We discuss below the individual galaxies in more detail, based mainly on the new spectral data, and the observational data from the literature, appropriate for the present discussion. The main parameters of the studied galaxies are given in Table 4.

4.1 UM 151 = Mkn 1169

This galaxy does not look like a bona fide BCG. Its appearance resembles that of a face-on disk with somewhat disturbed outermost parts, but clearly without spiral arms. Salzer (1989a) classified it as a Dwarf Amorphous Nuclear Starburst (DANS). Its absolute magnitude quoted here (Table 4, $M_B=-17\fm96$ ) is within the range for DANS. A bright knot is seen near the galaxy center, but does not change the generally regular appearance of this galaxy. The low metallicity value from Telles (1996) was derived from a very low S/N spectrum (see Terlevich et al. 1991).

For the metallicity estimation both the Pilyugin (2001) calibration based on the strong oxygen lines and the recently reported calibration based on the [N II] line (Denicoló et al. 2002) give very close values of $\rm 12+log(O/H)$ , 8.50 and 8.47, respectively.

Balmer absorption lines and other features with reliable detections can be used to estimate the age of star-formation episodes.

The Equivalent Widths (EW) of Balmer absorption lines from the underlying continuum have been calculated following the prescription in González-Delgado et al. (1999) and the results are presented in Table 5. The comparison with the González-Delgado et al. (1999) models gives an age for the starburst consistent with an instantaneous starburst $\sim$ 10 Myr old. The EW of H $\beta$ emission is also consistent with the age of instantaneous starburst of $\sim$ 10 Myr (Leitherer et al. 1999). Following Raimann et al. (2000), we also measured the EW of the Ca II K-line and the G-band as well as the continuum flux ratios in the 3 bands (see Table 5). All but one of the parameters are consistent with a mixture of two starbursts with ages $\sim$ 10 Myr and a $\sim$ 50 Myr. The rather high EW of the G-band suggests an additional contribution of a stellar population with an age of a few hundred Myr.

Table 5: The absorption lines EW and continuum ratios of UM 151.
Abs. line Value Band Value

EW(H $_{\beta}$ ) 5.4 CaII-K 2.2

EW(H $_{\gamma}$ ) 4.0 G band 3.2

EW(H $_{\delta}$ ) 6.1 F₃₆₆₀/F₄₀₂₀ 0.71

EW(H₈) 7.4 F₃₇₈₀/F₄₀₂₀ 0.88

EW(H₉) 7.9 F₄₅₁₀/F₄₀₂₀ 0.83

EW(H₁₀) 4.0

**Table 5:** The absorption lines EW and continuum ratios of UM 151.
Abs. line	Value	Band	Value
EW(H $_{\beta}$ )	5.4	CaII-K	2.2
EW(H $_{\gamma}$ )	4.0	G band	3.2
EW(H $_{\delta}$ )	6.1	F₃₆₆₀/F₄₀₂₀	0.71
EW(H₈)	7.4	F₃₇₈₀/F₄₀₂₀	0.88
EW(H₉)	7.9	F₄₅₁₀/F₄₀₂₀	0.83
EW(H₁₀)	4.0

The known upper limit on H I flux of this galaxy (see Table 4) is rather high, so the upper limit on the ratio $M(H{\sc i})/L_B$ is consistent with the range typical of gas-rich starbursting galaxies.

4.2 UM 408

Appearing like a typical blue compact dwarf, this galaxy was classified by Salzer (1989a) as a Dwarf H II Hotspot Galaxy. The absolute B-band luminosity is M_B=-15.16 (Campos-Aguilar et al. 1993), almost at the lower end of the BCG luminosity distribution, with a quoted diameter of 2.1 kpc. The calculated metallicity reported by Masegosa et al. (1994) was 12 + log(O/H) = 7.63 using the data from the Spectroscopic Catalogue of H II Galaxies (Terlevich et al. 1991). Using the same set of data Telles (1996) reported a value of 12 + log(O/H) = 7.66. The new estimation with the present data suggests a significantly higher metallicity with a difference of 0.2 to 0.3 dex. This difference cannot be attributed to the reddening estimation or differences in slit positioning. Comparing the present data with Masegosa et al. (1994) and Telles (1996), the reddening coefficients are similar within the uncertainties and the values of both EW(H $\beta$ , emis) and integrated H $\beta$ flux are the same. Therefore the main reason for the discrepancy must be a poor estimation of the [O III] $\lambda$ 4363 due to low S/N. The comparison between the line intensities of both sets of data shows that a large difference is found not only in the faint [O III] line, but also in [O II] being larger by a factor of 2 for the TWIN data. The same is also true for the measured [N II] line. The low S/N on the continuum for this faint galaxy can account for the large difference in oxygen abundance.

The measured H I flux of this BCG (see Table 4) corresponds to a very high value of the parameter $M({\rm HI})/L_B$ . The $M({\rm HI})/L_B$ is comparable to values derived for the most extreme objects in the sample of dwarf galaxies with extended H I (van Zee et al. 1995).

