals augment the startle reflex in humansNeuroscience Letters 394:127–130.https://doi.org/10.1016/j.neulet.2005.10.012
-
Male steroid hormones and female preference for male body odorEvolution and Human Behavior 27:259–269.https://doi.org/10.1016/j.evolhumbehav.2005.11.002
-
Olfactory bulb volume and depth of olfactory sulcus in patients with idiopathic olfactory lossEuropean Archives of Oto-Rhino-Laryngology 267:1551–1556.https://doi.org/10.1007/s00405-010-1230-2
-
The hypothalamus and reproductionAmerican Journal of Obstetrics and Gynecology 114:423–442.https://doi.org/10.1016/0002-9378(72)90620-5
-
The chemical bases of human socialityTrends in Cognitive Sciences 17:427–429.https://doi.org/10.1016/j.tics.2013.05.008
-
Female body odour is a potential cue to ovulationProceedings of the Royal Society of London. Series B: Biological Sciences 268:797–801.https://doi.org/10.1098/rspb.2001.1589
-
Fast robust automated brain extractionHuman Brain Mapping 17:143–155.https://doi.org/10.1002/hbm.10062
-
Parallel processing of social signals by the mammalian main and accessory olfactory systemsCellular and Molecular Life Sciences : CMLS 63:1476–1484.https://doi.org/10.1007/s00018-006-6109-4
-
Social chemosignals from breastfeeding women increase sexual motivationHormones and Behavior 46:362–370.https://doi.org/10.1016/j.yhbeh.2004.06.002
-
Frequency of factors associated with habitual abortion in 197 couplesFertility and Sterility 66:24–29.
-
Menstrual synchrony pheromones: cause for doubtHuman Reproduction 14:579–580.https://doi.org/10.1093/humrep/14.3.579
-
BookSTAI* a Questionnaire for the Assessment of State and Trait Anxiety: A Hebrew Manual for the ResearcherTel Aviv, Israel: Tel Aviv University Publishers.
-
Interrater reliability and agreement of subjective judgmentsJournal of Counseling Psychology 22:358–376.https://doi.org/10.1037/h0076640
-
The vomeronasal cavity in adult humansChemical Senses 25:369–380.https://doi.org/10.1093/chemse/25.4.369
-
BookThe Rites of PassageChicago: Psychology Press.https://doi.org/10.7208/chicago/9780226027180.001.0001
-
Modification of the oestrous cycle of the mouse by external stimuli associated with the maleJournal of Endocrinology 13:399–404.https://doi.org/10.1677/joe.0.0130399
-
Consequences of removing the vomeronasal organThe Journal of Steroid Biochemistry and Molecular Biology 39:661–669.https://doi.org/10.1016/0960-0760(91)90265-7
-
Segmentation of brain MR images through a hidden markov random field model and the expectation-maximization algorithmIEEE Transactions on Medical Imaging 20:45–57.https://doi.org/10.1109/42.906424
-
Encoding human sexual chemosensory cues in the orbitofrontal and fusiform corticesJournal of Neuroscience 28:14416–14421.https://doi.org/10.1523/JNEUROSCI.3148-08.2008
-
Male-mediated prenatal loss: functions and mechanismsEvolutionary Anthropology: Issues, News, and Reviews 28:114–125.https://doi.org/10.1002/evan.21776
Decision letter
-
Andreas T SchaeferReviewing Editor; The Francis Crick Institute, United Kingdom
-
Catherine DulacSenior Editor; Harvard University, United States
-
Matthew N ZippleReviewer; Duke University, United States
In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.
Acceptance summary:
Human pregnancies often end in miscarriages, with most pregnancy losses remaining unexplained. Olfaction plays a key role in mammalian reproduction in general. Strikingly, Rosenkrantz, Sobel and colleagues now find that – as a population – women that experience unexplained repeated pregnancy loss have altered olfactory perceptual ability, reduced olfactory bulb volume, and increased hypothalamic response to men's body odours. This elegant study combines a variety of approaches to convincingly correlate olfaction with repeated pregnancy loss in humans, potentially providing a new entry point for clinical investigations. Furthermore, it highlights the importance and behavioural relevance of the sense of smell in humans in general.
Decision letter after peer review:
Thank you for submitting your article "Unexplained Repeated Pregnancy Loss is Associated with Altered Perceptual and Brain Responses to Men's Body-Odor" for consideration by eLife. Your article has been reviewed by five peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Catherine Dulac as the Senior Editor The following individual involved in review of your submission has agreed to reveal their identity: Matthew N Zipple (Reviewer #4).
The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.
All reviewers agree that the data you present are both of high quality and importance. You present the intriguing finding that human females with repeated pregnancy loss (RPL) are better able to identify their spouse odour compared to controls. Moreover, you demonstrate differences in the anatomy of their olfactory system and differences in functional activity when exposed to male odours. All reviewers agree that the data – while correlative as you also discuss – are very clear, clearly presented, and compelling. However, there are major concerns that were consistently raised by virtually all reviewers (coming from very different fields):
Most importantly, while the correlations between olfactory function and RPL are compelling, the parallels to the Bruce-effect are not. Please do substantially tune down the Bruce effect analogies by restricting them to a discussion point and removing them from Abstract and Introduction. In the Discussion, however, it is necessary to have a more general discussion of the human (and in general primate) Bruce effect, its definition and the wider literature as outlined in particular by reviewers 4 and 5 (for that purpose I append the full reviews below).
Other major concerns (that you can all address with rewriting, adding raw data tables, and additional discussion):
• The Materials and methods indicate that for the experiments presented in Figures 1-2 each woman sniffed the SSD either 4-6 times. It is not clear why some women underwent two additional attempts. The “individual” data points represented in the figures are an average of these 4-6 trials. Please include a table of the raw data points for each individual to reveal the variance in response and identify the conditions where the additional trials occurred.
• Was the identity of the odor (spouse or stranger) blind to the researcher that presented the SSDs to each individual (I cannot find this in the Materials and methods)? If not, subliminal cues by the experimenter could alter the individual's perception and should be stated as a potential reason for behavioral differences observed.
• It is possible that the effects for spouse-odor identification are not at all related to RPL women's olfactory system but that they reside in their partners. As acknowledged by the authors, the men in RPL relationships may have unique body odors that are easy to be identified. One way to rule this out is to use the same set of odors in both groups, such that the spouse odor in one group is the non-spouse odor in the other. If the effects are driven by distinctive spouse-odors in the RPL group, women in both groups should be able to correctly identify this odor resulting in similar performance across groups. That is, the RPL group could correctly identify their spouse in the RPL group, and the control group could correctly identify these stand-out odors as the non-spouse odor. Ideally, the authors would rule out this important possibility in a control experiment. At the very least, however, this should be acknowledged and discussed. This confound is absent in many of the other experiments including the imaging studies. So while this potential explanation of the spouse-odor identification results needs to be taken seriously, it does not distract from the overall importance of this study.
• Because this study involves natural groups rather than randomization, there are a number of important confounds that need to be controlled for. For instance, groups may differ in the number of total pregnancies and other factors that are related to RPL. These factory could have an effect on olfaction in RPL women or on their ability to recognize their spouse's odor. Moreover, other factors may include the number of sexual partners, their living arrangements, etc. These and other factors could explain the present association and this should be tested as possible. In general, more needs to be said in the main text about how the groups were matched and how potential confounds were ruled out.
• A curious finding is the negative correlation between the composite olfactory score and the spouse identification performance in the RPL group. Would we not expect a positive correlation and importantly present across all participants?
Reviewer #1:
This manuscript describes a series of experiments testing for "Bruce effect"-like effects in humans. The key findings are that compared to controls, human females with repeated pregnancy loss (RPL) were able to identify the body odor of their spouses, despite relatively comparable olfactory perceptual ability. Compared to controls, these women rated stranger's body odors differently and also showed increased activity in the hypothalamus to such odors. In addition, women with RPL had significantly smaller olfactory bulbs.
This is a expertly conducted study on an important topic with interesting results. The findings are important because they may lead to new interventions for RPL. That said, there are a few issues that should be considered.
1) The effects for the spouse-odor identification test are strong and convincing. However, the results of the follow-up tests, aimed at identifying the specific cause of this advantage, are much weaker and less conclusive. From this perspective, it is less clear that women with RPL have an heightened sense of smell, and thus these findings should be interpreted with more caution. If there are no clear differences between groups, this opens up the possibility of alternative explanations. For instance, it is tempting to speculate that non-peripheral mechanisms, likely downstream of olfactory bulb are at play. Such effects could for instance be related to improved odor memory.
2) Alternatively, it is possible that the effects for spouse-odor identification are not at all related to RPL women's olfactory system but that they reside in their partners. As acknowledged by the authors, the men in RPL relationships may have unique body odors that are easy to be identified. One way to rule this out is to use the same set of odors in both groups, such that the spouse odor in one group is the non-spouse odor in the other. If the effects are driven by distinctive spouse-odors in the RPL group, women in both groups should be able to correctly identify this odor resulting in similar performance across groups. That is, the RPL group could correctly identify their spouse in the RPL group, and the control group could correctly identify these stand-out odors as the non-spouse odor. Ideally, the authors would rule out this important possibility in a control experiment. At the very least, however, this should be acknowledged and discussed. This confound is absent in many of the other experiments including the imaging studies. So while this potential explanation of the spouse-odor identification results needs to be taken seriously, it does not distract from the overall importance of this study.
3) Because this study involves natural groups rather than randomization, there are a number of important confounds that need to be controlled for. For instance, groups may differ in the number of total pregnancies and other factors that are related to RPL. These factory could have an effect on olfaction in RPL women or on their ability to recognize their spouse's odor. Moreover, other factors may include the number of sexual partners, their living arrangements, etc. These and other factors could explain the present association and this should be tested as possible. In general, more needs to be said in the main text about how the groups were matched and how potential confounds were ruled out.
4) A curious finding is the negative correlation between the composite olfactory score and the spouse identification performance in the RPL group. Would we not expect a positive correlation and importantly present across all participants?
Reviewer #2:
Rozenkrantz et al. investigate links between repeated pregnancy loss (RPL) and women's perception of male body odor, hypothesizing an underlying Bruce-like effect. The authors find several perceptual and neuroanatomical differences between RPL and control groups. These effects are certainly intriguing, especially the reported higher ability of RPL women to identify their spouses, as well as the neuroanatomical differences between this group and controls. However, these data do currently not allow to conclude a Bruce-like effect in humans. In my opinion, this study would therefore only be publishable in eLife conditional on addressing the concerns outlined below.
1) Extensive parallels are drawn to the Bruce effect, the mechanisms of which have most extensively been described in rodents. In mice, pheromone detection via the vomeronasal pathway and subsequent hypothalamic activation and dopamine release results in a loss of pregnancy only prior to implantation. In contrast, vomeronasal signalling is likely absent in humans and miscarriages (or pregnancies for that matter) would only be diagnosed long after embryo implantation. Even though the authors explicitly state that the Bruce effect is unlikely to be present in humans, they extensively refer to this phenomenon and e.g. place the hypothalamic activation in RPL women in this context. Although this limited equivalence is being addressed in the Discussion, allusions to a Bruce(-like) effect should feature less prominently throughout the Abstract/Introduction and Results sections.
2) The authors' use targeted analysis of their structural and functional imaging data to identify differences (OB volume, hypothalamic activation) between RPL women and controls. Yet, when the same data are analyzed in a hypothesis-free manner, the same differences cannot be identified (apart from when data are not corrected for multiple comparisons). This should be discussed.
3) A Bruce-like effect would rely on the correct identification of a smell as belonging to a Stranger. The study design chosen here however asked participants to distinguish between Spouse, Stranger and blank. While the authors argue that the correct identification of a familiar Spouse smell is a prerequisite for identifying other smells as Stranger, I am curious why participants were not asked to choose between e.g. two (or several) Spouse jars and one Stranger jar.
Reviewer #3:
In "Unexplained Repeated Pregnancy Loss is Associated with Altered Perceptual and Brain Responses to Men's Body-Odor" Rozenkrantz et al. explore the hypothesis that repeated pregnancy loss (RPL) in humans may be olfactory driven, similar to the Bruce effect in mice. When assayed for the ability of women to guess the odor of their spouse verses another male, they find the population of women that experienced RPL to perform better than women that were easily able to conceive. The correlation is extended to anatomical differences in the size of RPL olfactory bulbs and functional activity differences to male odors assayed by fMRI. Overall, the data is correlative, as is expected with human studies, there is no direct evidence that olfaction is contributing to RPL. As I read the manuscript, I tallied a list of major concerns, and then found them largely clearly stated in the last two paragraphs of the Discussion. The Bruce effect is only superficially analogous to the observations and experiments here, and the claims are much more limited than what is implied in the Abstract. With such a devastating condition as RPL, I do not think the story should be speculatively comingled with a unique aspect of behavior of a subset of species. I recommend removing most references to the Bruce effect prior to publication.
1) Though there is a statistical difference between RPL and control in Figures 1A/2/3B the response of both populations are very broad and the overall spread is very similar between RPL and control women. This looks like an example of a coincidental statistical difference – but not a biologically meaningful difference. I expect that if one was to look across all women (impossible) the results would not hold. With such a large variance in both populations in all measures of study, I do not think the data supports the conclusions.
2) The Materials and methods indicate that for the experiments presented in Figures 1-2 each woman sniffed the SSD either 4-6 times. It is not clear why some women underwent two additional attempts. The “individual” data points represented in the figures are an average of these 4-6 trials. Please include a table of the raw data points for each individual to reveal the variance in response and identify the conditions where the additional trials occurred.
3) Was the identity of the odor (spouse or stranger) blind to the researcher that presented the SSDs to each individual (I cannot find this in the Materials and methods)? If not, subliminal cues by the experimenter could alter the individual's perception and should be stated as a potential reason for behavioral differences observed.