4.3 A 1228+12 = RMB 132 = VCC 1313

This galaxy is one of the most compact, almost starlike in appearance, of the BCG family (Drinkwater & Hardy 1991). It resides in the Virgo cluster, and because of the surrounding environment, its properties are probably somewhat affected by more frequent interactions with surrounding galaxies and the hot intracluster medium (ICM). There is a number of sufficiently massive candidate galaxies in the vicinity of this BCG, which could trigger its current SF burst, including NGC 4478 at the projected distance of $18.7\arcmin$ ( $\sim$ 90 kpc) and M 87 at $20.8\arcmin$ ( $\sim$ 100 kpc), whose relative radial velocities are lower than 100 km s $^{\rm -1}$ .

Of the three galaxies studied in this paper, this is the faintest system with an absolute B magnitude of -14.1. Taking into account its compactness and luminosity, this is the type of galaxy classified as a Searle-Sargent object by Salzer (1989a), or H II galaxies by Campos-Aguilar et al. (1993). This galaxy was one of the first BCGs studied for metallicity purposes (Kinman & Davidson 1981, hereafter KD81). From this study the metallicity reported of $\rm 12+log(O/H)=7.64\pm0.07$ is in reasonable agreement with the present value of $7.73\pm0.06$ , based on the new, higher S/N data from the 6 m telescope. The agreement can be considered as a good one, taking into account that the areas sampled are 29 arcsec² for KD81 and 7.2 arcsec² in this work.

However, it is evident that the abundance calculations by KD81 could have some systematic differences, since these authors did not account for temperature gradients. They used one temperature for all zones, and noticed that this would lead to slight underestimation of oxygen and neon abundances. Therefore we recalculated oxygen and neon abundances from their relative intensities and the cited errors using our methodology. This resulted in an upward shift of 0.04 dex for the KD81 oxygen abundance, giving $\rm 12+log(O/H)=7.68\pm0.10$ and $\rm log(Ne/O)=-$ 0.75. The difference between the KD81 O/H abundance and the recalculated one appears to be from a somewhat larger abundance of ${\rm O}^{+}$ ions: $\rm 12+log(O^{+}/H)$ = 6.87 from KD81, and 7.05 in the model with the lower $T({\rm O}^{+})$ .

One of the interesting features in the spectrum of A 1228+12 is the appearance of an emission line centered at $\lambda$ 4591 Å, with flux $\sim$ (2.5 $\pm$ 1.2)% of H $\beta$ and FWHM= 28 Å. The only reasonable identification is as the Si III $\lambda$ 4565 line, characteristic of WR stars. The implied radial velocity of this feature is then $\sim$ 450 km s $^{\rm -1}$ higher than the system velocity of the galaxy found in other (narrow) emission lines. Given the large line width and low S/N ratio, the velocity shift is likely not significant. However the appearance of WR stars in the starburst region of this BCG would not be unexpected. Indeed, the observed EW of H $\beta$ (93 Å) according to Fig. 85e (corresponding to metallicity z= 0.001, the closest to the oxygen abundance of A 1228+12) from Starburst99 (Leitherer et al. 1999), corresponds to an instantaneous starburst of age 3-3.5 Myr. We adopted the Salpeter IMF with $M_{\rm low}=0.1~M$ $_{\odot}$ for this estimate. For the metallicity of z=0.001 this is exactly the age range where the models predict significant numbers of WR-stars (0.5 to 2.5% of the number of O-stars, see Schaerer & Vacca 1998). Since we did not detect other characteristic WR features (N III/N IV $\lambda$ 4640 and broad component of He II $\lambda$ 4686), that should have comparable EW, doubt about the reality of the Si III $\lambda$ 4565 line remains. Deeper spectroscopy is necessary in order to measure the strength of the probable WR features in this BCG.

The $M({\rm HI})/L_B$ ratio for this BCG is quite large (see Table 4). Accounting for very large EW of strong emission lines and respective significant brightening ( $\Delta B \gtrsim ~1\fm0$ ), this implies a large gas mass-fraction. The latter is difficult to understand for a low-mass galaxy affected both by the ICM ram pressure (unless the BCG is only now entering the hot gas) and tidal interactions from massive neighbours.

In fact, the very existence of such a low-metallicity BCG in a dense environment such as the Virgo Cluster (despite the opposite tendency of additional enrichment by heavy elements for the Virgo cluster BCGs, noticed by Izotov & Guseva 1989), poses interesting questions on the evolutionary history of this and similar objects. It is worth mentioning that there are two more BCGs with well determined low metallicities ( $Z\sim Z_{\odot}/20$ ) probably belonging to the Virgo Cluster or its outskirts: the Optical Counterpart of HI 1225+01 (Salzer et al. 1991) and VV 432=IC 3105=VCC 241 (Zasov et al. 2000). Less intriguing, but also not well understood, is the appearance of SBS 0335-052 (E+W), the pair of extremely metal-deficient BCGs, situated at the outskirts of the loose galaxy group LGG 103 (Pustilnik et al. 2001; Peebles 2001).

4 Discussion and summary

4.1 UM 151 = Mkn 1169

4.2 UM 408

4.3 A 1228+12 = RMB 132 = VCC 1313

4.4 Conclusions