4) The functional fMRI differences are confounded by watching video clips during assay. Why is this necessary? Of relevance here is the activity response towards odors only. The conclusions would be stronger if the stimuli were clear and controlled.
Reviewer #4:
The data that you present are of high quality and may be of great value to the RPL/medical literature, although I am unable to assess that aspect of your paper in detail. My reviewing expertise lies in the connection between your manuscript and the Bruce effect literature, and I believe that you would be better off avoiding discussion of the Bruce effect entirely. It is unclear that there is or should be any mechanistic or functional connection between RPL in humans and the Bruce effect, and I fear the addition of this manuscript to the Bruce effect literature would do more to confuse than to clarify.
I don't wish to belabour the below points more than I already have. In areas where my explanation may lack detail, I would recommend a thorough read of
deCatanzaro (2015) "Sex steroids as pheromones in mammals: the exceptional role of estradiol." Hormones and Behavior.
and
Zipple et al. (2019) Male-mediated prenatal loss: Functions and mechanisms." Evolutionary Anthropology.
1) The mechanistic argument: In order for use of the term "Bruce effect" or even "Bruce-like effect" to be justified, it seems necessary to show that there is some reason to believe that terminations of pregnancies are occurring after women are exposed to non-sire males.
You attempt to make this connection by showing that, compared to control women, women that display RPL are (1) more discriminating in their olfaction, especially their ability to distinguish between a spouse and a non-spouse, (2) have smaller olfactory bulbs, and (3) show differential brain response to men's body odors. The underlying logic appears to be that women who display RPL are sensitive to the smell of other men, and after smelling a non-spouse man, a chemical cascade is triggered that results in the termination of their pregnancies.
Such a line of argumentation would be much firmer if we had reason to believe that exposure to the odors of non-sire males is the trigger that leads to pregnancy termination across the many species that exhibit the Bruce effect, including some primates. Under such a scenario the lack of causality in the paper might justify the speculation that the presented data support a Bruce effect in humans. However, at present there is no evidence in any species that pregnancy termination (as opposed to implantation failure) is triggered by exposure to the scents of non-sire males. Even in mice, where the odor of non-sire males has been argued to trigger a Bruce effect by preventing implantation, it is unclear that exposure to the odor alone is sufficient to trigger the Bruce effect in the absence of exogenous estradiol.
What's more, it seems unlikely that such a complex and costly behavior of pregnancy termination could be triggered in humans simply as a result of a conspecific chemical cue, especially in a species in which females routinely encounter non-spouse males. Outside of rodents, there has so far been no evidence that odor or any other exogenous chemical mechanism is at play in species that exhibit the Bruce effect, such as dogs, horses, or primates.
While it is possible that olfactory cues play a role in post-implantation Bruce effects, it is also possible that the relevant cues in these cases are entirely social/visual and require no chemical cue whatsoever. Animals that are capable of recognizing each other visually, both because they have the cognitive capacity for such an assessment and because they live in a visual environment that allows for it, seem much more likely to rely on sociocognitive rather than exogenous chemical cues to trigger termination. A woman (or for that matter a female horse, dog, or lion) does not need to smell a non-sire male to know that he is present in her social group nor does she need to smell a male to know that he might pose a threat to the survival of her future children. Rather, she is able to assess changes in the dynamics of her group to make such determinations in a much more accurate way.
So, from a mechanistic standpoint, the speculation about associations between olfactory sensitivity and RPL does not appear justified.
2) The evolutionary argument: I also have concerns about the evolutionary story that you are telling here. Your argument appears to be that the ability of control women not to miscarry following exposure to the scent of non-sire men is an adaptation to communal living. The implication appears to be that termination following exposure to the scent of non-sire males is the ancestral state in primates, a claim that seems unlikely given the above discussion of mechanisms. It is also a confusing claim, because a great many primates live "communally," but there are certainly specie in which the Bruce effect does not occur (e.g. chacma baboons and yellow baboons). So when did this adaptation occur? Was it a prerequisite for group-living? If so, then it must have occurred millions of years ago.
If we take this interpretation as accurate, the unstated evolutionary argument that you are making is that the ancestral chemical cascade that had previously induced termination following exposure to non-sire males was broken via a mutation in an olfactory receptor gene (OR4C16G>A). For millions of years since that cascade was broken, it has remained otherwise entirely functional except that this change in one olfactory reception gene has caused it to no longer be triggered. So, when women are born with a mutation in the olfactory reception gene that makes it functional once more, the necessary chemical pathways that lead from exposure to male odors to pregnancy termination kick in and continue to work just as they would have millions of years previously. This seems to me to be a very unlikely scenario. There are additional issues that I see in terms of the evolutionary narrative at play, but I will not belabor these points further.
I therefore see no reason to refer to the Bruce effect in this manuscript, except as perhaps a passing comment in the Discussion that suggests it is possible that there may be some connection. Even this feels like more speculation than may be warranted.
Reviewer #5:
This manuscript describes experiments suggested by the argument that human females exhibit a Bruce Effect in which pregnant women spontaneously abort gestation when exposed to the odor of men other those who impregnated them. The experiments involve exposing women with and without histories of repeated pregnancy loss to odors of partners as well as other men to determine, in general, whether those with such histories detect partner odor better than controls, whether their olfactory biology differs from controls, and whether they react differently to arousing stimuli when exposed to male body odors. The pattern of results support the argument that women with a history of repeated pregnancy loss appear different from controls in way consistent with a Bruce Effect in women.
Those of us inclined to believe that humans exhibit Bruce Effect like behavior will find your arguments and findings powerful. Skeptics will offer caveats many of which you anticipated, and both directly and indirectly defend against. But to further avoid capitalizing on confirmation bias among readers, I suggest you alert them to two circumstances. First, the literature that explains, rather than describes, the "Bruce Effect" characterizes the termination of gestation when the environment threatens the survival of offspring as adaptive. Natural selection presumably conserved mutations that produce the Effect because they increase the reproductive fitness of females. They should, therefore, appear frequently among females in a stable, self-sustaining population. You, however, study a clinically defined sub-group of the population. A sub-group that exhibits a non-adaptive behavior, very frequent pregnancy loss. Your findings could be characterized as identifying risk factors for inclusion in a relatively small group that does not contribute to a self-sustaining population (i.e., 2-4% of all women who suffer frequent clinically recognized fetal loss), rather than providing evidence for an adaptive mechanism that, if the Bruce Effect argument is correct, should appear in a large fraction of women. You need, in short, to discuss the external validity of your findings and its implications for theory.
Second, you should spend a bit of page space making clear that the literature includes tests of a generalized Bruce Effect in humans. That work shows that, consistent with the theory underlying the Bruce Effect, populations witnessing unexpected death among children suffer greater than expected male fetal loss. The work remains controversial because it invokes "Bruce Effect" to describe the termination of gestation following something other than exposure to the odor of men. But the work does allow you to argue that your reasoning and findings may apply to circumstances other than rare cases of repeated pregnancy loss.
https://doi.org/10.7554/eLife.55305.sa1Author response
Most importantly, while the correlations between olfactory function and RPL are compelling, the parallels to the Bruce-effect are not. Please do substantially tune down the Bruce effect analogies by restricting them to a discussion point and removing them from Abstract and Introduction. In the Discussion, however, it is necessary to have a more general discussion of the human (and in general primate) Bruce effect, its definition and the wider literature as outlined in particular by reviewers 4 and 5 (for that purpose I append the full reviews below).
We have removed the Bruce effect from the Abstract, Introduction, and Results, and now only carefully allude to it in the final paragraph of the Discussion. This was initially hard to do… because the Bruce-effect frame was not an after-thought of our results, but rather genuinely the impetus for conducting this study in the first place. It was our hypothesis. Thus, emotionally, it was hard to let go. However, now in reading our final revised version, I must admit that this was the right thing to do, and this made for a better manuscript.
Other major concerns (that you can all address with rewriting, adding raw data tables, and additional discussion):
• The Materials and methods indicate that for the experiments presented in Figures 1-2 each woman sniffed the SSD either 4-6 times. It is not clear why some women underwent two additional attempts. The “individual” data points represented in the figures are an average of these 4-6 trials. Please include a table of the raw data points for each individual to reveal the variance in response and identify the conditions where the additional trials occurred.
Some participants had different numbers of trials for the following reason: This study was conducted over many years, and naturally, we evolved some of the paradigms on the way, in order to improve them. In this case, it dawned on us that it would be more powerful to be able to assess significance also within each participant alone and not only across groups, so we added trials per participant. That said, critically, the number of trials remained equal across cohorts. In other words, for every RPL participant with a reduced number or trials there was a matched control with equally reduced trials. Thus, this shortcoming cannot underlie the group effect that we report. Moreover, in consideration of this comment, we now tested again, using only the first four trials from all those participants that had 6 trials, and we obtain ostensibly the same results. Finally, there were 3 cases where we had only two trials (I will remind that these are 3AFC trials, not two alternative). If we restrict the entire cohort to just two trials, the effect again still remains significant. To conclude, the group effect we reported was not influenced by the number of trials conducted per participant. Finally, consistent with the reviewer suggestion, we have now included a table (Supplementary file 2) which details all trials for each participant, and we added these sub-analyses using restricted portions of the data for verification.
• Was the identity of the odor (spouse or stranger) blind to the researcher that presented the SSDs to each individual (I cannot find this in the Materials and methods)? If not, subliminal cues by the experimenter could alter the individual's perception and should be stated as a potential reason for behavioral differences observed.
This experimental design was not double-blind in the strictest sense of the term, but its arrangement largely prevented the risk of experimenter-generated cues. More specifically, during each trial the experimenter did not hand the participant one SSD at a time, but rather placed three arbitrarily marked SSDs on the table in front of the participant at the beginning of the trial. The experimenter then moved away, and the participant alone interacted with a computer that provided sampling instructions. The participant then entered their selection using the trackpad of the computer, such that the experimenter did not know what the participant answered at that time. There was no voice interaction between the experimenter and participant during a trial. In other words, the timing of each sniff, its content, and answer, were all independently controlled by the participant without experimenter involvement. Finally, we used different experimenters across participants. Nevertheless, we acknowledge that this is not genuinely double-blind, and moreover, we acknowledge that this is a drawback. Indeed, in retrospect, an unnecessary drawback. This has now been better detailed, and explicitly acknowledged, in the manuscript.
• It is possible that the effects for spouse-odor identification are not at all related to RPL women's olfactory system but that they reside in their partners. As acknowledged by the authors, the men in RPL relationships may have unique body odors that are easy to be identified. One way to rule this out is to use the same set of odors in both groups, such that the spouse odor in one group is the non-spouse odor in the other. If the effects are driven by distinctive spouse-odors in the RPL group, women in both groups should be able to correctly identify this odor resulting in similar performance across groups. That is, the RPL group could correctly identify their spouse in the RPL group, and the control group could correctly identify these stand-out odors as the non-spouse odor. Ideally, the authors would rule out this important possibility in a control experiment. At the very least, however, this should be acknowledged and discussed. This confound is absent in many of the other experiments including the imaging studies. So while this potential explanation of the spouse-odor identification results needs to be taken seriously, it does not distract from the overall importance of this study.
This is a keen observation, that we had indeed acknowledged in the manuscript, but apparently we had not done so sufficiently. That men in RPL relationships have a unique body-odor fingerprint is a tantalizing possibility. Alas, our experiments were not designed to address this question. Although the stranger in each experiment was indeed the spouse of a different participant (as here suggested), whether this stranger was an RPL-man or a control-man was not well balanced across groups (as we had not imagined this would be a factor). We are certain to conduct a future study probing this exact question, and here currently further stress this possibility as here advised. In addition to the note in the body of the manuscript, we now note in the Discussion:
“We would like to reiterate the possibility mentioned in the results whereby theoretically, this advantage may reflect not better olfactory memory/identification in uRPL women, but rather a unique body-odor in uRPL men. Alas, our experimental design did not permit addressing this question in the current results, and this remains an open and tantalizing question for future research. That said, if there is a body-odor difference in uRPL men, we speculate that this is in addition and not instead of altered body-odor perception in uRPL women. We say this because that in addition to better performance in spouse identification, uRPL women also reported different perceptual qualities associated with a group of independent men. Women with and without uRPL rated the same group of men, and the women with RPL perceived them differently (Figure 2). In combination, we conclude that women with uRPL have altered perception of men's body-odor.”
• Because this study involves natural groups rather than randomization, there are a number of important confounds that need to be controlled for. For instance, groups may differ in the number of total pregnancies and other factors that are related to RPL. These factory could have an effect on olfaction in RPL women or on their ability to recognize their spouse's odor. Moreover, other factors may include the number of sexual partners, their living arrangements, etc. These and other factors could explain the present association and this should be tested as possible. In general, more needs to be said in the main text about how the groups were matched and how potential confounds were ruled out.
We were of course very cognizant of matching controls in this study. Because these are human participants and not mice, we cannot match them on all fronts. Thus, we are forced to rank-order the importance of the matching criteria, and maximally match on the most important fronts, at the risk of variability on other counts. Our two most important matching criteria were: 1. Age, 2. Number of living children and their age. Beyond that, we always tried to match means on all other fronts. In other words, even if we did have individual differences on lower-order considerations, such as living arrangements, these differences were not systematically associated with group (uRPL vs. Control).
We further note that as to: "groups may differ in the number of total pregnancies", this is of course the case by definition. Although we are not aware of any study suggesting that number of pregnancies influences olfaction, this is of course a possibility, but this difference is inherent to our study. Given equal number of live children, the uRPL group will always have more pregnancies. To ask if number of pregnancies is related to these results, what we can do is to take the most prominent behavioral result we have: spouse identification, and see if it is associated with number of pregnancies within each group (collapsing across both groups would be biased since RPL women were better at the task and, as said, have higher number of pregnancies). If the correlation reflects a relation of olfaction and pregnancies, it should materialize within the groups as well. Doing so we find no significant correlations. This control analysis has been added to the text. Regarding number of sexual partners and living arrangements, all women in the identification task, and the overwhelming majority of women in the experiment, were married and living with their spouses at time of the experiment. We now note this in the Materials and methods as well:
“Because this study involves natural groups rather than randomization, we paid close attention to controlling for factors which may confound our findings. For each comparison performed, we matched control participants to uRPL participants based on their age and number of children (see Tables 1-4 for details). If multiple choices were available, our choice was based on matching the average of the control group as a whole. For the olfactory bulb analysis, following recent findings regarding the impact of pregnancy on maternal brain anatomy after birth (Kim et al., 2010; Hoekzema et al., 2017; Kim et al., 2018), we additionally matched for age of offspring, particularly age of the youngest child (Table 3). Additional factors, like marital status and living arrangements, were similar between groups: the overwhelming majority of women in our study were married and living with their spouses. One factor that is by definition confounded with group differences is number of pregnancies, as RPL is defined as 2 or more consecutive miscarriages. To ask whether number of pregnancies may be associated with our results, we examined the most prominent behavioral result, namely spouse identification, and its association with number of pregnancies within each group. We observed no significant correlations (RPL group: Spearman’s Rho = 0.19, p = 0.3; Control group: Spearman’s Rho = -0.29, p = 0.11), implying that number of pregnancies alone does not impact this result."
• A curious finding is the negative correlation between the composite olfactory score and the spouse identification performance in the RPL group. Would we not expect a positive correlation and importantly present across all participants?
First, we would like to caution that this negative correlation was not statistically significant, so we would not characterize it as a "finding". Nevertheless, a trend is there, and it is counter-intuitive. Indeed, we would expect a positive correlation, and that was certainly not there. Although we could hand-wave, we have absolutely no good explanation for this. Thus, all we can do is make sure to clearly note this, and retain it as an open observation. We now stress this in the manuscript.
Reviewer #1:
This manuscript describes a series of experiments testing for "Bruce effect"-like effects in humans. The key findings are that compared to controls, human females with repeated pregnancy loss (RPL) were able to identify the body odor of their spouses, despite relatively comparable olfactory perceptual ability. Compared to controls, these women rated stranger's body odors differently and also showed increased activity in the hypothalamus to such odors. In addition, women with RPL had significantly smaller olfactory bulbs.
This is a expertly conducted study on an important topic with interesting results. The findings are important because they may lead to new interventions for RPL. That said, there are a few issues that should be considered.
1) The effects for the spouse-odor identification test are strong and convincing. However, the results of the follow-up tests, aimed at identifying the specific cause of this advantage, are much weaker and less conclusive. From this perspective, it is less clear that women with RPL have an heightened sense of smell, and thus these findings should be interpreted with more caution. If there are no clear differences between groups, this opens up the possibility of alternative explanations. For instance, it is tempting to speculate that non-peripheral mechanisms, likely downstream of olfactory bulb are at play. Such effects could for instance be related to improved odor memory.
We completely agree, and this is in fact the outcome we were trying to convey. If there is any overall olfactory advantage in RPL, it is very minimal. We just didn't want to be in the place of even remotely obscuring such an advantage, as a claim that the RPL advantage is limited to only social odors, although consistent with our initial hypothesis, may be a stretch. Nevertheless, we now better clarify this in the Discussion.
2) Alternatively, it is possible that the effects for spouse-odor identification are not at all related to RPL women's olfactory system but that they reside in their partners. As acknowledged by the authors, the men in RPL relationships may have unique body odors that are easy to be identified. One way to rule this out is to use the same set of odors in both groups, such that the spouse odor in one group is the non-spouse odor in the other. If the effects are driven by distinctive spouse-odors in the RPL group, women in both groups should be able to correctly identify this odor resulting in similar performance across groups. That is, the RPL group could correctly identify their spouse in the RPL group, and the control group could correctly identify these stand-out odors as the non-spouse odor. Ideally, the authors would rule out this important possibility in a control experiment. At the very least, however, this should be acknowledged and discussed. This confound is absent in many of the other experiments including the imaging studies. So while this potential explanation of the spouse-odor identification results needs to be taken seriously, it does not distract from the overall importance of this study.
We completely concur. Beyond comments in the results, we now have a section in the Discussion addressing both this and the previous comment:
"We would like to reiterate two considerations related to this primary behavioral finding. First, considering the only marginal uRPL advantage at other olfactory tasks, this advantage may reflect some downstream uRPL advantage, perhaps even memory beyond the olfactory system. Second, as mentioned in the results, this advantage may reflect not better olfactory memory/identification in uRPL women, but rather a unique body-odor in uRPL men. Alas, our experimental design did not permit addressing this question in the current results, and this remains an open and tantalizing question for future research. That said, if there is a body-odor difference in uRPL men, we speculate that this is in addition and not instead of altered body-odor perception in uRPL women. We say this because in addition to better performance in spouse identification, uRPL women also reported different perceptual qualities associated with a group of independent men. Women with and without uRPL rated the same group of non-spouse men, and the women with RPL perceived them differently (Figure 2). Thus, we stand by the notion that women with uRPL have altered perception of men's body-odor."
3) Because this study involves natural groups rather than randomization, there are a number of important confounds that need to be controlled for. For instance, groups may differ in the number of total pregnancies and other factors that are related to RPL. These factory could have an effect on olfaction in RPL women or on their ability to recognize their spouse's odor. Moreover, other factors may include the number of sexual partners, their living arrangements, etc. These and other factors could explain the present association and this should be tested as possible. In general, more needs to be said in the main text about how the groups were matched and how potential confounds were ruled out.
As noted in the reply to Editors: We were of course very cognizant of matching controls in this study. Because these are human participants and not mice, we cannot match them on all fronts. Thus, we are forced to rank-order the importance of the matching criteria, and maximally match on the most important fronts, at the risk of variability on other counts. Our two most important matching criteria were: 1. Age, 2. Number of living children and their age. Beyond that, we always tried to match means on all other fronts. In other words, even if we did have individual differences on lower-order considerations, such as living arrangements, these differences were not systematically associated with group (uRPL vs. Control).
We further note that as to: "groups may differ in the number of total pregnancies", this is of course the case by definition. Although we are not aware of any study suggesting that number of pregnancies influences olfaction, this is of course a possibility, but this difference is inherent to our study. Given equal number of live children, the uRPL group will always have more pregnancies. To ask if number of pregnancies is related to these results, what we can do is to take the most prominent behavioral result we have: spouse identification, and see if it is associated with number of pregnancies within each group (collapsing across both groups would be biased since RPL women were better at the task and, as said, have higher number of pregnancies). If the correlation reflects a relation of olfaction and pregnancies, it should materialize within the groups as well. Doing so we find no significant correlations (both p > 0.1). This control analysis has been added to the text. Regarding number of sexual partners and living arrangements, all women in the identification task, and the overwhelming majority of women in the experiment, were married and living with their spouses at time of the experiment. We now note this in the Materials and methods as well:
“Because this study involves natural groups rather than randomization, we paid close attention to controlling for factors which may confound our findings. For each comparison performed, we matched control participants to uRPL participants based on their age and number of children (see Tables 1-4 for details). If multiple choices were available, our choice was based on matching the average of the control group as a whole. For the olfactory bulb analysis, following recent findings regarding the impact of pregnancy on maternal brain anatomy after birth (Kim et al., 2010; Hoekzema et al., 2017; Kim et al., 2018), we additionally matched for age of offspring, particularly age of the youngest child (Table 3). Additional factors, like marital status and living arrangements, were similar between groups: the overwhelming majority of women in our study were married and living with their spouses. One factor that is by definition confounded with group differences is number of pregnancies, as RPL is defined as 2 or more consecutive miscarriages. To ask whether number of pregnancies may be associated with our results, we examined the most prominent behavioral result, namely spouse identification, and its association with number of pregnancies within each group. We observed no significant correlations (RPL group: Spearman’s Rho = 0.19, p = 0.3; Control group: Spearman’s Rho = -0.29, p = 0.11), implying that number of pregnancies alone does not impact this result."
4) A curious finding is the negative correlation between the composite olfactory score and the spouse identification performance in the RPL group. Would we not expect a positive correlation and importantly present across all participants?
As noted in the reply to Editors: First, we would like to caution that this negative correlation was not statistically significant, so we would not characterize it as a "finding". Nevertheless, a trend is there, and it is counter-intuitive. Indeed, we would expect a positive correlation, and that was certainly not there. Although we could hand-wave, we have absolutely no good explanation for this. Thus, all we can do is make sure to clearly note this, and retain it as an open observation. We now stress this in the manuscript.
Reviewer #2:
Rozenkrantz et al. investigate links between repeated pregnancy loss (RPL) and women's perception of male body odor, hypothesizing an underlying Bruce-like effect. The authors find several perceptual and neuroanatomical differences between RPL and control groups. These effects are certainly intriguing, especially the reported higher ability of RPL women to identify their spouses, as well as the neuroanatomical differences between this group and controls. However, these data do currently not allow to conclude a Bruce-like effect in humans. In my opinion, this study would therefore only be publishable in eLife conditional on addressing the concerns outlined below.
1) Extensive parallels are drawn to the Bruce effect, the mechanisms of which have most extensively been described in rodents. In mice, pheromone detection via the vomeronasal pathway and subsequent hypothalamic activation and dopamine release results in a loss of pregnancy only prior to implantation. In contrast, vomeronasal signalling is likely absent in humans and miscarriages (or pregnancies for that matter) would only be diagnosed long after embryo implantation. Even though the authors explicitly state that the Bruce effect is unlikely to be present in humans, they extensively refer to this phenomenon and e.g. place the hypothalamic activation in RPL women in this context. Although this limited equivalence is being addressed in the Discussion, allusions to a Bruce(-like) effect should feature less prominently throughout the Abstract/Introduction and Results sections.
We embrace this criticism, and have deleted all analogies with the Bruce effect from the Abstract, Introduction, and Results, and now retain only a very cautious consideration of this at the very final paragraph of the Discussion alone.
2) The authors' use targeted analysis of their structural and functional imaging data to identify differences (OB volume, hypothalamic activation) between RPL women and controls. Yet, when the same data are analyzed in a hypothesis-free manner, the same differences cannot be identified (apart from when data are not corrected for multiple comparisons). This should be discussed.
As noted in the reply to Editors: We acknowledge this, but stress that it is in no way an anomaly. For example, previous studies that uncover a hypothalamic response to Androstadienone were all based on region of interest (ROI) analyses:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3397979/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1129091/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4037295/ Nevertheless, we now further stress this in the Discussion.
3) A Bruce-like effect would rely on the correct identification of a smell as belonging to a Stranger. The study design chosen here however asked participants to distinguish between Spouse, Stranger and blank. While the authors argue that the correct identification of a familiar Spouse smell is a prerequisite for identifying other smells as Stranger, I am curious why participants were not asked to choose between e.g. two (or several) Spouse jars and one Stranger jar.
As noted in the reply to Editors: As our initial intentions were to test whether RPL women were better able to recognize their spouse’s body odor than control women, the task we designed was an identification task. Had we used the design suggested here, we would have been testing their discrimination abilities, that is: given two identical odors and one different odor, are they able to tell the different odor apart. Thus, we constructed the task in accordance with our motivation to examine spouse recognition.
Reviewer #3:
In "Unexplained Repeated Pregnancy Loss is Associated with Altered Perceptual and Brain Responses to Men's Body-Odor" Rozenkrantz et al. explore the hypothesis that repeated pregnancy loss (RPL) in humans may be olfactory driven, similar to the Bruce effect in mice. When assayed for the ability of women to guess the odor of their spouse verses another male, they find the population of women that experienced RPL to perform better than women that were easily able to conceive. The correlation is extended to anatomical differences in the size of RPL olfactory bulbs and functional activity differences to male odors assayed by fMRI. Overall, the data is correlative, as is expected with human studies, there is no direct evidence that olfaction is contributing to RPL. As I read the manuscript, I tallied a list of major concerns, and then found them largely clearly stated in the last two paragraphs of the Discussion. The Bruce effect is only superficially analogous to the observations and experiments here, and the claims are much more limited than what is implied in the Abstract. With such a devastating condition as RPL, I do not think the story should be speculatively comingled with a unique aspect of behavior of a subset of species. I recommend removing most references to the Bruce effect prior to publication.
We embrace this criticism. As noted in the reply to Editors, we have removed all mention of the Bruce effect from the Abstract, Introduction and Results, and have limited its consideration to a careful final comment in the Discussion.
1) Though there is a statistical difference between RPL and control in Figures 1A/2/3B the response of both populations are very broad and the overall spread is very similar between RPL and control women. This looks like an example of a coincidental statistical difference – but not a biologically meaningful difference. I expect that if one was to look across all women (impossible) the results would not hold. With such a large variance in both populations in all measures of study, I do not think the data supports the conclusions.
This is a keen observation, and to address this possibility we conducted bootstrap analyses for all three results. In each case, we randomly shuffled the assignment of the data to uRPL or Control 10,000 times, and reconducted the analysis. We observe that in all three cases, our result is far removed from chance. More specifically, the probability of obtaining these results coincidentally is 2 in 1000 for Figure 1A, 6 in 1000 for Figure 2, and 2 in 100 for Figure 3B. Based on this analysis, we think the results are quite robust. We have added this important control analysis to all the figures.
2) The Materials and methods indicate that for the experiments presented in Figures 1-2 each woman sniffed the SSD either 4-6 times. It is not clear why some women underwent two additional attempts. The “individual” data points represented in the figures are an average of these 4-6 trials. Please include a table of the raw data points for each individual to reveal the variance in response and identify the conditions where the additional trials occurred.
As noted in the reply to Editors: Some participants had different numbers of trials for the following reason: This study was conducted over many years (the recruitment rate was slow), and naturally, we evolved some of the paradigms on the way, in order to improve them. In this case, it dawned on us that it would be more powerful to be able to assess significance also within each participant alone and not only across groups, so we added trials. That said, critically, the number of trials remained equal across cohorts. In other words, for every RPL participant with a reduced number or trials there was a matched control with equally reduced trials. Thus, this shortcoming cannot underlie the group effect that we report. Moreover, in consideration of this comment, we now tested again, using only the first four trials from all those participants that had 6 trials, and we obtain ostensibly the same results. Finally, there were 3 cases where we had only two trials. If we restrict the entire cohort to just two trials, the effect again still remains significant. To conclude, the group effect we reported was not influenced by the number of trials conducted per participant. Finally, consistent with the reviewer suggestion, we have now included a table (Supplementary file 2) which details all trials for each participant, and we added these sub-analyses.
3) Was the identity of the odor (spouse or stranger) blind to the researcher that presented the SSDs to each individual (I cannot find this in the Materials and methods)? If not, subliminal cues by the experimenter could alter the individual's perception and should be stated as a potential reason for behavioral differences observed.
As noted in the reply to Editors: This experimental design was not double-blind in the strictest sense of the term, but its arrangement largely prevented the risk of experimenter-generated cues. More specifically, during each trial the experimenter did not hand the participant one SSD at a time, but rather placed three arbitrarily marked SSDs on the table in front of the participant at the beginning of the trial. The experimenter then moved away, and the participant alone interacted with a computer that provided sampling instructions. The participant then entered their selection using the trackpad of the computer, such that the experimenter did not know what the participant answered at that time. There was no voice interaction between the experimenter and participant during a trial. In other words, the timing of each sniff, its content, and answer, were all independently controlled by the participant without experimenter involvement. Nevertheless, we acknowledge that this is not genuinely double-blind, and moreover, we acknowledge that this is a drawback. Indeed, in retrospect, an unnecessary drawback.
4) The functional fMRI differences are confounded by watching video clips during assay. Why is this necessary? Of relevance here is the activity response towards odors only. The conclusions would be stronger if the stimuli were clear and controlled.
As noted in the reply to Editors: We agree that the direct response to odor alone is an interesting path to take. That said, we opted to go for modulation of activity related to arousal because of a set of considerations that we think are strong. Specifically: uRPL participants are a "valuable" participant type. That is, we cannot recruit endless participants as we could Controls, and therefore, we wanted to maximize chances of obtaining meaningful signal in our first shot. A subliminal odor alone drives very minimal patterns of activity, but is far more evident in the brain in its modulation of other, related task activations. This notion of social odors as a human brain "modulator" rather than brain "activator" has been well validated (Jacob, Hayreh and McClintock, 2001). Of course, had our interest been the nature of response to body-odors alone, this would have been more of a problem, but given that our interest is the difference between groups as a function of body-odor, we judged this to be the correct path to take. In light of this comment, we have added text to justify this path in the manuscript.
Reviewer #4:
The data that you present are of high quality and may be of great value to the RPL/medical literature, although I am unable to assess that aspect of your paper in detail. My reviewing expertise lies in the connection between your manuscript and the Bruce effect literature, and I believe that you would be better off avoiding discussion of the Bruce effect entirely. It is unclear that there is or should be any mechanistic or functional connection between RPL in humans and the Bruce effect, and I fear the addition of this manuscript to the Bruce effect literature would do more to confuse than to clarify.
I don't wish to belabour the below points more than I already have. In areas where my explanation may lack detail, I would recommend a thorough read of
deCatanzaro (2015) "Sex steroids as pheromones in mammals: the exceptional role of estradiol." Hormones and Behavior.
and
Zipple et al. (2019) Male-mediated prenatal loss: Functions and mechanisms." Evolutionary Anthropology.
We have now read both, and thank you for directing us in this
1) The mechanistic argument: In order for use of the term "Bruce effect" or even "Bruce-like effect" to be justified, it seems necessary to show that there is some reason to believe that terminations of pregnancies are occurring after women are exposed to non-sire males.
You attempt to make this connection by showing that, compared to control women, women that display RPL are (1) more discriminating in their olfaction, especially their ability to distinguish between a spouse and a non-spouse, (2) have smaller olfactory bulbs, and (3) show differential brain response to men's body odors. The underlying logic appears to be that women who display RPL are sensitive to the smell of other men, and after smelling a non-spouse man, a chemical cascade is triggered that results in the termination of their pregnancies.
Such a line of argumentation would be much firmer if we had reason to believe that exposure to the odors of non-sire males is the trigger that leads to pregnancy termination across the many species that exhibit the Bruce effect, including some primates. Under such a scenario the lack of causality in the paper might justify the speculation that the presented data support a Bruce effect in humans. However, at present there is no evidence in any species that pregnancy termination (as opposed to implantation failure) is triggered by exposure to the scents of non-sire males. Even in mice, where the odor of non-sire males has been argued to trigger a Bruce effect by preventing implantation, it is unclear that exposure to the odor alone is sufficient to trigger the Bruce effect in the absence of exogenous estradiol.
What's more, it seems unlikely that such a complex and costly behavior of pregnancy termination could be triggered in humans simply as a result of a conspecific chemical cue, especially in a species in which females routinely encounter non-spouse males. Outside of rodents, there has so far been no evidence that odor or any other exogenous chemical mechanism is at play in species that exhibit the Bruce effect, such as dogs, horses, or primates.
While it is possible that olfactory cues play a role in post-implantation Bruce effects, it is also possible that the relevant cues in these cases are entirely social/visual and require no chemical cue whatsoever. Animals that are capable of recognizing each other visually, both because they have the cognitive capacity for such an assessment and because they live in a visual environment that allows for it, seem much more likely to rely on sociocognitive rather than exogenous chemical cues to trigger termination. A woman (or for that matter a female horse, dog, or lion) does not need to smell a non-sire male to know that he is present in her social group nor does she need to smell a male to know that he might pose a threat to the survival of her future children. Rather, she is able to assess changes in the dynamics of her group to make such determinations in a much more accurate way.
So, from a mechanistic standpoint, the speculation about associations between olfactory sensitivity and RPL does not appear justified.
Although we feel slightly odd to here provide a short answer to this very detailed and long concern, we can briefly say that we accepted and embraced this point of view, and therefore completely reframed our manuscript, deleting nearly all reference to the Bruce effect from the Abstract, Introduction, and body, and retaining only minimal allusion to it in the Discussion, as suggested by the Editors.
2) The evolutionary argument: I also have concerns about the evolutionary story that you are telling here. Your argument appears to be that the ability of control women not to miscarry following exposure to the scent of non-sire men is an adaptation to communal living. The implication appears to be that termination following exposure to the scent of non-sire males is the ancestral state in primates, a claim that seems unlikely given the above discussion of mechanisms. It is also a confusing claim, because a great many primates live "communally," but there are certainly specie in which the Bruce effect does not occur (e.g. chacma baboons and yellow baboons). So when did this adaptation occur? Was it a prerequisite for group-living? If so, then it must have occurred millions of years ago.
If we take this interpretation as accurate, the unstated evolutionary argument that you are making is that the ancestral chemical cascade that had previously induced termination following exposure to non-sire males was broken via a mutation in an olfactory receptor gene (OR4C16G>A). For millions of years since that cascade was broken, it has remained otherwise entirely functional except that this change in one olfactory reception gene has caused it to no longer be triggered. So, when women are born with a mutation in the olfactory reception gene that makes it functional once more, the necessary chemical pathways that lead from exposure to male odors to pregnancy termination kick in and continue to work just as they would have millions of years previously. This seems to me to be a very unlikely scenario. There are additional issues that I see in terms of the evolutionary narrative at play, but I will not belabor these points further.
I therefore see no reason to refer to the Bruce effect in this manuscript, except as perhaps a passing comment in the Discussion that suggests it is possible that there may be some connection. Even this feels like more speculation than may be warranted.
Again, we embraced this point of view, and radically edited the manuscript to match.
Reviewer #5:
This manuscript describes experiments suggested by the argument that human females exhibit a Bruce Effect in which pregnant women spontaneously abort gestation when exposed to the odor of men other those who impregnated them. The experiments involve exposing women with and without histories of repeated pregnancy loss to odors of partners as well as other men to determine, in general, whether those with such histories detect partner odor better than controls, whether their olfactory biology differs from controls, and whether they react differently to arousing stimuli when exposed to male body odors. The pattern of results support the argument that women with a history of repeated pregnancy loss appear different from controls in way consistent with a Bruce Effect in women.
Those of us inclined to believe that humans exhibit Bruce Effect like behavior will find your arguments and findings powerful. Skeptics will offer caveats many of which you anticipated, and both directly and indirectly defend against. But to further avoid capitalizing on confirmation bias among readers, I suggest you alert them to two circumstances. First, the literature that explains, rather than describes, the "Bruce Effect" characterizes the termination of gestation when the environment threatens the survival of offspring as adaptive. Natural selection presumably conserved mutations that produce the Effect because they increase the reproductive fitness of females. They should, therefore, appear frequently among females in a stable, self-sustaining population. You, however, study a clinically defined sub-group of the population. A sub-group that exhibits a non-adaptive behavior, very frequent pregnancy loss. Your findings could be characterized as identifying risk factors for inclusion in a relatively small group that does not contribute to a self-sustaining population (i.e., 2-4% of all women who suffer frequent clinically recognized fetal loss), rather than providing evidence for an adaptive mechanism that, if the Bruce Effect argument is correct, should appear in a large fraction of women. You need, in short, to discuss the external validity of your findings and its implications for theory.
Put simply: we could not agree more, and although this is a document where the reviewer is commenting on our work, and not the reverse :-), we would nevertheless like to say that the above is very well put. That said, because the review consensus was that we remove most all reference to the Bruce effect, and we grudgingly accepted this, all this becomes rather moot. The Bruce effect is now largely out of this manuscript. We will soon set out on a line of added experimentation that if successful, will imply causation, and if we get there, we would like to use the reviewer comment verbatim.
Second, you should spend a bit of page space making clear that the literature includes tests of a generalized Bruce Effect in humans. That work shows that, consistent with the theory underlying the Bruce Effect, populations witnessing unexpected death among children suffer greater than expected male fetal loss. The work remains controversial because it invokes "Bruce Effect" to describe the termination of gestation following something other than exposure to the odor of men. But the work does allow you to argue that your reasoning and findings may apply to circumstances other than rare cases of repeated pregnancy loss.
Although, as noted, the Bruce effect has been largely stripped from this manuscript, we now address this perspective in the Discussion.
https://doi.org/10.7554/eLife.55305.sa2Article and author information
Author details
Funding
H2020 European Research Council (670798)
- Noam Sobel
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Acknowledgements
This study was funded by a European Research Council AdG. grant #670798 (SocioSmell) awarded to NS. General work in the Sobel lab is funded by the Rob and Cheryl McEwen Fund for Brain Research. Dr. E Furman-Haran holds the Calin and Elaine Rovinescu Research Fellow Chair for Brain Research. We thank Dr. Noa Ofen for fMRI consultation.
Ethics
Human subjects: All participants provided written informed consent to procedures approved by the Tel- Hashomer (behavioral studies, protocol reference number 8067-10-SMC) or Wolfson hospital (imaging studies, protocol reference number 0121-10-WOMC) Helsinki Committees. All participants were screened for good general health and no history of psychiatric disease.
Senior Editor
- Catherine Dulac, Harvard University, United States
Reviewing Editor
- Andreas T Schaefer, The Francis Crick Institute, United Kingdom
Reviewer
- Matthew N Zipple, Duke University, United States
Version history
- Received: January 20, 2020
- Accepted: August 18, 2020
- Version of Record published: September 29, 2020 (version 1)
Copyright
© 2020, Rozenkrantz et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.
Metrics
-
- 5,020
- Page views
-
- 265
- Downloads
-
- 7
- Citations
Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.
Download links
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Further reading
-
- Genetics and Genomics
- Neuroscience
A hexanucleotide repeat expansion in C9ORF72 is the most common genetic cause of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). A hallmark of ALS/FTD pathology is the presence of dipeptide repeat (DPR) proteins, produced from both sense GGGGCC (poly-GA, poly-GP, poly-GR) and antisense CCCCGG (poly-PR, poly-PG, poly-PA) transcripts. Translation of sense DPRs, such as poly-GA and poly-GR, depends on non-canonical (non-AUG) initiation codons. Here, we provide evidence for canonical AUG-dependent translation of two antisense DPRs, poly-PR and poly-PG. A single AUG is required for synthesis of poly-PR, one of the most toxic DPRs. Unexpectedly, we found redundancy between three AUG codons necessary for poly-PG translation. Further, the eukaryotic translation initiation factor 2D (EIF2D), which was previously implicated in sense DPR synthesis, is not required for AUG-dependent poly-PR or poly-PG translation, suggesting that distinct translation initiation factors control DPR synthesis from sense and antisense transcripts. Our findings on DPR synthesis from the C9ORF72 locus may be broadly applicable to many other nucleotide repeat expansion disorders.
-
- Cell Biology
- Neuroscience
The amyloid beta (Aβ) plaques found in Alzheimer’s disease (AD) patients’ brains contain collagens and are embedded extracellularly. Several collagens have been proposed to influence Aβ aggregate formation, yet their role in clearance is unknown. To investigate the potential role of collagens in forming and clearance of extracellular aggregates in vivo, we created a transgenic Caenorhabditis elegans strain that expresses and secretes human Aβ1-42. This secreted Aβ forms aggregates in two distinct places within the extracellular matrix. In a screen for extracellular human Aβ aggregation regulators, we identified different collagens to ameliorate or potentiate Aβ aggregation. We show that a disintegrin and metalloprotease a disintegrin and metalloprotease 2 (ADM-2), an ortholog of ADAM9, reduces the load of extracellular Aβ aggregates. ADM-2 is required and sufficient to remove the extracellular Aβ aggregates. Thus, we provide in vivo evidence of collagens essential for aggregate formation and metalloprotease participating in extracellular Aβ aggregate removal.